«Открытый ИИ — это пузырь»! Откровения из Силиконовой долины | Братья Либерман

Что произойдет с миром, если алгоритмы начнут управлять людьми? Можно ли сохранить свободу, когда решения принимают машины? Может ли искусственный интеллект стать новой формой власти? Скоро ли настанет будущее, когда роботы будут работать за и на человека? И почему в мире, где растет зависимость от данных и кода, главным капиталом снова становится человек?

*https://www.youtube.com/watch?v=wp7izqZmiWM
**https://300.ya.ru/v_OOTQyGao

таймкоды

00:00:00 Введение и контекст

• Обсуждение бума в индустрии и предсказаний о больших потрясениях.
• Упоминание о росте доходов компании с миллиарда до пятнадцати миллиардов долларов за год.

00:00:25 Технологии и общество

• В Албании анонсировали доступ к суперинтеллекту для каждого человека.
• Технологии меняют жизнь, но распределены неравномерно.
• Риски утечки личной информации и роста безработицы.

00:00:58 Представление гостей

• Давид и Даниил Либерманы — основатели компаний, продавших одну из них Snapchat за 60 миллионов долларов.
• Создание фонда, инвестирующего в людей и их продукты.

00:01:37 Мнения о будущем

• Различия в мнениях о будущем: от автоматизации до риска краха, как в доткоме 2000 года.
• Сравнение текущего рынка с доткомом 2000 года.

00:03:00 Риски и примеры

• Примеры мошенничества в доткоме 2000 года, включая компании, притворявшиеся Amazon и Google.
• Текущие компании, такие как Amazon, Google, Facebook и Microsoft, продолжают зарабатывать.

00:04:30 Ожидания и реальность

• Ожидания быстрого развития технологий часто не оправдываются.
• Пример с ожиданием появления AI через год.

00:05:58 Инвестиции и риски

• Большие инвестиции в новые компании, многие из которых могут не оправдаться.
• Риск схлопывания компаний из-за огромных расходов.

00:07:12 Долгосрочная перспектива

• В долгосрочной перспективе большинство инвестиций оправдываются.
• Статистически успешные компании создают значительные ценности.

00:08:56 Пример компании

• Компания с миллиардом долларов дохода за год достигла пятнадцати миллиардов.
• Её продукты влияют на крупнейшие компании.

00:09:41 Стоимость талантов

• Таланты могут стоить миллиарды долларов, что меняет рынок.
• Эксперты понимают, во что инвестировать, но остальные рискуют.

00:10:28 Аккредитив для бизнеса

• Аккредитив от Сбера позволяет безопасно совершать сделки, не завися от добросовестности контрагента.
• С помощью аккредитива можно покупать оборудование, недвижимость, доли компании или интеллектуальную собственность.
• Банк проверяет документы и переводит средства поставщику только после подтверждения права собственности.

00:11:25 Преимущества аккредитива

• Средства замораживаются на защищённом счёте и переводятся только после проверки документов.
• Если сделка сорвётся, средства вернутся на счёт.
• Аккредитив упрощает сотрудничество с новыми контрагентами и увеличивает прибыль.

00:12:04 Концентрация венчурного капитала в Силиконовой долине

• В Силиконовой долине наблюдается высокая концентрация венчурного капитала в нескольких фондах.
• Инвесторы вкладывают средства в различные стартапы, включая компании, создающие децентрализованные технологии.
• Общая совокупная стоимость рынка потенциально высока и оправдывает инвестиции.

00:13:35 Риски фондового рынка

• Фондовый рынок растёт на ожиданиях повышения эффективности компаний.
• Существует риск обвала рынка, если ожидания не оправдаются.
• Рост фондового рынка сопровождается ростом цен на золото и биткоин, что вызывает опасения у инвесторов.

00:16:11 Стратегии инвестиций в дата-центры

• Многие компании вкладывают средства в строительство дата-центров с устаревшим оборудованием.
• Пример с биткоином показывает, что инвестиции в инновации могут быть более эффективными, чем в текущее оборудование.
• Борьба за доминирование на рынке дата-центров может привести к замедлению прогресса.

00:18:25 Последствия переоценённости рынка

• Переоценённость рынка может привести к замедлению развития технологий.
• Деньги перераспределяются между участниками рынка, некоторые теряют, другие выигрывают.
• Кризис может временно замедлить рост, но рынок вернётся к здоровому состоянию через несколько лет.

00:20:49 Важность экспертизы в инвестициях

• Для успешных инвестиций необходима экспертиза и умение отличать реальные инновации от пузырей.
• Технологии меняют жизнь и профессии, поэтому важно не упустить возможность инвестировать в перспективные проекты.

00:21:49 Конкуренция между США и Китаем в области технологий

• США являются бенефициарами технологического бума.
• Китай активно развивает свои технологии, но информация о них ограничена.
• США пытаются переманивать китайские кадры.

00:22:45 Влияние на другие страны

• Появление локальных поисковых систем, таких как Яндекс в России, иллюстрирует возможность конкуренции.
• Однако для создания собственных технологий требуется значительный объём ресурсов.
• Существует риск доминирования одного гиганта, который сможет контролировать рынок.

00:23:32 Анализ рынка и геополитические решения

• Три года назад был проведён анализ рынка технологий.
• В 2020 году было заключено соглашение между Microsoft, Meta, Илоном Маском и NVIDIA о разделении рынка GPU.
• США ввели лицензирование для моделей с более чем триллионом параметров.

00:26:02 Изменение баланса сил

• Два-три года назад США и Китай имели паритет в вычислительных мощностях.
• Сейчас у США 90% вычислительных мощностей.

00:28:55 Роль NVIDIA и стандарты

• NVIDIA создала стандарт для вычислений, который определяет производительность чипов.
• Этот стандарт позволяет контролировать рынок вычислительных мощностей.
• Без этого стандарта многие страны могли бы иметь альтернативы.

00:30:46 Переговоры и ограничения

• Переговоры между США и другими странами включают ограничения на покупку GPU.
• Ограничения кажутся незначительными, но они создают сложности для других стран.

00:32:02 Пример DeepMind

• DeepMind потратил два миллиарда долларов на создание своей модели.
• Успешные трейдеры, такие как Алекс Герка, используют большое количество GPU для своих операций.
• Эти примеры показывают, что для создания мощных моделей требуются значительные ресурсы.

00:33:37 Китайские технологии и чипы

• Китайские компании инвестируют миллиарды долларов в разработку технологий.
• Несмотря на ограничения в покупке чипов, Китай пытается обойти санкции.
• 90% чипов остаются у американских компаний.

00:34:51 Монополия и альтернативный путь

• Существует риск формирования монополии одной страны на рынке чипов.
• Альтернативный путь — децентрализация и опенсорс, который может превзойти текущие стандарты.

00:36:00 Биткоин как инфраструктурный проект

• Биткоин рассматривается как финансовый инструмент, но его потенциал шире.
• Биткоин может служить финансовым фреймворком для финансирования инфраструктуры.
• Распределённая инфраструктура биткоина достигает 26 гигаватт.

00:40:05 Инновации в майнинге биткоина

• Майнинг биткоина характеризуется быстрой инновацией оборудования.
• Потребление энергии для вычислений биткоина снизилось в 300 тысяч раз.
• Специализированные майнинговые компании выпускают обновления каждые полгода.

00:42:03 Философия децентрализации

• Децентрализованная модель позволяет каждому участнику вносить свой вклад.
• Централизованные компании могут попасть под контроль государства.
• Альтернатива — модель, где каждый может улучшить оборудование и заработать на этом.

00:44:27 Экономические аспекты децентрализации

• Текущие затраты на оборудование для майнинга высоки и недоступны большинству.
• Увеличение производительности чипов в 100 тысяч раз может значительно снизить затраты.
• Децентрализованная модель может сделать оборудование более доступным.

00:45:42 Стабильность биткоин-сети

• Увеличение количества дата центров и вычислений не изменило суть биткоина.
• Количество биткоинов и транзакций в секунду остаётся стабильным.

00:45:58 Децентрализованные вычисления и суперинтеллект

• Децентрализованные вычисления могут сделать суперинтеллект доступным каждому человеку.
• Инициативы по децентрализованным вычислениям уже реализуются в виде стартапов.
• Пример: взаимодействие Павла Дурова с президентом Казахстана.

00:46:54 Изменение восприятия децентрализации

• Ранее считалось, что децентрализация неэффективна для вычислений.
• Теперь признано, что децентрализация хорошо работает для определённых задач.
• Социальные приложения обеспокоены феноменом ChatGPT.

00:47:30 Влияние ChatGPT на социальные приложения

• ChatGPT пока не имеет социальных инструментов, но уже привлекает много пользователей.
• Приложение Sora с социальной функцией быстро распространяется.
• Meta выпустила свой аналог, но он не получил широкого признания.

00:50:13 Экономические аспекты мессенджеров

• Мессенджеры боятся тратить много денег на улучшение ботов и моделей.
• Модель заработка на пользователях меняется: пользователи платят за дополнительные функции.
• Инициатива Дурова и Маска по улучшению Telegram провалилась.

00:52:01 Будущее децентрализованных вычислений

• Децентрализованные вычисления могут стать дешёвыми и доступными для всех.
• Крупные компании доминируют благодаря большим ресурсам графических процессоров.
• Необходимость нового подхода к разработке продуктов.

00:54:19 Роль GPU и децентрализованных протоколов

• GPU необходимы для обработки больших объёмов данных.
• Казахстан и другие страны имеют ресурсы GPU, но не знают, как их использовать.
• Требуется свободный протокол для совместного использования ресурсов.

00:56:15 Финансовый фреймворк для децентрализованных систем

• Нужен финансовый фреймворк, аналогичный биткоину, для мотивации участников.
• Монета протокола может служить субсидией для строительства инфраструктуры.
• Ожидание доли от рынка как стимул для развития.

00:57:27 Реальность реализации децентрализованных идей

• Для реализации децентрализованных идей нужен мощный катализатор.
• Люди должны осознать проблематику и финансовую модель на собственном опыте.
• Пока неясно, как эти идеи могут реализоваться в реальности.

00:58:53 Запуск проекта

• Проект разрабатывался два года и был запущен полтора месяца назад.
• За месяц количество участников выросло с 80 до 450.

00:59:23 Финансирование и репутация

• Три года назад проект получил 18 миллионов долларов в первом раунде финансирования.
• После этого раунда новых инвестиций не было.

01:00:29 Оптимизация кода

• Продукт помогает находить ошибки и оптимизировать код с помощью искусственного интеллекта.
• Он ускоряет работу приложений за счёт оптимизации кода.

01:01:45 Создание децентрализованной сети

• Два года назад решено создать децентрализованный протокол.
• Проект не привлекал инвестиции, чтобы избежать проблем с регуляцией.

01:03:43 Открытый код

• Протокол полностью передан в открытый доступ.
• Развитие сети контролируется сообществом.

01:06:02 Сравнение с биткоином

• Технология может вырасти быстрее, чем биткоин, из-за повторения тренда.
• Пять компаний и несколько китайских компаний захватывают рынок.

01:08:16 Отказ от инвестиций

• За последние четыре месяца было два предложения о покупке проекта более чем за 100 миллионов долларов.
• Философия проекта направлена на создание социальных инноваций и будущего изобилия.

01:09:48 Перспективы

• Продукт может стать востребованным через год-два.
• Возможно, ведущие компании придут покупать продукт позже.

01:09:57 Конкуренция и тренды в IT

• Крупные лаборатории быстро создают альтернативы, вытесняя стартапы.
• Пример: OpenAI выпускает новые продукты, захватывая рынок агентов и генерации видео.
• Важно не только заработать деньги, но и повлиять на рынок.

01:10:57 Стратегия и перспективы

• Создание децентрализованной альтернативы может принести сотни миллиардов долларов.
• Сравнение с биткоином: возможность заработать больше, чем в Силиконовой долине.

01:12:34 Модель протокола и её отличия

• В биткоине 99% вычислительных мощностей тратятся на безопасность сети, а не на транзакции.
• Новый протокол использует 99% мощностей для полезных вычислений, а 1% — для безопасности.
• Полезное действие в протоколе — вычисление и тренировка LLM-моделей.

01:14:04 Принцип работы протокола

• Пользователи подключают свои машины к сети и участвуют в вычислениях.
• Машины образуют единый сервер для вычислений, зарабатывая монеты за операции.
• Разработчики загружают вычисления в распределённую сеть.

01:15:33 Безопасность и квантизация

• Протокол обеспечивает безопасность ответов, проверяя степень квантизации моделей.
• Попытка мошенничества приводит к потере вознаграждения.

01:16:50 Примеры успешных проектов

• Filecoin хранит 10 миллионов терабайт данных.
• Render.io использует распределённые серверы для рендеринга мультиков.
• Эти проекты занимают нишевые рынки, но имеют потенциал для роста.

01:18:23 Перспективы рынка ИИ-вычислений

• Рынок ИИ-вычислений сегодня составляет 40 миллиардов долларов, а через 10 лет может достичь 2 триллионов долларов.
• Децентрализованные сети имеют шанс занять значительную долю рынка.

01:20:15 Риски и перспективы для альтернативных решений

• Большие лаборатории могут монополизировать рынок.
• Спрос на альтернативы будет расти из-за риска монополизации.
• Компании, такие как Canva и Flo, будут использовать децентрализованные решения.

01:21:56 Адаптация и долгосрочные перспективы

• Адаптация к новым технологиям займёт 5–10 лет.
• После адаптации эффект на рынок будет колоссальным.
• Пример самоуправляемых машин показывает экспоненциальный рост технологий.

01:22:23 Риски контроля в майнинге

• Обсуждение количества GPU в системе: 450 высокопроизводительных GPU.
• Опасность контроля со стороны крупных компаний: возможность перенаправления вычислений на чужие мощности.
• Подчёркивание важности масштаба для защиты от контроля.

01:23:20 Пример биткоина

• Вопрос о возможности банков контролировать 51% биткоина в 2013 году.
• Объяснение, почему биткоин не был контролирован крупными игроками.

01:23:40 Распространение знаний и рост мощности

• Роль распространения знаний в увеличении мощности сети.
• Высокая стоимость оборудования для майнинга: около миллиона долларов в месяц.
• Аренда GPU в облаке по цене около 2 долларов в час.

01:25:25 Инвестиции и рост

• Привлечение инвесторов и рост мощности за месяц.
• Естественный рост числа участников и обеспечение безопасности сети.

01:27:11 Гуманистические принципы биткоина

• Концепция биткоина основана на предположении о честности большинства участников.
• Энтузиасты, строящие сети, стремятся к безопасности и будущему.

01:29:00 Прогнозируемый рост

• Ожидание удвоения мощности каждый месяц.
• Прогноз достижения 100 тысяч GPU менее чем за год.

01:31:08 Будущее майнинга

• Переход от домашних компьютеров к специализированным чипам.
• Роль больших компаний и фондов в майнинге.
• Сравнение с добычей золота: предоплата за будущие ресурсы.

01:33:04 Другие проекты

• Обсуждение других проектов компании, включая фонд и оптимизацию приложений для корпораций.
• Увеличение оценки холдинговой компании с 200 до 400 миллионов долларов.
• Продолжение инвестиций в людей и ожидание выхода из инвестиций через 8 лет.

01:35:14 Инвестиционный проект

• Успехи фаундеров, в которых вложили средства.
• Цель — реализовать доли в будущем фаундеров за значительно большие деньги.
• Дивиденды составляют лишь небольшую часть стоимости.

01:36:08 Взаимодействие с другими компаниями

• Взаимодействие с людьми Питера Диамандиса, создателя Singularity University.
• Запуск нового проекта с использованием инвестиционной модели.

01:37:09 Долгосрочная стратегия

• Необходимость трекшна для привлечения новых фондов.
• Долгосрочная игра на рынке, терпение и ожидание развития рынка.

01:38:07 Биотех и квантовые компьютеры

• Использование открытий родителей для создания квантовых компьютеров.
• Долгосрочные эксперименты и подготовка технологической базы.

01:39:02 Риски квантовых компьютеров для криптовалют

• Технологии шифрования будут адаптироваться к появлению квантовых компьютеров.
• Биткоин уже адаптируется к изменениям, например, введение дополнительного типа шифрования.

01:42:52 Проблемы экономики и искусственный интеллект

• Ухудшение ситуации с пенсионными фондами и мотивацией молодёжи.
• Угроза искусственного интеллекта и рост безработицы.

01:44:42 Сценарии развития ИИ

• Два сценария развития ИИ: быстрое и медленное.
• В медленном сценарии ИИ увеличивает производительность, но доступ к нему будет зависеть от социального статуса.
• В быстром сценарии ИИ может стать инструментом, доступным лишь немногим.

01:47:17 Будущее с ИИ

• ИИ как инструмент, выполняющий команды, а не порабощающий человечество.
• ИИ обладает всеми знаниями человечества и способностью рационально мыслить.

01:47:28 Конкуренция в производстве роботов

• Роботы могут выполнять любые задачи с разной степенью точности.
• Стоимость роботов зависит от их функциональности: от миллиона до десяти тысяч долларов.
• Существует множество компаний, производящих роботов, что создаёт высокую конкуренцию.

01:47:57 Скорость разработки роботов

• Прототипы роботов создаются за семь месяцев.
• Компании, работающие над роботами, сталкиваются с трудностями и неудачами.
• В базе данных одного из разработчиков более пятисот компаний, производящих роботов.

01:48:23 Влияние доступности технологий на рынок

• Если технологии будут доступны всем, конкуренция станет безумной.
• Пример с интернетом и телефонами иллюстрирует, как доступность технологий может изменить рынок.
• В мире, где роботы стоят пять тысяч долларов, миллиард человек может их себе позволить.

01:49:17 Роль роботов в повседневной жизни

• Роботы могут выполнять различные задачи: готовить еду, работать на заводе, регулировать перекрёстки.
• Они могут работать 20 часов в день, заряжаясь четыре часа.
• Роботы могут распределять свою работу между домом, работой и общественными целями.

01:50:12 Влияние на безработицу

• Роботы могут снизить уровень безработицы, выполняя многие задачи.
• Однако это возможно только при условии, что «мозги» роботов не контролируются одной компанией.
• Пример с роботами от Google показывает, как одна компания может контролировать рынок.

01:51:01 Модели самоуправляемых машин

• Uber не смог построить самоуправляемые машины из-за судебного иска от Google.
• Илон Маск предлагает модель, где каждая машина управляется сетью, но пользователь не владеет машиной.
• Маск планирует использовать доходы от роботов-такси для финансирования проекта по полёту на Марс.

01:52:42 Будущее с децентрализацией

• Децентрализованная сеть с открытым кодом может изменить будущее безработицы.
• В будущем каждый человек сможет иметь робота, способного выполнить любую работу.
• Экономика изменится, так как роботы будут выполнять множество задач.

01:53:33 Адаптация к изменениям

• В начале внедрения роботов может быть расслоение общества и рост безработицы.
• В США доходы от роботов будут распределяться через пособия, но это может повлиять на суверенитет стран.
• В некоторых странах, например, в Саудовской Аравии, пенсионные фонды инвестируют в опенсорс-роботов.

01:57:23 Философские вопросы будущего

• Главная проблема будущего — смысл жизни и выбор задач для роботов.
• Децентрализованная модель противоречит текущей структуре власти и доминированию.
• Идея децентрализации меняет ход истории и правила игры.

01:58:56 Эволюция операционных систем

• В конце 90-х годов серверы управлялись частными операционными системами, такими как Microsoft.
• К 2001 году Linux занимал менее 1% рынка серверов, а сегодня — 75%.
• Android-телефоны работают на Linux, а все интернет-протоколы и серверные языки основаны на открытом коде.

02:01:09 Децентрализация и облака

• Облачные вычисления не стали децентрализованными из-за отсутствия фреймворка, аналогичного биткоину.
• Биткоин построил более крупную инфраструктуру, чем облачные сервисы.

02:02:24 Технические аспекты биткоина

• Текущее железо не подходит для вычислений в биткоине.
• Специализированные чипы для биткоина становятся всё более эффективными.
• Ожидается значительное увеличение скорости вычислений в биткоине.

02:05:00 Будущее биткоина и блокчейнов

• Биткоин остаётся основой децентрализованных блокчейнов.
• Возможно появление новых протоколов, не претендующих на роль «золота».
• Альтернативные проекты должны пересмотреть свою структуру и перейти на Proof of Stake.

02:07:34 Пример Бутана

• Бутан использует гидроэлектроэнергию для майнинга биткоинов.
• Страна подписала договор с Bitmain и импортировала специализированные чипы.
• Бутан планирует развивать международную кооперацию в области блокчейнов.

02:11:08 Перспективы будущего

• В ближайшие 5–10 лет мир ждёт значительных изменений и потрясений.
• Децентрализованные технологии могут значительно улучшить благосостояние людей.
• Даже те, кто сейчас находится за пределами рынка, могут достичь уровня жизни развитых стран.

02:12:30 Влияние роботов на общество

• Роботы могут значительно улучшить жизнь людей, особенно в развивающихся странах.
• Доступность роботов будет варьироваться: у богатых — дорогие модели, у бедных — более доступные.
• Роботы изменят способы заработка, выполнения работы и создания вещей.

02:13:11 Неравномерное распределение будущего

• Будущее уже здесь, но оно будет распределено неравномерно.
• В некоторых странах будут потрясения, в других — рост и изобилие.
• Каждому человеку важно задуматься о своём месте в мире, где роботы могут производить всё лучше.

02:14:10 Вопросы о будущем

• Вопрос о том, станет ли ИИ умнее человека, рассматривается как победа.
• Желание увидеть, как далеко зашло человечество, и возможность посетить будущее на 10 тысяч лет вперёд.

02:15:03 Влияние ИИ на науку

• ИИ влияет на биохимию, химию, физику и создание новых материалов.
• В ближайшие два года ИИ изменит производство лекарств, материалов и батареек.
• Ведущие химические и фармацевтические компании уже начинают заказывать новые молекулы у ИИ-компаний.

02:15:57 Навыки предпринимателя будущего

• Главный навык предпринимателя — умение брать на себя ответственность и существовать в неопределённости.
• Необходимость принимать решения в условиях, когда сценарии развития событий непредсказуемы.
• Умение адаптироваться к неопределённости — ключевая часть предпринимательства.

02:17:27 Заключение

• Обсуждение изменений в мире сильно влияет на восприятие действительности.
• Вопрос о будущем мира — самый главный для адекватных людей.
• Благодарность за интересный разговор и призыв к подписке на канал.

В этом видео

Поисковая расшифровка
Chapter 1: Trailer
0:00
An absolutely crazy boom. This will be a world of great upheaval. We gathered in the courtyard somewhere near Altman’s house. We’ll place bets on when it will happen. Most are honest,
0:09
9 seconds
go mine. This market is about 98% of them having a billion dollars in revenue a year ago. Now
0:16
16 seconds
they have 15 billion dollars. I’ve seen Nvidia graphics. Oh, the Chinese. The Chinese, and it died out. Well, it died out. And why Kazakhstan? Wait, wait. We’re sliding towards a world that is completely different from the one we would like to
0:25
25 seconds
live in. Just recently, in Albania, they announced that they will have a minister. Superintendent will be available to everyone . Zero talk. Sounds great. The future is already here, but it is unevenly distributed. A robot
0:35
35 seconds
for every person. This is your robot. It goes to work at the factory for you. Sounds provocative, but interesting. You’re corresponding today, and you hope no one reads it. And
0:43
43 seconds
they’ll read it. It’s funny. Your names, Elon’s correspondence, are in court documents. Not funny. These networks are actually built by enthusiasts. It’s all already gold. Everything has already been sawed up and divided. Huge
0:53
53 seconds
social stratification, growing unemployment, and so on. He’s changing our lives right now. Friends, hello everyone. A few facts about my guests today. Ah, they’ve been building companies for over 15 years.
Chapter 2: David and Danil Lieberman
1:04
1 minute, 4 seconds
They sold one of their companies to Snapchat for $60 million. For their next company, after selling Product Science, they raised $18 million. They’ve also been building a fund for several years now
1:15
1 minute, 15 seconds
that invests not just in companies, but in people and, accordingly, receives income from all the company’s products that these people have created. In fact, this is only part of
1:25
1 minute, 25 seconds
what they do. Plus, we already recorded a podcast a year and a half ago, and it has already garnered almost 1.5 million views. Actually, David and Daniil Liberman, please welcome. Hello, gentlemen.
1:36
1 minute, 36 seconds
Hello. Welcome. What’s interesting? Last time, we recorded an episode and talked about
1:42
1 minute, 42 seconds
how our world would change, how the economy would change, if artificial intelligence really became a part of our lives. Literally a month passes, probably, and an
1:52
1 minute, 52 seconds
absolutely crazy boom begins, which we are still witnessing today. And opinions differ. Some people say our world will never be the same. Basically, everything will be automated, technologically advanced, and so
Chapter 3: Firsthand, what’s really going on in the AI world
2:02
2 minutes, 2 seconds
on. There are a huge number of skeptics who say it will all collapse like the dot-com boom in 1999. And in fact, only a small fraction of what we see today will remain.
2:11
2 minutes, 11 seconds
And all this is being inflated so that venture investors and banks can simply make money at this point in time. You are at the very epicenter, so to speak, of events.
2:20
2 minutes, 20 seconds
Firstly, it’s the valley where all this is happening. Secondly, you yourself are actively involved in product development. And, in fact, you have a huge circle of contacts. Let’s try to figure out what
2:29
2 minutes, 29 seconds
we actually have now, in your opinion. How much truth is there about a paradigm shift and
2:35
2 minutes, 35 seconds
a new future, and how much is there to the idea that all of this will depreciate and only a small part will remain? On the one hand, if you look at the chart of the 2000 bubble, you’ll see that
2:46
2 minutes, 46 seconds
it no longer looks like a bubble, because the current market is an order of magnitude larger than what existed then and was considered overly inflated. Oh, Amazon was worth a lot less back then than it is now.
3:00
3 minutes
Well, really, it subsequently fell by 98%, right? Look, maybe we can
3:04
3 minutes, 4 seconds
look at Bitcoin for a slightly more direct picture, one that would be more understandable to all of us. When this
3:12
3 minutes, 12 seconds
excitement arises, wow, excitement, then at that moment money can be attracted to projects that are nothing special
3:19
3 minutes, 19 seconds
, because it’s very difficult to distinguish, yes, they are still very small, they all look
3:24
3 minutes, 24 seconds
like those that can become these giants, but it’s very difficult to distinguish. And what happened in 2000 , there were several pure frauds. Literally senators participated in this process.
3:36
3 minutes, 36 seconds
Worldcom and Enron. These were companies that pretended to be Amazon and Google and were fooled into
3:46
3 minutes, 46 seconds
inflating their valuations. And then it all collapsed when their documentation, internal financial documentation, suddenly surfaced somewhere. When was this visible? This was in 2000.
3:57
3 minutes, 57 seconds
Yeah. There were even cases of outright fraud that spread everywhere. But now we see
4:05
4 minutes, 5 seconds
that not all of this was pure money, because Amazon, Google, Facebook, Microsoft, all that Anything worth these trillions is now making money. Back then, this boom was based on certain expectations.
4:17
4 minutes, 17 seconds
For example, it was based on the expectation that most products would be sold online. Back then, a very small percentage of products were sold online. Ah, but it was this expectation that gave rise to these gigantic
4:28
4 minutes, 28 seconds
valuations. Now, most products are sold online. It happened. It’s been 25 years,
4:34
4 minutes, 34 seconds
right? It always takes longer. That’s precisely the main reason for the bubble,
4:40
4 minutes, 40 seconds
because this excitement is based on the idea that everything will happen tomorrow, yes, just like people usually overestimate how much we can do in a year and underestimate what we can do in 10. Yes, something like that. Yes.
4:49
4 minutes, 49 seconds
And this will definitely happen with AI too. Ah, there were a lot of statements that AGI would happen
4:55
4 minutes, 55 seconds
in a year, and so on. It didn’t happen through HI—it’s this general artificial intelligence, yes, we’ll talk about that a little later. This is the ultimate goal of the majority and
5:06
5 minutes, 6 seconds
companies right now—to create this universal, human-superior artificial intelligence.
5:14
5 minutes, 14 seconds
However, those who look more closely can no longer say that this is an unfounded
5:24
5 minutes, 24 seconds
expectation. I mean, the AI ​​bubble isn’t just a bubble. It’s clear where it really works.
5:32
5 minutes, 32 seconds
And what’s more, people are paying. The end user is already paying those same $20 a month because they clearly understand how it changes their lives. Well, look, let me
5:41
5 minutes, 41 seconds
throw out some questions that are already swirling around in my head. For example, I looked at
5:45
5 minutes, 45 seconds
the forecasts of that same Open AI, and they’re planning losses of
5:53
5 minutes, 53 seconds
$112 billion by 2030, I think, yes. Accordingly, that’s a colossal amount of money. This is just the tip of the iceberg. If we look now, well, you’ll see all this already in the investments
Chapter 4: What will happen to the numerous AI startups?
6:03
6 minutes, 3 seconds
that are happening, how they’re hunting people for billions of dollars. These are simply absolutely incredible figures: a new company is opening, with seemingly endless amounts of money being poured into it. And so I
6:13
6 minutes, 13 seconds
immediately have a question. Of course, among the hundreds of these companies that have emerged, there will be these diamonds in the rough that will, in fact, become profitable in 10 years, earn
6:22
6 minutes, 22 seconds
trillions, change the world, and so on, just like Amazon and Microsoft and other companies. But I have a feeling that 98% of what we’re seeing now is happening to the same thing, as
6:32
6 minutes, 32 seconds
you mentioned with the examples of Amazon duplicates and the like. Moreover, I personally have a question: when a company like Openi, for example, is pumped with so much money, does anyone even admit that this could simply be a mathematical implosion, that it might all just…
6:45
6 minutes, 45 seconds
well, the company won’t break even, it will incur huge expenses, and I don’t know, there will also be problems with electricity and with servicing all of this. And what will happen then? This market, in general,
6:56
6 minutes, 56 seconds
is 98% dying and will continue to die, as it has been for the last 30 years. That’s precisely why it’s so difficult to understand
7:04
7 minutes, 4 seconds
from the outside, because it seems like everyone made a mistake, investing in those 98
7:10
7 minutes, 10 seconds
companies out of the 100 that died. But if you look at the long-term outlook
7:15
7 minutes, 15 seconds
and try to calculate all the money invested in ventures over that time, and then look at how much value was actually created, how much was returned to the investor, what was it?
7:27
7 minutes, 27 seconds
5 times the amount invested, the total amount of results achieved. Results achieved. And
7:32
7 minutes, 32 seconds
in reality, it’s normal that 98% die. And the main thing is that even if you try, you won’t
7:39
7 minutes, 39 seconds
be able to distinguish which of them will actually become this huge giant. Whether it will be Open AI in the end or not, no one knows. Could it be that Open AI will disappear?
7:52
7 minutes, 52 seconds
Maybe, but this thesis is still based on the fact that, on average, all these attempts are justified by
8:01
8 minutes, 1 second
the statistically significant creation of such large scores. That is, it’s no longer possible to look at companies
8:07
8 minutes, 7 seconds
like Google or Microsoft or Amazon as simply inflated cash. They have a capitalization
8:16
8 minutes, 16 seconds
ratio of capitalization to their actual profit. This ratio is lower than Walmart’s. Yeah.
8:24
8 minutes, 24 seconds
So, they want companies that grow much faster than them, because these are already companies
8:32
8 minutes, 32 seconds
that continuously generate a very large amount of profit. Yeah. Net profit. And
8:40
8 minutes, 40 seconds
they justify all the investments made in a new company. And in this regard, yes,
8:51
8 minutes, 51 seconds
You might think that PNAI is highly valued, but a year ago
8:58
8 minutes, 58 seconds
they had a billion dollars in revenue, now they have $15 billion in revenue.
9:06
9 minutes, 6 seconds
You don’t see this kind of speed; no other company has ever experienced this kind of growth. It’s constantly at the top of the
9:18
9 minutes, 18 seconds
Android and Apple app stores. And, in fact, every new product release shakes things up.
9:30
9 minutes, 30 seconds
All the largest companies don’t know what to do with it anymore. And so the hype around
9:37
9 minutes, 37 seconds
these companies and the hype around the reality of talent—you know, we’re big fans of the value of talent. So, the fact that it’s now become the norm
9:48
9 minutes, 48 seconds
that talent can be purchased for $100 million a year in salary or billions
9:53
9 minutes, 53 seconds
is changing the market entirely. Yes, but considering what they’re creating, what they’re creating uniquely—it’s not like
10:00
10 minutes
everyone can create such systems—it absolutely justifies the investment
10:11
10 minutes, 11 seconds
. Even though it’s true, it’s incredibly scary for someone who’s not familiar with the market. Well,
10:17
10 minutes, 17 seconds
because it’s incredibly scary to be wrong 98% of the time. There are experts who understand what to
10:24
10 minutes, 24 seconds
invest in, while everyone else understands that if they don’t jump in now, they’ll miss this wave and potentially miss everything they already have. They jump in, but with less
10:34
10 minutes, 34 seconds
professionalism. Any entrepreneur knows that risks are commonplace in business. Buying equipment, goods, and services for a business is always stressful. Everyone
10:44
10 minutes, 44 seconds
keeps telling you, «Check the supplier, don’t transfer money upfront, be careful.» And you’re like, «I’ll decide for myself.» But sometimes, deciding for myself can lead to very unpleasant consequences. You transfer hundreds
10:55
10 minutes, 55 seconds
of thousands and in a second you’re left with nothing if the supplier suddenly turns out to be a fraud. To secure transactions that don’t depend on the counterparty’s integrity, there’s a reliable tool from Sberbank:
11:05
11 minutes, 5 seconds
a letter of credit for business. It allows you to securely conduct various transactions: from purchasing equipment and real estate to company shares or intellectual property.
11:14
11 minutes, 14 seconds
Sber annually processes up to 4 trillion of these secure transactions for its clients. It works simply. For example, you decide to buy equipment secondhand. You don’t need to transfer money
11:24
11 minutes, 24 seconds
directly to the supplier. You register for the service in the SberBusiness app and deposit funds into a secure account. The bank freezes the funds and transfers them to the supplier only when you become the full
11:34
11 minutes, 34 seconds
owner of the goods. All documents will be verified by the bank, so you don’t risk losing money or equipment and can complete the transaction with peace of mind. And if the transaction falls through, the funds will be returned to
11:43
11 minutes, 43 seconds
your account. The bank is responsible for reliability, making it easier for partners to decide whether to cooperate, especially if you have a young business or are entering into a new business. The result: more transactions,
11:54
11 minutes, 54 seconds
less downtime, and higher profits. If you are planning a major purchase for your business or entering into a new business, follow the link in the description or scan the QR code on the screen. Let’s get started.
Chapter 5: What has changed in Silicon Valley with the advent of AI?
12:04
12 minutes, 4 seconds
Why is Silicon Valley currently experiencing the largest concentration of capital in the history of venture capital, spread across the smallest number of funds? Yeah. A huge concentration.
12:18
12 minutes, 18 seconds
Because, in reality, this large amount of outside capital from venture capitalists is already accustomed to them dropping in, dropping in, and constantly dropping in awkwardly. But here, okay, okay,
12:28
12 minutes, 28 seconds
these five are such experts that we trust them and pour all our capital into them. Moreover, if
12:34
12 minutes, 34 seconds
you look at funds like Andreon Horowitz, they’re literally like an index now; they’ve invested
12:44
12 minutes, 44 seconds
in every company you know, although previously it was considered wrong that you
12:50
12 minutes, 50 seconds
should invest in just one, it’s some kind of conflict of interest, and not investors in OpenA, XAI, Antropic,
12:57
12 minutes, 57 seconds
SSI Super Intelligent, Safe Super Intelligent, or Sutsver I and Mmir Murrati’s company,
13:05
13 minutes, 5 seconds
they’re in everything, and add at least two more startups creating decentralized AI. Basically,
13:12
13 minutes, 12 seconds
there are eggs in all baskets. And in this regard, and here, mm investor
13:20
13 minutes, 20 seconds
, no one really claims to know specifically which of them will win. Yeah. But everyone understands that this
13:27
13 minutes, 27 seconds
market will definitely be different and that its total aggregate value is potentially so high
13:34
13 minutes, 34 seconds
that tens of trillions justify all these efforts. That is, the point is that when I’m talking about expertise; it’s simply about not investing in any kind of bullshit.
13:44
13 minutes, 44 seconds
But in everything else that’s clearly created by a business’s consciousness, even if it loses, money will still be invested. You don’t know who will win. Let’s consider a hypothetical situation. Look.
Chapter 6: Let’s imagine what will happen without AI. And can AI collapse?
13:55
13 minutes, 55 seconds
Let’s still draw the analogy with datacoms. It’s clear that some companies remained, 98% or 99% of companies, they simply went into liquidation. Let’s just imagine, you say,
14:04
14 minutes, 4 seconds
no one knows. There are companies like Open AI, for example, that are pumping money into them. And in many ways, if we look at what’s happening with the markets now, this bubble,
14:13
14 minutes, 13 seconds
the hypothetical stock market bubble in general, it’s largely driven by these expectations that companies
14:19
14 minutes, 19 seconds
will become more efficient, that there will be staff reductions, higher profits, and, in fact, expectations are growing and the markets are growing. Our stock market is simply looking at the S&P 500 this way right now.
14:29
14 minutes, 29 seconds
Look, let’s imagine a hypothetical case. Some
14:35
14 minutes, 35 seconds
time passes, a year, two, three, I don’t know how long. And Open AI says, «No, that’s it, we understand
14:40
14 minutes, 40 seconds
that the math, well, guys, it doesn’t add up, yeah, the math doesn’t add up and that’s it.» And they collapse, and then, as you can imagine, a bunch of other companies collapse. So, let’s
14:50
14 minutes, 50 seconds
draw this here. Why am I asking this? Because many people are afraid of this right now. Many people, for example, are afraid to buy the stock market, or stocks in general, because they understand
14:58
14 minutes, 58 seconds
that everything could go to zero, and then 90% of the time it would work out. Moreover, in contrast, we see, for example, that this is an exceptional event, that on one hand, we have the
15:07
15 minutes, 7 seconds
American stock market growing, and on the other, gold is growing, right? It seems like you’re seeing both risky assets and gold growing. It’s clear that the government is buying Bitcoin, which is still rising. So,
15:16
15 minutes, 16 seconds
everything is growing. And that’s completely abnormal. And people who come in like this, I have, I don’t know, say, a million dollars there, I want to invest, they look at all this and think:
15:24
15 minutes, 24 seconds
“No, there’s something buried here. What will happen, in your opinion, in general to the US economy and to the economies of the world, if some giant like this just closes down, goes bankrupt and, naturally, drags
15:34
15 minutes, 34 seconds
down a number of other companies in general, and just as bubble peaks are usually unjustified, so bubble
15:42
15 minutes, 42 seconds
troughs are also usually unjustified, and most likely there will be a very sharp correction,
15:50
15 minutes, 50 seconds
especially considering that, and I just defended the companies, but at the same time, We don’t really agree
15:57
15 minutes, 57 seconds
with the current strategy, because you’re right that a very large amount of capital is
16:04
16 minutes, 4 seconds
not being invested in innovation. Yeah. A lot of capital is now being invested in scaling the current status, yes,
16:12
16 minutes, 12 seconds
data centers with even the current hardware. Fabulous amounts of money are simply, uh, in their reports they write down depreciation for 6 years, equipment that in 2 It should be completely obsolete in a year.
16:24
16 minutes, 24 seconds
But we’re seeing a protest right now, using Bitcoin as an example, that, so to speak, previous ASICs are simply ineffective. Imagine your hashrate, can you imagine how much, yes,
16:32
16 minutes, 32 seconds
I’m talking about small numbers here, but just imagine the volume of capital, well, the model is roughly the same. And Bitcoin itself, over the 13 years of its
16:43
16 minutes, 43 seconds
existence, Bitcoin ASICs have increased their productivity, how many mathematical operations are produced per invested capital and per applied electricity.
16:54
16 minutes, 54 seconds
Yes, they’ve increased this productivity by 300,000 times. And in this regard, in this regard, you’re really
17:03
17 minutes, 3 seconds
investing in innovation in this area, simply in building the current, yes, the current version
17:17
17 minutes, 17 seconds
of the equipment. This is indeed, from our point of view, a rather serious mistake. In this
17:26
17 minutes, 26 seconds
regard, too, it’s being done with very precise calculations, it’s being done with very precise calculations,
17:32
17 minutes, 32 seconds
and because now There’s still a very intense struggle going on for market dominance, for who will
17:38
17 minutes, 38 seconds
own it, and everyone is ready to forget about the possibility of tenderness. Yes, yes, yes. Ah, but in any case, we
17:47
17 minutes, 47 seconds
think that this is rather the wrong strategy. You’re right that if at some point fear arises
17:56
17 minutes, 56 seconds
and, well, such a chain fear, and because of what they’re doing, then, of course, we can… for several
18:04
18 minutes, 4 seconds
Roll back progress for years, when for a few years, as happened quite recently, in 2002, 2000, and
18:14
18 minutes, 14 seconds
2024, and because of the extremely high expectations that 2021 created. Yeah.
18:23
18 minutes, 23 seconds
For a few years, it was really difficult to raise money. It was absolutely impossible to raise money. Well, there were some interesting cases, and some of them, well, obviously, these are smaller
18:32
18 minutes, 32 seconds
analogies, but nevertheless, there’s the case of Farfech, yes, when the pandemic started, they said: «We’ll disrupt the fashion market and now everyone will buy online.» So. And they pumped them with money, valuations soared. And then it turned out that, basically, any brand can create its own online store.
18:44
18 minutes, 44 seconds
And about when the pandemic ended, because people generally love stores, and they still go to stores, and the company was auctioned off for a dollar. These very expectations,
18:54
18 minutes, 54 seconds
they actually lead to… Come on, let’s just say in this specific case, what will happen ?
18:59
18 minutes, 59 seconds
They’ll pump the market with a huge number of GPUs, and then it suddenly turns out that, for example, there’s no demand for them, or something else appears. Since all this will collapse, then what will this actually lead to?
19:11
19 minutes, 11 seconds
It could lead to the development of chip speeds being slowed for a while, because
19:17
19 minutes, 17 seconds
no one will provide additional funds for this development, but those chips won’t disappear. It ‘s not like there will be a boom and all of AI will disappear. No, it will simply be there. There will simply be
19:27
19 minutes, 27 seconds
a redistribution of who owns what. Yeah. It’s just a redistribution of money from one pocket to another, right? Someone will lose something, someone will be indebted, someone will buy
19:37
19 minutes, 37 seconds
it for a song. That’s what happened with real estate in 2017. It simply changed hands, and then everyone got
19:43
19 minutes, 43 seconds
all their money back, like, two years later. Not all, but rather those who received all this asset returned it. At the same time, Open Aa is very smart in terms of how they raise money.
19:54
19 minutes, 54 seconds
Most of this money doesn’t go back into the company, but to third-party companies that receive loans for building these data centers. So, OP won’t even lose much. And OP’s investors
20:03
20 minutes, 3 seconds
won’t lose much either. It’s more like an investor in these third-party companies with Oracle data. Not Oracle, in the sense that it’s
20:10
20 minutes, 10 seconds
all pumped into Oraac, which is building a data center and buying it for itself. He’s number one right now,
20:16
20 minutes, 16 seconds
yes, the richest man in the world. Well, thanks to this, Aa himself might not even wait
20:23
20 minutes, 23 seconds
until it suddenly turns out that all this hardware is no longer needed, and other things are needed. So, the idea is that you assume that even if this happens, it could be overvalued now, and
20:33
20 minutes, 33 seconds
it might also fall for a while, but it won’t be a story where the market reverses and goes back up. So, it’s inevitable. It’s just like any crisis, it’ll bounce back and,
20:43
20 minutes, 43 seconds
in a couple of years, return to normal and, let’s say, be healthier,
20:48
20 minutes, 48 seconds
growing and developing naturally, right? Another 20 years from now. Well, yes, it will cost a size 10, and in that sense,
20:53
20 minutes, 53 seconds
on the one hand, it’s simply inevitable; it’s just the psychology of it, yes, it’s very difficult to stop yourself
21:01
21 minutes, 1 second
when something like this happens and everyone rushes to invest at least some capital in it.
21:08
21 minutes, 8 seconds
And reality. But to invest in the capital of reality, you need expertise, you need to be able to distinguish where innovation is actually happening and
21:19
21 minutes, 19 seconds
where it isn’t, for example. And again, this is all very important; this is a truly life-changing
21:25
21 minutes, 25 seconds
product. It’s changing the lives of all of us right now, in every profession. And in that sense,
21:31
21 minutes, 31 seconds
it’s not a fake, it’s not some kind of super-bubble. So every country in the world
21:38
21 minutes, 38 seconds
or every company looks at this and says, «If I wait another year and
21:44
21 minutes, 44 seconds
this doesn’t go bust, I might just end up irrelevant.» Look, I’ll expand on
21:50
21 minutes, 50 seconds
this immediately. This is probably one of the main global issues right now. Yes, we have the United States of America, which is, naturally, the beneficiary,
Chapter 7: China, USA or Russia? Who will win in the AI race?
21:59
21 minutes, 59 seconds
as usual, just as it has been for many years now, the beneficiary of this entire boom. On the other hand, we have China, which we also see little about in open sources. My understanding is
22:09
22 minutes, 9 seconds
that everything there is top secret, conditionally, but something is being created. Meanwhile, America is trying Buying up Chinese personnel for insane amounts of money. So, we have a kind of
22:18
22 minutes, 18 seconds
competition going on between these two powers, so to speak. And here several questions arise.
22:24
22 minutes, 24 seconds
The first question is: firstly, what will happen if one of them ultimately wins, and how will events unfold? And the second point is, what will happen to everyone else? Because,
22:33
22 minutes, 33 seconds
as you know, it’s hypothetical, well, if we draw an analogy, of course, at some point Google appeared, for example, and it took over the world. But then country-specific, yes, local
22:44
22 minutes, 44 seconds
networks appeared. For example, Yandex appeared in Russia, and somewhere else, something else appeared. For example, Thailand has its own search engines, Turkey has its own search engines, and they are much more popular
22:51
22 minutes, 51 seconds
than Google. But still, when we talk about a search engine, it involves a certain amount of resources. When
22:57
22 minutes, 57 seconds
we talk about e-commerce, yes, the amount of capital and resources required for this is, like,
23:05
23 minutes, 5 seconds
thousands. And it creates the feeling that, for example, some country—it doesn’t matter which one, Kazakhstan, Russia, Albania, Brazil—they can’t afford to do anything about it on their own. And in
23:15
23 minutes, 15 seconds
this model, there’s a feeling that at some point there will be one giant that defeats everyone
23:21
23 minutes, 21 seconds
and, say, takes over the world, and no one can counter it after that. That’s how you see this situation now—the US, China, and beyond, accordingly, the data coming out of it,
23:32
23 minutes, 32 seconds
you know? But what we’re doing now is the result of our analysis from more than three years
23:40
23 minutes, 40 seconds
ago. More than three years ago, we looked at everything that was happening. And you have to understand that, as far as
23:46
23 minutes, 46 seconds
I’m concerned, this isn’t something that comes from public sources. We’re close friends with Greg Brockman and Ilya
23:51
23 minutes, 51 seconds
Sutskiver. We’re friends with Altman and all the other players in this market. And we saw what was happening
23:58
23 minutes, 58 seconds
and what they were betting on. At first, around 2017, it was like, you know, we gathered in the courtyard
24:05
24 minutes, 5 seconds
somewhere around Altman’s house in Polualta and were placing bets on when Ajai would happen. In 30 years, in
24:12
24 minutes, 12 seconds
50 years, or maybe in 10 years. And it was like a joke. Yeah. But already three years ago, it was
24:18
24 minutes, 18 seconds
a real turning point for us. We were like, this is happening, we were wrong. We really thought that quantum computers would be needed to achieve what we’re about to achieve.
24:27
24 minutes, 27 seconds
And we started analyzing what was happening, what was happening in politics, literally,
24:32
24 minutes, 32 seconds
in politics regarding this. And then there was a moment, I don’t know, two years ago, two years ago, a little more
24:40
24 minutes, 40 seconds
than two years ago, when a small group of Open AI, Microsoft,
24:52
24 minutes, 52 seconds
XI, Elon, and I don’t remember who else got together, and Ividi, plus some government representatives.
25:00
25 minutes
And an agreement was made that none of you could have a monopoly on GPUs,
25:07
25 minutes, 7 seconds
and Nvidia had to divide all of its GPUs between you, right? What a geopolitical decision, essentially,
25:13
25 minutes, 13 seconds
yes, that’s quite specific. So. And as soon as that geopolitical decision was made, and, in parallel, Biden’s Executive Order was passed, uh, uh, which stated that
25:23
25 minutes, 23 seconds
this was a licensed activity, that no one could
25:29
25 minutes, 29 seconds
train models with more than a trillion parameters without permission from the US government. These two factors, and then the rest of the BS,
25:37
25 minutes, 37 seconds
which prohibits the sale of chips, despite the fact that it’s a Taiwanese company, TSMC. But Nvidia is
25:45
25 minutes, 45 seconds
an American company, even though it produces them in Taiwan, but nevertheless, control over who it can sell these chips to exists and is determined by the US government. At that moment, it’s like,
25:55
25 minutes, 55 seconds
«Wow, we’re absolutely done, we’re already, we’re already at this point, we’re there. Not now, not in two years, we’re already at this point two years ago.» That is, it apparently hasn’t happened yet,
26:04
26 minutes, 4 seconds
but conditional agreements have already been concluded on how things will go. Well, everything was already being split and divided three years ago. And two years ago, it was recorded. And so, over these
26:13
26 minutes, 13 seconds
two or three years, China and the US had a certain advantage in terms of
26:21
26 minutes, 21 seconds
computing power, something like 25-20% or something like that. Well, that was something like that, and now the US has
26:27
26 minutes, 27 seconds
90% of the members. Friends, disorganization in work communications is a tricky thing. At first, it seems
26:33
26 minutes, 33 seconds
like everything is under control, and then it starts. Tasks are in correspondence, documents are scattered across cloud storage, and there’s constant confusion in areas of responsibility. As a result, the system disappears.
26:42
26 minutes, 42 seconds
And with it, the manager’s peace of mind. This creates a whole host of problems for the company: important files are lost, and project correspondence worth millions of rubles simply disappears, along with the chats
26:52
26 minutes, 52 seconds
they were stored in. For those tired of constant chaos, there’s a solution. Strive is a Russian service for team collaboration with tasks, projects, and documentation, helping you bring order
27:02
27 minutes, 2 seconds
to your company’s operations. More than 20,000 teams and organizations have already tried it. See how it works. You create separate spaces for departments, simple
27:12
27 minutes, 12 seconds
yet packed with functionality. Boards allow you to create tasks and subtasks with deadlines, priorities, and assigned team members. A built-in chat lets you discuss details and attach files
27:21
27 minutes, 21 seconds
and links. Regulations and instructions are stored in one place, reducing the time it takes to train new employees.
27:26
27 minutes, 26 seconds
In addition to the web version, there’s also a full-fledged mobile app. Strive replaces dozens of tools, including a knowledge base, documents, and task statistics. You can see who’s online and what they’re doing,
27:36
27 minutes, 36 seconds
where deadlines are looming, and how busy the team is. Managers gain transparency and control, while the team gains structure and convenience. Strive is currently running a special offer. Anyone who
27:46
27 minutes, 46 seconds
submits a request via the link in the description or the QR code on the screen will receive a free onboarding for your business. During the call, a manager will show you how your company can streamline tasks,
27:55
27 minutes, 55 seconds
eliminate chat chaos, and save your team time right from the start using Strive. Try it yourself to
28:01
28 minutes, 1 second
see how results change when everything is under control. Let’s get started. In fact, I’ve been
28:08
28 minutes, 8 seconds
betting for some time now on completely dominating this market. And in this regard,
28:14
28 minutes, 14 seconds
as long as we continue along this path—and you’re right, what will happen if we don’t deviate from this path—there will be
28:21
28 minutes, 21 seconds
these two policies, and their economic power will be enormous, much greater than even today,
28:31
28 minutes, 31 seconds
thanks to the control that comes from being the only one who can produce such
28:39
28 minutes, 39 seconds
models and let businesses use them. Ah, and that’s the plan for now. This plan has a very clear,
28:48
28 minutes, 48 seconds
well, clear calculation that most of these calculations are now done on Nvidia cards.
28:57
28 minutes, 57 seconds
I saw this Nvidia graph before, I just saw it, maybe even differently. It’s just that before, Nvidia was known mainly because of money, yes, those who played computer games, yes, yes. But
29:07
29 minutes, 7 seconds
now everyone knows, because it’s the most valuable company. And many years ago, it created a standard
29:16
29 minutes, 16 seconds
that made these calculations possible. Many, many years ago. We worked with this standard around 2002-3. And in this regard, at first everyone thought this standard: «Well, why do we
29:28
29 minutes, 28 seconds
need it? It’s only for scientists.» It was even very strange how much effort they put into it.
29:33
29 minutes, 33 seconds
And it seemed very… Then he switched to games and won. Standard for what? What kind of standard? A standard
29:38
29 minutes, 38 seconds
for how mathematics is calculated on cards, how a program task is translated into
29:45
29 minutes, 45 seconds
a specific operation on a specific architecture. But, in essence, this is the performance of the chip, what determines, yes, this determines your ability, your ability to even write
29:56
29 minutes, 56 seconds
a program that will be executed on this chip is called. Well, in fact, with the help of this, you were already writing games on these, as it were, according to these rules. Even before games, even when
30:05
30 minutes, 5 seconds
this had not fully entered games, we used this thing for distributed mathematical calculations.
30:11
30 minutes, 11 seconds
Yeah. And it is thanks to this standard that they are now very much in
30:17
30 minutes, 17 seconds
control. If there were no standard, we would see a lot of players, we would see that there is still capital in the world A lot, not just in the United States and China.
30:28
30 minutes, 28 seconds
We would see that all companies could afford, all countries could afford, some
30:33
30 minutes, 33 seconds
alternatives. Yeah. But not now. As you know, every trade negotiation between the United States,
30:40
30 minutes, 40 seconds
for example, with Saudi Arabia or the Emirates, includes how much we’ll be allowed to buy. And I even saw the latest agreement, when Trump flew to the Middle East, yes,
30:50
30 minutes, 50 seconds
accordingly, in all these agreements there’s a quantity, yes, the quantity of GPUs that we’ll be allowed to buy and so on. And these are all very ridiculous figures. Yes, there are 100, 100 billion, like,
31:00
31 minutes
Yes, it’s a drop in the ocean given the current volume of things happening, I understand. And Open A and XAI are currently producing a million of these GPUs each. I’ve seen it. It’s absolutely ridiculous,
31:09
31 minutes, 9 seconds
of course. Yes. That’s it. But all countries understand that they need to at least get this. And very complex negotiations are underway about this. And we are actually moving in this direction for now.
31:22
31 minutes, 22 seconds
And this direction is actually not very pleasant for everyone else, and most likely for
31:32
31 minutes, 32 seconds
the rest of the world. Because it’s clear that
31:38
31 minutes, 38 seconds
everyone else will have to pay for all these $100 billion, billions of dollars that are currently being invested in data centers. And when you have no
31:43
31 minutes, 43 seconds
other choice, when you don’t have a cheap alternative of your own, you can’t do anything but pay. And in this regard, ah, I didn’t mention this to us, for three years this particular scenario has been unsettling.
31:57
31 minutes, 57 seconds
It’s almost inevitable. Ah, but we see a small scenario of US dominance in this game. Yes. Yes.
32:05
32 minutes, 5 seconds
Well, it’s interesting that, as you remember, DePS came out at the beginning of this year. Only a year has passed, not even a year has passed since the whole Strauss world. And then, at the beginning of the year, it was all about the Chinese, the Chinese women.
Chapter 8: Why is DeepSeek no longer heard about?
32:15
32 minutes, 15 seconds
Well, well, it died down, because in fact, a lot of untruths were said about
32:20
32 minutes, 20 seconds
DeepSky. It was all pure PR. But, but it’s true. DeepSky works. It’s a very good
32:27
32 minutes, 27 seconds
model. And rewind another year and a half to before that, we were running around trying to raise money to do roughly the same thing that DeepSky eventually did, because they had the money.
32:38
32 minutes, 38 seconds
Yeah. The company that made the model had $2 billion to do it.
32:43
32 minutes, 43 seconds
Wait a second. All this talk about it being $50 million, right? No, they had a ton of money and a lot of computers. These are very successful traders. These are traders. They had a ton of money.
32:52
32 minutes, 52 seconds
They bought a lot of these Nvidia cards. I’ll give you another example. There’s Alex Gerko,
33:00
33 minutes
the largest private taxpayer in England. That’s how Alex trades so successfully on
33:09
33 minutes, 9 seconds
the super-mega-robots. No, I mean mega-robots. He had 50,000 GPUs. That’s right, he had
33:15
33 minutes, 15 seconds
a comparable number of these cards for calculations to the number needed to
33:20
33 minutes, 20 seconds
train GPT, but he simply used them for trading. They’re
33:29
33 minutes, 29 seconds
a couple hundred very smart mathematicians who interview for months
33:35
33 minutes, 35 seconds
to get into Alex’s company. Amazing. And they interview them for months to see if they’re fit for purpose. And then these guys come up with models for using GPUs to
33:45
33 minutes, 45 seconds
trade Forex effectively. And it’s the second-largest private trading company in the world after Citadel. These guys are the Chinese version of that. They had a ton of money.
33:59
33 minutes, 59 seconds
Like, a ton. The main thing is that they allocated a couple of billion dollars to this project. And so our point, our point three years ago, yes, two years ago, was that it was possible.
34:10
34 minutes, 10 seconds
Everyone was already saying, «No, we can’t catch up, we can’t catch up with Apple, or Google, or Open AI, or Google, we can’t catch up.» We were like, «Yeah, it’s possible to catch up.» Exactly, it’s possible, it just takes
34:18
34 minutes, 18 seconds
a lot of resources. Yeah. The Chinese did it, and they’re kind of behind, but they’re, as they say in Russian,
34:24
34 minutes, 24 seconds
in English, choked, stifled, eh, stifled by the fact that they were deliberately pushed away
34:33
34 minutes, 33 seconds
from buying these chips two years ago. They’re trying to buy them through every possible crack. Cards go to them through Singapore, knimkarts go through Malaysia, chips get to China through every possible smuggling channel. But
34:44
34 minutes, 44 seconds
it’s crumbs, because 90% of these chips still end up with five American companies. Yeah. And
34:51
34 minutes, 51 seconds
right now, everything you’re talking about is just confirmation that there will be a kind of market monopolist in the form of one country, which will then impose the rules of the game, the overall
Chapter 9: Reverse scenario and alternative path: where does Bitcoin have the real scale
35:00
35 minutes
structure of the economy and the world, on everyone else. Unless we… And what’s the opposite scenario,
35:05
35 minutes, 5 seconds
right? Unless we—and that’s us, your audience, literally every one of us—don’t see and invest
35:15
35 minutes, 15 seconds
our efforts, our resources, our time, and our attention in an alternative path. And there’s a small chance,
35:21
35 minutes, 21 seconds
maybe, I don’t know, a 1% chance, that this price will actually happen. But this alternative path is completely
35:27
35 minutes, 27 seconds
open source, completely decentralized, and shared resources invested in one standard and one protocol.
35:36
35 minutes, 36 seconds
The one that ultimately surpasses it will probably lag behind for a couple of years, maybe even. But, as I showed
35:42
35 minutes, 42 seconds
with DeepSky, it took them a year to catch up enough to be able to
35:49
35 minutes, 49 seconds
use it effectively. And an example of this alternative path already
35:54
35 minutes, 54 seconds
exists. We’re just all looking at it through a different lens. That’s right. If you look at
36:01
36 minutes, 1 second
Bitcoin, everyone’s mainly discussing it as a financial instrument, yes, as the new gold,
36:08
36 minutes, 8 seconds
digital gold. Bitcoin makes up 10% of the gold market. Everyone’s mainly interested in how much Bitcoin is worth today, how it’s fallen or risen.
36:18
36 minutes, 18 seconds
But there’s another perspective: if we look at this as an infrastructure project. Right now,
36:24
36 minutes, 24 seconds
as you know, Open AI, XAI, Microsoft—they’re all building these big datacenters,
36:31
36 minutes, 31 seconds
Stargates. Some giant, yes, Kolass, he calls it, built it, and it has a million GPUs,
36:40
36 minutes, 40 seconds
and it’s considered the fastest construction project in terms of infrastructure. Tens of
36:46
36 minutes, 46 seconds
billions of dollars have been invested. But if we look at Bitcoin as a financial
36:52
36 minutes, 52 seconds
framework for funding infrastructure, then anyone can, say,
36:58
36 minutes, 58 seconds
become a miner, buy some machines, plug in their own power outlet, and basically become one. Now wait a second. Anyone can… well, how would I make my own… Can you
37:07
37 minutes, 7 seconds
imagine the scale of Bitcoin today? Well, I understand that, roughly speaking, it’s a highly distributed infrastructure. Highly distributed, but what’s its size? What’s its
37:17
37 minutes, 17 seconds
scale? 26 GW. Yeah. It’s hard to estimate. 26 GW is more than Microsoft, Amazon, Google,
37:27
37 minutes, 27 seconds
Open AI, XA. All combined. And I said Bitcoin is the future. Yeah. Yeah. Well, Bitcoin is simply
37:34
37 minutes, 34 seconds
the future. Bitcoin is already established, the most real. Yeah, I said it was the future, and today it’s
37:40
37 minutes, 40 seconds
established. It’s the largest data center in the world. All over the world today, it’s all over the news
37:49
37 minutes, 49 seconds
that Open AI or Elon Musk’s XAI have each built 1
37:58
37 minutes, 58 seconds
GW of data centers this year. Yeah. During the same time, Bitcoin added 5 GW. Yeah. This is incredible infrastructure,
38:05
38 minutes, 5 seconds
which we’ve achieved precisely because it’s distributed, because there’s no dominance
38:12
38 minutes, 12 seconds
based on your position or regulation. Yes, there’s still a certain conglomerate of mining companies
38:19
38 minutes, 19 seconds
that own a huge share of it all. You need to understand that. Can you imagine there being, like, 10,000 players who’ve divided up this entire market? It’s not a million anymore, like
38:26
38 minutes, 26 seconds
there were at the peak in the 1950s, 1960s, 1970s. It’s just 10,000 players, friends. When you go to a restaurant, the manager always greets you first and shows you
38:36
38 minutes, 36 seconds
to a free table. Then the waiter comes, takes your order, and gives it to the chef. But, of course, he starts preparing the dish. Everything runs like clockwork. Now imagine a restaurant where the chef himself buys ingredients, runs to tables, and prepares food.
38:49
38 minutes, 49 seconds
He also creates content, updates equipment and supplies. It sounds absurd, but this is exactly how a business operates without an COO, when the owner tries to be everywhere but ultimately
38:58
38 minutes, 58 seconds
gets nothing done. If you feel like your business is stalling due to process breakdowns, your team is disorganized, and you’re wasting days on routine instead of strategy, it’s time
39:08
39 minutes, 8 seconds
to systematize. Take the COO program from Edyuon Academy and ensure your business is sustainable, even when you’re not around. What’s in the course? 385 interactive lessons,
39:18
39 minutes, 18 seconds
49 business cases and assignments, and 20 ready-made templates for implementation. Modules on financial, team, and risk management from Itzhak Adis and experts from Skolkovo, Alfa-Bank, and the Higher School of Economics.
39:29
39 minutes, 29 seconds
Support from a personal mentor, lifetime access to the course and all updates, and a professional retraining certificate. This isn’t a typical course where you’re pressed for deadlines. Everything here is tailored
39:38
39 minutes, 38 seconds
specifically for people with businesses, projects, and families. There’s also the opportunity to complete an internship at Grруoup to practice your acquired skills.
39:47
39 minutes, 47 seconds
You can start optimizing your routine even before the course begins. After you submit your application, Edюon managers will send you a file with 30 Prompts for GPT chat that save hours on your operating system. Follow the link in the description,
39:58
39 minutes, 58 seconds
register for the program, and enter promo code «Sokolovsky» to get a 55% discount and a second
40:03
40 minutes, 3 seconds
course free. Let’s go. But besides the incredible volume, simply surpassing
40:11
40 minutes, 11 seconds
anything centralized players have ever produced, it was also the fastest innovation in
40:21
40 minutes, 21 seconds
the hardware itself. Yeah, that produced the hardware. These guys, these are the conductors themselves.
40:26
40 minutes, 26 seconds
Back in 2012, 2013, 2014—though 2014 might
40:32
40 minutes, 32 seconds
have been completely out of date—bitcoins could be mined on home GPUs
40:39
40 minutes, 39 seconds
, either on a home computer, or Nvidia was already selling special crayfish for mining crypto.
40:46
40 minutes, 46 seconds
They’ve created a new market. And where is Nvidia in Bitcoin mining now? But you just
40:52
40 minutes, 52 seconds
couldn’t compete with the speed of innovation that these, like, specialized bitmen came along, and so on, who started focusing specifically on this,
41:01
41 minutes, 1 second
and releasing updates every six months. A new machine, yes. Yes. And in 2010, it took
41:09
41 minutes, 9 seconds
5 million joules of energy to calculate one trashesh. Now,
41:19
41 minutes, 19 seconds
15 years later, one unit of Bitcoin calculation is 300,000 times smaller. Just 15 joules are needed to calculate a hash.
41:26
41 minutes, 26 seconds
It was an incredible innovation. If we can simply implement the same innovation, we’ll simply scale the number of chips, which now burn a lot of electricity, but the number
41:36
41 minutes, 36 seconds
of mathematical operations performed by these chips on this entire volume of gigawatts of electricity
41:43
41 minutes, 43 seconds
is tens of thousands of times greater than the number of all the mathematical operations performed by the entire cloud in the world. And in this regard, this is exactly how we should be approaching infrastructure now.
41:56
41 minutes, 56 seconds
There was no need to buy modern equipment for these hundreds of billions of dollars. What was needed was
42:01
42 minutes, 1 second
to launch a mechanism where these skilled people around the world have a simple model, where they
42:11
42 minutes, 11 seconds
understand that if I make this equipment 10% better, I earn 10% more. That I don’t
42:19
42 minutes, 19 seconds
need distribution, I don’t need to negotiate with Microsoft or Google, but that I can simply plug
42:26
42 minutes, 26 seconds
this equipment into an outlet, and I’ll be more successful. Well, yes, you see, the main idea here, mm,
42:32
42 minutes, 32 seconds
such a key difference, in my opinion, it’s more about a certain philosophy, right, because why did Bitcoin achieve such widespread adoption? Because the idea is that it’s a people’s
42:42
42 minutes, 42 seconds
tool, and everyone, let’s implement this, you know, somehow the model itself found the idea, and
42:49
42 minutes, 49 seconds
the consumer, as it were, also found this philosophy, but here, you see, on the contrary, they’re like, «We’ll do it ourselves, and you’ll be left out there.» As a result, people are sitting around, panicking about unemployment and wondering
42:59
42 minutes, 59 seconds
what will happen to them, while the beneficiaries are trying to work. I’m saying, and that’s why I’m saying, our only chance now is for all of us not to be left behind, and that there will be a single center. Because
43:11
43 minutes, 11 seconds
even when we say these are five American companies, well, that doesn’t change the essence. No, wait, wait. We’re talking about five companies. Okay, there’s no monopoly here. Well, sort of, they will be, but this is
43:21
43 minutes, 21 seconds
a matter of national security. How quickly will these companies come under US government control?
43:28
43 minutes, 28 seconds
They’re already under that control, but how quickly will they come under complete control? Well, especially if they start taking jobs from millions of people. Then it’s clear
43:39
43 minutes, 39 seconds
that this will be very restrictive compared to other options. And we’re simply sliding into a situation where it’s all in one
43:44
43 minutes, 44 seconds
set of hands. Completely in one set of hands. And now these one set of hands are deciding, and whether you behave well or not
43:50
43 minutes, 50 seconds
is enough for us to give you these resources. Yeah. And that’s it. And we’re sliding toward
43:56
43 minutes, 56 seconds
a world that’s not at all the one we’d like to live in. But on the other hand, if you imagine that
44:03
44 minutes, 3 seconds
this alternative could make this equipment 300,000 times cheaper, I, as an engineer,
Chapter 10: Competitive advantage of people using AI
44:11
44 minutes, 11 seconds
am already spending about $2,000 a month on AI. Few can afford that, yes,
44:20
44 minutes, 20 seconds
but it speeds up my work so much that it’s very difficult to compete with me, when I have a real
44:27
44 minutes, 27 seconds
example like you. So, you, owl, spend $2,000 simply buying subscriptions, subscriptions, subscriptions, yes, I mean monthly subscriptions. No, no, that’s just one
44:37
44 minutes, 37 seconds
2,000, I mean, they spend it just on programming. What about the API. And all the other
44:44
44 minutes, 44 seconds
subscriptions on top can probably add another 500 plus dollars. And that’s out of reach for most
44:50
44 minutes, 50 seconds
people. Well, of course, that’s understandable. Yes. And this creates a competitive advantage for me, which means
44:55
44 minutes, 55 seconds
other people will have less work, and I will temporarily have more work. For now, I won’t have any,
45:01
45 minutes, 1 second
because, ah, Elon Musk, when he uses Xi, he uses it for hundreds of thousands of dollars,
45:08
45 minutes, 8 seconds
because you can always include more calculations, more variations, and… But if you divide
45:16
45 minutes, 16 seconds
the 2,000 dollars I spend by 300,000 by 300,000 of this optimization
45:22
45 minutes, 22 seconds
that can happen, it’s zero. It’s literally zero. You don’t even need to count it. I mean, if we
45:27
45 minutes, 27 seconds
increase the performance of these chips by 100,000 times, then we need to divide the $2,000 it spends
45:35
45 minutes, 35 seconds
by 100,000, and we get zero. And that’s what it is, zero, comma, something, but still zero. That’s
45:42
45 minutes, 42 seconds
what’s possible. So, for the Bitcoin network, the fact that the number of data centers and
45:49
45 minutes, 49 seconds
the amount of computing has grown hasn’t changed the very essence of Bitcoin. Yes, it’s the same, it hasn’t changed at all. The same number of bitcoins, the same number of transactions per second,
45:57
45 minutes, 57 seconds
nothing I understand. So, if we apply this same principle to AI and computing,
46:03
46 minutes, 3 seconds
it could actually be a salvation from the scenario you described. That is, the AI
Chapter 11: Can AI become available to everyone? And concern about the appearance of ChatGPT
46:11
46 minutes, 11 seconds
​​in the scenario we’re describing could become so cheap that superintelligence would be available to everyone
46:18
46 minutes, 18 seconds
, and uncontrolled by anyone. Well, it’s a model for the uniform distribution of global resources,
46:26
46 minutes, 26 seconds
yes, and as far as I understand, confirm or deny, these same initiatives are already underway. Firstly, there are several startups that, yes,
46:35
46 minutes, 35 seconds
are somehow trying to implement this idea of ​​decentralized AI. I’m not directly in the context, but I’ve at least seen some reports there. There. And, secondly, even, uh, your colleagues tipped me off,
46:44
46 minutes, 44 seconds
suggested that what was happening in Kazakhstan between Durov and, actually, the president, is also somehow connected to this. Yes, the guys are looking somewhere in that direction,
46:53
46 minutes, 53 seconds
what’s happening now. So, uh, on the one hand, over the last year and a half, from
47:02
47 minutes, 2 seconds
the position that decentralization can never be sufficiently effective for such
47:08
47 minutes, 8 seconds
calculations, everything has shifted to the position that declaring decentralization works very well for such calculations. Well,
47:14
47 minutes, 14 seconds
actually, what happened two years ago was that practically everyone, uh, creators of social
47:23
47 minutes, 23 seconds
apps or messengers, uh, were very concerned about the GPT chat phenomenon. Yeah. It’s
47:32
47 minutes, 32 seconds
definitely true that there weren’t any social tools there yet. It’s still a question
47:39
47 minutes, 39 seconds
of time. They all understand, yes, it’s a question. They’re still a single-player game. Yeah, yeah, but
47:46
47 minutes, 46 seconds
how many people are already communicating there, the fact that this app is constantly popping up at the top of
47:53
47 minutes, 53 seconds
all the App Stores, it’s certainly worried everyone. Yes, speaking colloquially, I have a commensurate number
48:01
48 minutes, 1 second
of conversations on Telegram there every day. And recently, you may have heard that they’ve launched it
48:07
48 minutes, 7 seconds
in the United States. This is their social video generation app. And it’s called Sora,
48:16
48 minutes, 16 seconds
right? What we discussed before simply added a small social feature where you
48:22
48 minutes, 22 seconds
can mention or not mention another person, and they’ll be in the video. So, you know, listen, I haven’t seen this one, by the way. Sorry, but yes, this is the main, this is the main phenomenon.
48:32
48 minutes, 32 seconds
They added a small social feature, and it also immediately took first place, just
48:37
48 minutes, 37 seconds
number one, the app is the latest, and viral spread is what everyone was afraid of, but
48:47
48 minutes, 47 seconds
all social app makers from Mark Zugkirch to Pablo Dorov began to fear it. But then
Chapter 12: How did Meta screw up?
48:54
48 minutes, 54 seconds
Suckerberg, in fact, we saw it all. MTA released its similar product in 5 days
49:02
49 minutes, 2 seconds
or whatever, they tried to release their version, but they didn’t include the video app. Whatever it was called, oh
49:09
49 minutes, 9 seconds
my God, I don’t remember. They were so afraid of what was going on, but it’s
49:16
49 minutes, 16 seconds
clear everyone knows who’s doing what, that they released their version in 5 days, 5 days before
49:22
49 minutes, 22 seconds
, and now they’re just in shock inside the meta because of what they
49:29
49 minutes, 29 seconds
They released it, but no one uses it. It’s not just a joke, but everyone uses it. So, basically,
49:37
49 minutes, 37 seconds
if I understand your point correctly, this issue of product launches—yes, a huge meta releases a product and it’s not interesting, but there’s so much engagement that they
49:45
49 minutes, 45 seconds
release it and everyone talks about it. Yes, a technological approach. They’ve really figured out
49:52
49 minutes, 52 seconds
how to make products for messengers. Messengers, what happened? Two years ago, everyone
49:59
49 minutes, 59 seconds
got scared. How can they spend so much money like that? It’s expensive. If our billions of people,
50:06
50 minutes, 6 seconds
our billions of users, communicate with each other, we’ll never recoup our costs. They have a very limited
50:12
50 minutes, 12 seconds
budget, which they can’t earn, of course. Yes. Yes. But they changed the model and
50:18
50 minutes, 18 seconds
pay them twenty a month. And in this regard, they allowed themselves and
50:26
50 minutes, 26 seconds
released all the messengers for free. So, was it with WhatsApp or was it with Telegram in terms of bots? They all released
50:33
50 minutes, 33 seconds
some kind of MM agents. They were all simplified. They all had simplified models. And they
50:40
50 minutes, 40 seconds
were so bad in quality that people used to say, «Meta will never do anything in AI,
50:46
50 minutes, 46 seconds
and Durov will never do anything.» And everyone was just like, «Oh, we have a GPT double-double on
50:52
50 minutes, 52 seconds
chat GPT.» As I understand it, that’s where Durov and Musk’s initiative, the Telegram group, came from, yes,
50:57
50 minutes, 57 seconds
which failed miserably, but had a good foundation. Where does it actually come from? Because
51:03
51 minutes, 3 seconds
Durov has an audience, but no AI. Yes. Yes. And he needs AI for his audience, but Musk
51:10
51 minutes, 10 seconds
doesn’t have an audience, because OpenI already has the entire audience. And why doesn’t Musk have an audience if
51:16
51 minutes, 16 seconds
there’s, say, Twitter? X Because people read on a reader, so they don’t come for
51:25
51 minutes, 25 seconds
that experience, they come to Open to chat. And this created this
51:34
51 minutes, 34 seconds
tension, when on the one hand, all social companies weren’t ready to spend so much money on
51:41
51 minutes, 41 seconds
user requests, yes. And on the other hand, well, apparently, something was gradually,
51:49
51 minutes, 49 seconds
well, gradually eating away at their market. That is, Open was still small back then, now they have almost
51:55
51 minutes, 55 seconds
a billion users. They are already comparable in size to all the losers. And yet, they are growing
52:01
52 minutes, 1 second
much faster than Telegram, or 15 years, and here, say, a couple of years. Yes. And in this regard,
52:11
52 minutes, 11 seconds
M. Dorum and I discussed two years ago that the only alternative is de-stylization,
52:18
52 minutes, 18 seconds
because de-stylization can make it so, and cheap. And what can allow
52:28
52 minutes, 28 seconds
this to be implemented? Yes, I understand. That’s it. But then Nikolai said, «De-stallization won’t work. And I actually
52:35
52 minutes, 35 seconds
have my own ideas about how to do it.» Two years later, we see that no, it didn’t work, and
52:41
52 minutes, 41 seconds
now they’re clearly trying to do something at the last minute. So. And in this regard, this is
52:47
52 minutes, 47 seconds
more of a confirmation that de-stallized AI is the only future,
52:57
52 minutes, 57 seconds
the only alternative besides these few American companies, which are currently achieving their goals primarily due to their dominant position with graphics processors.
53:09
53 minutes, 9 seconds
And we really love the Doro brothers and we really love Telegram. We use it every day, basically , as our main messenger. Ah, well, the approach needs to be completely different. These are completely different products.
53:20
53 minutes, 20 seconds
The story here is specifically about the product, about making a convenient thing with user experience, yes, with a cool twist. But AI is a completely different approach. The interface there might
53:31
53 minutes, 31 seconds
be completely broken, and there might be a single-window interface with one button, but the question is what’s on your back end, well, yeah, and you can’t draw it quickly, yeah. And then there’s another thing. The amount of
53:40
53 minutes, 40 seconds
your computation. In fact, in any case, to simulate a model, you need a lot of computation. A lot, to catch up, you need a lot. Well, and even just to provide
53:50
53 minutes, 50 seconds
this to your users, you need even more computation.
53:58
53 minutes, 58 seconds
Every one of these companies is simply stupid, whoever we can call something like Kanvart, whatever,
54:09
54 minutes, 9 seconds
chatfuls, minichats, who needs LMs, who needs graphic text, and so on.
54:15
54 minutes, 15 seconds
There are many, many companies. We can even do anything here. Everyone needs computation if each of them fights for their piece of the GPU. But they’ll never catch up with the top five.
54:29
54 minutes, 29 seconds
Each of which has millions of GPUs. Uh-huh. Yeah. Well, Kazakhstan doesn’t have a million GPUs and never will. And
54:35
54 minutes, 35 seconds
it’s clear that no one will give it a quota, right? Yeah. And in this sense, Durov is trying to do something for himself, because it’s clear that nothing can go on without AI. Can he
54:45
54 minutes, 45 seconds
create an alternative for the entire world? He can’t. And why Kazakhstan? Because Kazakhstan bought GPUs and they need to somehow use them, right? Ah, I get it. So, they don’t understand
54:55
54 minutes, 55 seconds
what to do with them. They have GPUs, and they seem to understand what to do with them.
55:01
55 minutes, 1 second
I’ve already flown to 24 countries this year. Each country has its own thousands. And somewhere after you, I’ll be thinking about what to do with them. After you, we’ll meet with the Minister of AI here in Dubai. Before
Chapter 13: Countries don’t know what to do?
55:11
55 minutes, 11 seconds
that, I just flew in and met with government officials from Vietnam. And before that, and before that, and before that, they all have many, many, what do thousands, thousands of GPUs mean, each of them.
55:22
55 minutes, 22 seconds
Everyone’s wondering what to do with them now to remain sovereign in terms of their AI. Yes,
55:27
55 minutes, 27 seconds
but for this to be relevant, all this needs to be done, and a free protocol in which
55:35
55 minutes, 35 seconds
everyone is willing to participate, so that taken together it would be something significant,
55:43
55 minutes, 43 seconds
so that taken together it would be commensurate with the scale of the major laboratories and the scale
55:52
55 minutes, 52 seconds
of DeepSky. And then we need to launch this process so that innovation in chips could
56:05
56 minutes, 5 seconds
allow all these distributed players to overtake these big
56:11
56 minutes, 11 seconds
players. Otherwise, there is no other way. Otherwise, the ability, that is, open source itself,
56:17
56 minutes, 17 seconds
won’t be able to do anything, except for an onsource. And also a similar or even identical financial framework, the same exact financial framework as Bitcoin.
56:30
56 minutes, 30 seconds
With the same motivation behind why money does this, you print some new currency,
56:38
56 minutes, 38 seconds
the value of which is determined by the market of traders, uh, speculation about whether it will have
56:46
56 minutes, 46 seconds
a large share of the market you’re targeting. Like with Bitcoin, everyone’s targeting whether it will be 10% of the gold market
56:51
56 minutes, 51 seconds
or not. And that was speculation, whether it will or won’t. It’s a skertil skerti. There’s
56:57
56 minutes, 57 seconds
a skertil of metal, and here we have one that’s primarily in electricity, which has become
57:03
57 minutes, 3 seconds
a derivative of electricity. And within such a framework, there could be an expectation that
57:14
57 minutes, 14 seconds
this coin of this protocol could serve as a subsidy for the
57:22
57 minutes, 22 seconds
construction of this infrastructure now. The expectation of market share was about the future. Now let’s move a little bit from
57:29
57 minutes, 29 seconds
our imaginary fantasies into the real world. To what extent, in your opinion,
57:34
57 minutes, 34 seconds
is this really possible? Because, like, what
57:40
57 minutes, 40 seconds
happened, for example, with Bitcoin, yes, you also need to understand the context of how this model unfolded. There was a crisis in 2008. People lost money they hadn’t saved their entire lives, blah-blah-blah,
57:50
57 minutes, 50 seconds
they were left without homes and lost confidence in the system where central banks printed money, and
57:57
57 minutes, 57 seconds
the dollar devalued. And then a solution came. And conceptually, I mean, it didn’t happen quickly either,
58:03
58 minutes, 3 seconds
but conceptually, gradually, gradually, gradually, almost 20 years have passed, just a second, yes, let’s not forget that. It’s like, right now, it’s turned into $2 trillion,
58:12
58 minutes, 12 seconds
yes, there you have a large decentralized network. As I listen to you now, there’s zero ideological talk . It sounds great. I don’t understand to what extent this can be realized and what needs to happen for it to
58:22
58 minutes, 22 seconds
happen, because, let’s say, some proactive group of guys said, «Yes, let’s make this decentralized thing.» They created protocols for this,
58:30
58 minutes, 30 seconds
created some kind of internal motivation, that profit sharing or something like that should work,
58:35
58 minutes, 35 seconds
I don’t know. But beyond that, that is, people need to understand the problem firsthand and,
58:42
58 minutes, 42 seconds
as it were, the financial model in order to want to do this. And for that, some very powerful catalyst is needed. That is, I still can’t figure out how this could actually happen.
58:52
58 minutes, 52 seconds
And let’s do it here. We can get straight to the point, because some guys have already gotten together and done it. Yeah. On the one hand, and some guys, that’s you. We got together and did it.
59:04
59 minutes, 4 seconds
We’ve been writing this for the last two years. Yeah. Ah, and until we’ve written it and made sure it works, it’s just not even
Chapter 14: What happened 2 years ago? Creating an alternative
59:09
59 minutes, 9 seconds
We didn’t tell anyone about this, and we launched it a month and a half ago, and it’s working and growing. I mean,
59:16
59 minutes, 16 seconds
the number of participants is growing. We grew from 80 GPUs to 450 GPUs in just a month. Let’s
59:23
59 minutes, 23 seconds
pause for a second. This is important to keep the chronology straight. So, look, as far as I understand, three years ago, I think, yes, or something like that, you actually
59:33
59 minutes, 33 seconds
launched Product Science. Yeah, we raised, I think, dollars in the first round. First of all, that’s a pretty big round for SED. So, it’s like, well, the amount is usually smaller. Secondly,
59:42
59 minutes, 42 seconds
if I looked at TCR correctly, it was 2012, or what year was it? 2021, yes, we raised a day. Since then, I looked at yours, you have,
59:52
59 minutes, 52 seconds
what’s the name of the product, the one with the errors in the code. Well, and inside there’s a code tuner, a code tuner,
59:58
59 minutes, 58 seconds
yes, a code tuner. So, you haven’t raised any money since then, right? And you know what my question is? So, look, on the one hand, you’re deep in the AI ​​context of what
1:00:07
1 hour, 7 seconds
‘s happening, you’re in the valley, you have a reputation, blah-blah-blah-blah-blah. At this moment, there’s an absolute explosion of startups in this LMA industry. Yes, startups in this field are starting
1:00:17
1 hour, 17 seconds
to pump in crazy amounts of money. And you’re not rushing around and raising round after round. Let’s take this as a starting point. How did that happen? Tell us
1:00:27
1 hour, 27 seconds
how it started and what it is. And let’s move on to what we have now. Two years ago, there was this moment, yes, when we were actually already working on an AI product
1:00:39
1 hour, 39 seconds
that worked with large companies, in short. You come in, run this thing on your code, it helps you find errors, yes, it helps you optimize. In fact, it helps you
1:00:49
1 hour, 49 seconds
speed up your offering right to the max. It’s like he’s playing Tetris, finding
1:00:55
1 hour, 55 seconds
where the code isn’t optimally placed, and optimizing it is like compressing the code into a tighter density so it
1:01:02
1 hour, 1 minute, 2 seconds
runs faster. And artificial intelligence does this beautifully. So, we, uh, we built this company, and it worked. Our clients were Walmart, Airbnb, large companies. That’s it.
1:01:15
1 hour, 1 minute, 15 seconds
But two years ago, and because of this explosion, we, uh, uh, just realized that creating such
1:01:23
1 hour, 1 minute, 23 seconds
a protocol and creating this destration infrastructure, and this, on the one hand, is the
1:01:32
1 hour, 1 minute, 32 seconds
only thing we need to do now, and on the other hand, this could be a once-in-a-lifetime
1:01:41
1 hour, 1 minute, 41 seconds
opportunity to create something of this scale. Yes. And then we decided to make what’s
1:01:49
1 hour, 1 minute, 49 seconds
called a beer, uh, and incubate such a protocol within Product Science. And at
1:01:56
1 hour, 1 minute, 56 seconds
that very moment, as soon as we decided this, we imposed a tab on ourselves, well, like, a,
1:02:03
1 hour, 2 minutes, 3 seconds
a restriction, uh, that we didn’t raise money for this project. How are you? It wasn’t with Product
1:02:09
1 hour, 2 minutes, 9 seconds
Science’s money. It was funded by Product Science. But what’s the difference? We raised money for Product Science
1:02:15
1 hour, 2 minutes, 15 seconds
for a completely different project than the one we later became involved in. So, we never promised any of our investors that we’d build this decentralized network with their money.
1:02:28
1 hour, 2 minutes, 28 seconds
And that’s important. Why? Because, as you know, when you launch decentralized networks,
1:02:34
1 hour, 2 minutes, 34 seconds
when you launch coins, any promise to investors in advance that they’ll give you money to
1:02:41
1 hour, 2 minutes, 41 seconds
build it in the United States—that’s not allowed by these regulations. You’ll immediately be
1:02:50
1 hour, 2 minutes, 50 seconds
considered a security. But if we want to build Bitcoin, it has to be commodity,
1:02:57
1 hour, 2 minutes, 57 seconds
it has to be crypto, it has to be a lithium token used for a specific
1:03:06
1 hour, 3 minutes, 6 seconds
product. So, we set ourselves the limitation that we wouldn’t communicate with investors about this,
1:03:11
1 hour, 3 minutes, 11 seconds
we wouldn’t raise any money, we just do it and launch. And we had enough resources
1:03:17
1 hour, 3 minutes, 17 seconds
at the time. And what’s more, creating these systems requires a very small number of very talented engineers and products. It doesn’t require, uh, thousands of employees.
1:03:31
1 hour, 3 minutes, 31 seconds
And we decided to focus on this, and we decided to impose a limitation on ourselves, that we don’t use these, uh, opportunities to raise money. Because if we can, uh,
1:03:42
1 hour, 3 minutes, 42 seconds
launch this project without raising money, raising money, then it will become possible from a regulatory standpoint. And then, and then, there’s an opportunity to create a phenomenon.
1:03:54
1 hour, 3 minutes, 54 seconds
Which won’t be stopped due to any errors at the time of creation. And what are we doing? We created this protocol, and at the time of launch, we completely released it to the public domain.
1:04:08
1 hour, 4 minutes, 8 seconds
It’s like in [the original text — context needed]. We made it open source. And we launched it a month ago.
1:04:18
1 hour, 4 minutes, 18 seconds
It’s not controlled by us, it’s already controlled by the community. We have a very strong
1:04:24
1 hour, 4 minutes, 24 seconds
voice in this community. And this is precisely a decentralized network, which in the future will develop absolutely
1:04:33
1 hour, 4 minutes, 33 seconds
only with the joint participation of all participants. And we believe that such a limitation is,
1:04:40
1 hour, 4 minutes, 40 seconds
although it is clear that it creates some obstacles, but this is the only way such projects can be launched and should be launched.
1:04:51
1 hour, 4 minutes, 51 seconds
That is, say, some Near Foundation with Ilya Polosukhin or Durov Sto
1:04:57
1 hour, 4 minutes, 57 seconds
, and so on and so forth. And many people now might be thinking, «Dura Sermiev.» Well, well, but you remember
1:05:04
1 hour, 5 minutes, 4 seconds
how they launched this project, raised $1.8 billion. And the court, of course, came sec. I’m talking about all
1:05:12
1 hour, 5 minutes, 12 seconds
this, and that’s exactly what we completely avoided with the way we made it so that it can’t happen to us.
1:05:17
1 hour, 5 minutes, 17 seconds
When it does, everything is so pure with us. We weren’t playing that game because we
1:05:23
1 hour, 5 minutes, 23 seconds
were playing a different game, creating a decentralized protocol, not pumping up
1:05:30
1 hour, 5 minutes, 30 seconds
staking in the foundation. There’s no foundation. So, in this regard, our history over
1:05:41
1 hour, 5 minutes, 41 seconds
the past two years was built on this, and we saw that it was very important to create this. And we saw
1:05:49
1 hour, 5 minutes, 49 seconds
that, on the one hand, the chance of this isn’t that great, yes, but this chance
1:05:56
1 hour, 5 minutes, 56 seconds
increases as the threat—this fear—increases. That is
1:06:05
1 hour, 6 minutes, 5 seconds
, more and more people are joining in over the past year and a half. You’re right that it’s still a niche of people who understand,
1:06:11
1 hour, 6 minutes, 11 seconds
but it’s the same with Bitcoin. It was a niche for many years. It’s not like everyone rushed in and said, «Oh, we don’t
1:06:17
1 hour, 6 minutes, 17 seconds
like central banks and everyone rushed to buy Bitcoin.» Well, no, we’re talking about tens of
1:06:23
1 hour, 6 minutes, 23 seconds
thousands of people in the second year, hundreds in the third, and then reaching a million in the fourth. We’re talking about
1:06:30
1 hour, 6 minutes, 30 seconds
how it’s been growing for a really long, long time. If this technology now simply follows the same, I mean, same trajectory, but faster. Why faster? Because, oh, everyone’s already
1:06:40
1 hour, 6 minutes, 40 seconds
seen Bitcoin. Well, it’s a second trend. A second time. The second time, people won’t want to miss out on what
1:06:46
1 hour, 6 minutes, 46 seconds
they missed the first time. And another thing is that right before our eyes,
1:06:54
1 hour, 6 minutes, 54 seconds
these five companies, plus several Chinese companies, are taking over this market at incredible speed. And it’s scary, and it’s only going to get scarier and scarier. Look, before we go any further,
Chapter 15: Why is the philosophy of humanity more important to us?
1:07:04
1 hour, 7 minutes, 4 seconds
first of all, we’ll definitely talk about how it works, because, frankly, I don’t fully understand it yet. But look, I was talking to the guys
1:07:10
1 hour, 7 minutes, 10 seconds
before recording the podcast, and we discussed something similar to what we discussed last time, about what you’re doing now. And this question actually came up, actually, from my colleague, who said,
1:07:19
1 hour, 7 minutes, 19 seconds
«Here’s one story I don’t understand.» So, the brothers, I mean, they started working in AI a few years ago
1:07:25
1 hour, 7 minutes, 25 seconds
, even raised money, were essentially already in the market, had seen everything that was happening, had seen products, and so on. And when this super-megatrend, this super-megawave, this
1:07:35
1 hour, 7 minutes, 35 seconds
crazy helicopter money return, and so on, why are you, instead of joining
1:07:42
1 hour, 7 minutes, 42 seconds
these guys who are operating with huge amounts of money and currently have insane valuations, and becoming
1:07:49
1 hour, 7 minutes, 49 seconds
beneficiaries of this, instead of going with what you’re talking about now, this libertarianism or whatever you want to call it, like, «Yeah, give it to the people, give it to society,» and so on?
1:07:59
1 hour, 7 minutes, 59 seconds
From a business perspective, you have a reputation, you’ve had successful exits in the Valley, you’re known, you can easily pitch, attract money, you’ve already built a very
1:08:08
1 hour, 8 minutes, 8 seconds
large sedron, and so on. So, essentially, the track could have just gone boom, and now the market capitalization is several billion. Why? Why not this vector over the last four months?
1:08:19
1 hour, 8 minutes, 19 seconds
We’ve had two offers from leading companies and laboratories, the very leading ones, to pay
1:08:27
1 hour, 8 minutes, 27 seconds
us over $100 million to buy us. And this decision can be understood from
1:08:38
1 hour, 8 minutes, 38 seconds
two perspectives. This is if we’re in reality, if you listen to our previous interview with us, those
1:08:44
1 hour, 8 minutes, 44 seconds
People who know, know that for us, philosophy is a philosophy of thought—thinking about the world and about the world
1:08:51
1 hour, 8 minutes, 51 seconds
of this future abundance. For us, it’s about creating tools, social tools, social
1:08:58
1 hour, 8 minutes, 58 seconds
innovation, that can take humanity to a completely different point of abundance. Now
1:09:05
1 hour, 9 minutes, 5 seconds
we can talk about AI abundance. Back then, it was still a bit crazy to talk about abundance. Well, even back then, you remember, we discussed it in general. But for us, this is a much more important thing, something we
1:09:15
1 hour, 9 minutes, 15 seconds
want to create in our lives. You know, how Elon Musk wants to take people to another planet—that’s more important to him. It’s kind of forgotten what PayPal did and everything else did—that’s his
1:09:24
1 hour, 9 minutes, 24 seconds
goal. It’s the same for us. Money is a resource for us to achieve this goal.
1:09:29
1 hour, 9 minutes, 29 seconds
And now, even talking about money, even talking about money, we’re now talking about, I don’t know, a couple hundred million dollars. What about those leading companies who make us offers
1:09:40
1 hour, 9 minutes, 40 seconds
, or we refuse to create our own WinSF? Yes, because we already made a product
1:09:46
1 hour, 9 minutes, 46 seconds
that used AI to speed up applications. And everyone will need it for another year or two . Maybe then they’ll come to us to buy Product Science, but we’ll see. And
1:09:56
1 hour, 9 minutes, 56 seconds
we could also make, well, a couple of billion dollars there for sure. Not bad. Not bad.
1:10:03
1 hour, 10 minutes, 3 seconds
Not bad at all. And in general, it wouldn’t have any impact on the world in terms of where it’s going. Well, at least there would be a resource, yes, which, you know, is temporary, to be honest, I look at all this and understand:
1:10:13
1 hour, 10 minutes, 13 seconds
someone makes a cursor, yes, and five leading labs will make another one, which
1:10:20
1 hour, 10 minutes, 20 seconds
will completely disconnect them from this channel and that’s it. And then it all disappears, like a quarrel replacing
1:10:26
1 hour, 10 minutes, 26 seconds
TikToks, YouTubes, and everything else. Gradually, if this AI can do absolutely everything possible, then why don’t they make all the products, noticing all the other competitors? Yeah, right.
1:10:39
1 hour, 10 minutes, 39 seconds
They’ll do it, and they’re doing it. And just now, Open AI released their new product with agents. There
1:10:45
1 hour, 10 minutes, 45 seconds
were so many agent-building startups. Now, Open AI is taking over their market, too. Now, Open AI is making an agent builder. Now, Open AI is making video generation. Now, Open is creating some kind of trend.
1:10:57
1 hour, 10 minutes, 57 seconds
You can catch up as much as you want, earn money, maybe even manage to reach a couple of billion dollars, but in the end, you’ll lose everything and be left out of it. Or you watch and see
1:11:06
1 hour, 11 minutes, 6 seconds
what’s really changing in the world. And we’re not like, if we’re great and if we manage to create exactly what
1:11:14
1 hour, 11 minutes, 14 seconds
we want—namely, a direct alternative that will become a major part of the AI ​​market, only a decentralized source alternative—then we, as the founders, take these 10% of the coins. Yeah, right.
1:11:27
1 hour, 11 minutes, 27 seconds
And if the race is to a trillion, that is, if it’s like Bitcoin plus, but actually bigger
1:11:35
1 hour, 11 minutes, 35 seconds
than Bitcoin, we’re much more concerned about that than this race in Silicon Valley with the rise of
1:11:42
1 hour, 11 minutes, 42 seconds
several hundred million dollars and billions in sight. We’ll make hundreds of billions
1:11:48
1 hour, 11 minutes, 48 seconds
if we bet wisely. Now look, I’ve
Chapter 16: What could be potentially bigger than Bitcoin?
1:11:55
1 hour, 11 minutes, 55 seconds
been closely following the crypto market for quite a few years now. So, let’s say there were Bitcoin killers, yes, if
1:12:00
1 hour, 12 minutes
you can say so, there were quite a few of them. And guys would come along, create utilities, say, «We have a mining model,» and so on and so forth. It all ended the same way,
1:12:09
1 hour, 12 minutes, 9 seconds
right? So, let’s say Bitcoin, as you said at the very beginning, is digital gold. It’s much more mobile than gold. Plus, you can transact, plus it’s more liquid,
1:12:17
1 hour, 12 minutes, 17 seconds
and so on and so forth. It’s under the government now, yes, at least for the time being. And you’re very clear on how it would apply to your specific model right now. First of all,
1:12:26
1 hour, 12 minutes, 26 seconds
yes, frankly, I’m skeptical, because as soon as some kind of coin appears, I immediately feel a certain skepticism. Now let’s break down
1:12:35
1 hour, 12 minutes, 35 seconds
the protocol model, why people should be motivated by it, and why you think it even has some, yes, a small percentage,
1:12:44
1 hour, 12 minutes, 44 seconds
but nevertheless, the potential to become something very big. Let’s draw a connection. I need a clear connection with Bitcoin. With Bitcoin, we see all these 26 gigawatts of electricity. And they’re being spent on solving the problem.
1:12:56
1 hour, 12 minutes, 56 seconds
A mathematical operation that’s in no way related to the specific function you ‘re talking about. A transaction. So, the transaction itself takes up microseconds. Well, obviously, mining is the main thing;
1:13:06
1 hour, 13 minutes, 6 seconds
it’s still mining. Mining is mining—it’s really just burning electricity, brute-forcing code, yes, brute-forcing code. There. That’s the current state of Bitcoin. And it’s enough for it,
1:13:16
1 hour, 13 minutes, 16 seconds
because it created the transaction function. And all this burning electricity is needed for, uh,
1:13:22
1 hour, 13 minutes, 22 seconds
network security. Well, so that no one becomes a majority and cheats the whole network. Okay. So,
1:13:29
1 hour, 13 minutes, 29 seconds
what we do, we immediately said: «No, 99% +% 99% of this computing
1:13:36
1 hour, 13 minutes, 36 seconds
power should be spent on useful actions and only 1% on network security,
1:13:42
1 hour, 13 minutes, 42 seconds
on a secure network. And this is the big, this is the main difference. Like, the protocol
1:13:49
1 hour, 13 minutes, 49 seconds
that we are making is like this, and the useful action in it is the calculation of existing LM models and the training of new LLM models. How does it work? I don’t fully understand how it works technically.
1:14:00
1 hour, 14 minutes
Just give it to me. So, I took, I bought several machines from you GP, yes, in general,
1:14:06
1 hour, 14 minutes, 6 seconds
and on this GPU I can do the calculations, yes. And you, as in the case of Bitcoin, you installed
1:14:13
1 hour, 14 minutes, 13 seconds
certain software on it and connected it to the network. At this point, your machine starts communicating with
1:14:23
1 hour, 14 minutes, 23 seconds
hundreds of machines in the network and they begin to jointly form, and orchestration, orchestration, a single
1:14:33
1 hour, 14 minutes, 33 seconds
server, and, which can perform calculations. Yeah. And then and an external API. And then the same
1:14:40
1 hour, 14 minutes, 40 seconds
as in In Bitcoin, there’s something called proof of work. Yes, you earn coins by proving
1:14:48
1 hour, 14 minutes, 48 seconds
that you can perform transactions. Yeah. So, in our case, your server also proves that
1:14:55
1 hour, 14 minutes, 55 seconds
it can perform transactions. And the more transactions it can perform, the more coins
1:15:00
1 hour, 15 minutes
it takes every day. That’s it. And in this regard, your experience is no different from
1:15:07
1 hour, 15 minutes, 7 seconds
Bitcoin mining. You connected, it’s your machine, your work, yes. The only question is where
1:15:12
1 hour, 15 minutes, 12 seconds
the computing power goes. Yes. Yes. But at the same time, what happens on the other side is that developers
1:15:18
1 hour, 15 minutes, 18 seconds
of any application, any company, I can, for example, come and
1:15:24
1 hour, 15 minutes, 24 seconds
load all my calculations onto your network. And they are sent to this distributed network. Yes, and then they
1:15:33
1 hour, 15 minutes, 33 seconds
find out you wrote a question, it was read, and an answer came back. Also, you receive this answer only thanks to
1:15:42
1 hour, 15 minutes, 42 seconds
the distributed network. Each question is distributed among those who will answer it. And this
1:15:48
1 hour, 15 minutes, 48 seconds
answer then goes back to you. And all this is done in a secure manner, so that you can be sure…
1:15:58
1 hour, 15 minutes, 58 seconds
That the response you received isn’t a fake response or a forged response. That is, on our
1:16:04
1 hour, 16 minutes, 4 seconds
side, within the protocol, there’s an algorithm that ensures that the model
1:16:11
1 hour, 16 minutes, 11 seconds
the developer requested hasn’t been modified to the extent of quantization of that model. Perhaps
1:16:18
1 hour, 16 minutes, 18 seconds
this isn’t entirely clear. It’s like the term, meaning you can simplify. This is called quantization;
1:16:23
1 hour, 16 minutes, 23 seconds
you can use 16 bits, 8 bits, or 4 bits. And our model determines that if you try,
1:16:30
1 hour, 16 minutes, 30 seconds
if you try to deceive the developer and give them a different degree of simplification, our model
1:16:37
1 hour, 16 minutes, 37 seconds
will catch it, and you won’t receive a reward. In this regard, we really are the ones. This product
1:16:43
1 hour, 16 minutes, 43 seconds
isn’t a competitor to Bitcoin. We’re not creating new gold. We are, but on the other hand,
1:16:53
1 hour, 16 minutes, 53 seconds
you can try on Bitcoin, conditionally, yes, at the same time, what has happened over the past 15 years is
1:16:58
1 hour, 16 minutes, 58 seconds
that it wasn’t just Bitcoin, there were various projects, and in particular, those projects that some created
1:17:04
1 hour, 17 minutes, 4 seconds
Uh-huh. For example, there’s Fcoin or Fcoin. Oh yeah, it doesn’t have a market cap in the billions of dollars,
1:17:12
1 hour, 17 minutes, 12 seconds
it does have a market cap in the trillions of dollars. Fcoin currently
1:17:19
1 hour, 17 minutes, 19 seconds
stores 10 million TBT of data. Yeah. In fact, as an infrastructure, it’s a very successful project. 10 million
1:17:26
1 hour, 17 minutes, 26 seconds
TBT is more than YouTube. That is, all YouTube videos could be placed on this network. It’s a distributed network. Honestly, I admit, as much as I’m ashamed, I don’t know
1:17:36
1 hour, 17 minutes, 36 seconds
that this is a bad product. That’s the whole thing. I’m hearing about it for the first time. Yeah. Yeah. And Protocol Labs made them.
1:17:43
1 hour, 17 minutes, 43 seconds
That’s Fcoin. Uh, there’s also a project called renderer, it’s now, I think, valued at somewhere around 6 billion, 5-6 billion dollars. And they, uh, also distributed servers for rendering cartoons.
1:17:55
1 hour, 17 minutes, 55 seconds
Yeah. They’re actually used. It’s the cheapest way to render cartoons now, to use them online. These were all really niche markets. These were all niche markets,
1:18:05
1 hour, 18 minutes, 5 seconds
like file storage and image rendering. But AI is an order of magnitude larger market. Yeah. So, we’re not
1:18:15
1 hour, 18 minutes, 15 seconds
competing to become the next gold. We’re competing to see if it’s possible to be a large percentage
1:18:22
1 hour, 18 minutes, 22 seconds
of the AI ​​market. And unlike the rest of this financial, crypto component, yes, that is,
1:18:29
1 hour, 18 minutes, 29 seconds
the smart contract market today is about $2 billion. Yeah. This is the limitation
1:18:36
1 hour, 18 minutes, 36 seconds
that, well, this is the market, it’s small. The AI ​​computing market is already $40 billion today,
1:18:45
1 hour, 18 minutes, 45 seconds
and in 10 years, the computing market will be about $2 trillion. Yeah. So, when we say
1:18:54
1 hour, 18 minutes, 54 seconds
this is an eternal market, I mean, how much will they pay per year? That means we’re talking about a market capitalization in the tens of thousands. It shrinks, of course. A coefficient, a multiplier. That’s it. And in this regard,
1:19:04
1 hour, 19 minutes, 4 seconds
ah, in reality, ah, um, there is, in any case, skepticism, because you can’t compare yourself to the biggest… yes, you compare yourself to Bitcoin, but Bitcoin is unique,
1:19:14
1 hour, 19 minutes, 14 seconds
but if you compare yourself to various existing infrastructure, crypto projects, yes, have they taken 10% of the centralized market? Yes, they have.
1:19:27
1 hour, 19 minutes, 27 seconds
Bitcoin took 10% of the centralized gold market. That’s for now. Can we, you see,
1:19:32
1 hour, 19 minutes, 32 seconds
you’re still talking about «for now?» Can we theoretically imagine an inside the multiverse, a cross-section of this
1:19:40
1 hour, 19 minutes, 40 seconds
multiverse, in which a decentralized GPU network takes 10% of this future computing market,
1:19:48
1 hour, 19 minutes, 48 seconds
which we just valued at tens of trillions of dollars. Yeah. And if it can, then we, when
1:19:54
1 hour, 19 minutes, 54 seconds
we go from today’s point, today’s valuation of this small network, to the point that it will
1:19:59
1 hour, 19 minutes, 59 seconds
grow to trillions of dollars over the next 510 years. Yep. Everything has a chance. A chance, yes. And indeed, right now there’s
1:20:09
1 hour, 20 minutes, 9 seconds
a very high chance that the big AI labs, ah, it will be impossible to compete with them. And
1:20:16
1 hour, 20 minutes, 16 seconds
the American government will strangle everything, no one will have access. If you believe in this chance, you need to short the rest of the market now, because the rest of the market will not survive in this new world.
1:20:27
1 hour, 20 minutes, 27 seconds
But that’s precisely why the demand for an alternative will grow, because an alternative is needed.
1:20:39
1 hour, 20 minutes, 39 seconds
No matter what happens, neither Google, nor Meta, nor O will use us. Well, of course, this is for those
1:20:44
1 hour, 20 minutes, 44 seconds
who can’t put up a million of their own G-notes, right? But we’ll be used by companies like Canva,
1:20:51
1 hour, 20 minutes, 51 seconds
Flo, multi-billion dollar companies that do a lot of computing, and they need it too. It’s precisely
1:20:57
1 hour, 20 minutes, 57 seconds
for them that the risk is high, and for them the risk is high that these five companies will take over their market, making presentations instead of them, calculating human health indicators instead of them.
1:21:10
1 hour, 21 minutes, 10 seconds
In this regard, and just like for Bitcoin, every financial crisis was a leap, yes,
1:21:17
1 hour, 21 minutes, 17 seconds
and so this audience that is looking for an alternative increases. Also, every news story about
1:21:23
1 hour, 21 minutes, 23 seconds
how cars have completely replaced everything, taxi drivers are no more. That’s it. That’s another piece
1:21:30
1 hour, 21 minutes, 30 seconds
of news. And another piece of news. And another. Year after year, we will see that it’s not happening. Do you think it will happen? No,
1:21:36
1 hour, 21 minutes, 36 seconds
it will happen. It’s just that it should have happened 10 years ago. Not 10-20, how long is enough, right? There it is, slowly approaching. Look at the chart. And Weima is the self-driving car of San Francisco. Yes, yes,
1:21:47
1 hour, 21 minutes, 47 seconds
I saw it. Yes, they’ve already eaten up 20% of the transportation market, and it’s simply an exponential growth curve.
1:21:56
1 hour, 21 minutes, 56 seconds
With all these technologies, they really do take longer. When, in fact,
1:22:01
1 hour, 22 minutes, 1 second
when they tell us, «You’re already too late,» no, we don’t believe it, because
1:22:07
1 hour, 22 minutes, 7 seconds
this adaptation will take several years, it will take 5-10 years. But once this happens, the sector
1:22:17
1 hour, 22 minutes, 17 seconds
-by-sector effect on the market will be colossal, which will be visible sector by sector. Look,
1:22:23
1 hour, 22 minutes, 23 seconds
I have a question: I understand the model. How many GPUs do you currently have in your children? We
1:22:32
1 hour, 22 minutes, 32 seconds
currently have 450 GPUs. These are high-end GPUs. These aren’t home GPUs, these are server hoppers and blacklists. Now
1:22:39
1 hour, 22 minutes, 39 seconds
I have a question. Look, how do I think from the perspective of a large company? So, let’s say
1:22:44
1 hour, 22 minutes, 44 seconds
you have 400 now, and then it becomes 50,000, 100,000, and so on. So, if I’m such a giant company,
1:22:51
1 hour, 22 minutes, 51 seconds
then I can easily take your system and hack it. That is, I can take and redirect all my
1:22:57
1 hour, 22 minutes, 57 seconds
The computations will be put on your computing power, and all your processing power will quickly be consumed. And then, bam,
1:23:04
1 hour, 23 minutes, 4 seconds
the whole story ends. And what’s going on in my head now in terms of control over you, despite the fact that you only become uninfluenced when you become
1:23:14
1 hour, 23 minutes, 14 seconds
very large. Everything before this, there are a lot of risks and potential threats that can be directed at you. How many computers did banks have? Well, one, probably,
1:23:24
1 hour, 23 minutes, 24 seconds
not owned by banks in 2013. Ah, well, I don’t know. Could the banks banning you in 2013 have brute-forced 51%
1:23:33
1 hour, 23 minutes, 33 seconds
of Bitcoin? They could. And that was already the fourth year of Bitcoin’s existence. Why
1:23:38
1 hour, 23 minutes, 38 seconds
didn’t they do it? What are we seeing? I realized that as soon as this idea of ​​connecting more GPUs appears in your mind at the same time as other people have the same idea.
Chapter 17: What do we underestimate in the development of technology and why things can go differently
1:23:50
1 hour, 23 minutes, 50 seconds
And this is what allowed Bitcoin, regardless of the fact that Bitcoin initially only had 1,500 GPUs, yes, this is what allowed Bitcoin to never get close to the point where someone had a control network.
1:24:03
1 hour, 24 minutes, 3 seconds
Because as soon as someone saw the news that, damn, this exists, they watched some podcast where they talked about it, and at that moment you think: «Oh, I’ll connect now.» At the
1:24:13
1 hour, 24 minutes, 13 seconds
same time, 10 other people are thinking the same thing. If you saw this news, everyone saw this news. That’s it.
1:24:19
1 hour, 24 minutes, 19 seconds
And in this regard, the spread of knowledge is key here. And because
1:24:26
1 hour, 24 minutes, 26 seconds
this spread of knowledge, it leads to the fact that every news about someone adding a GPU leads to someone else also wanting to add more GPUs. And on the one hand, it was possible
1:24:36
1 hour, 24 minutes, 36 seconds
to add more, it is possible, but it is very expensive equipment. Yeah. Today, it’s already around a million dollars a month. Oh, yeah. You spend on equipment without connecting the network. One machine costs a million dollars.
1:24:49
1 hour, 24 minutes, 49 seconds
One GPU, one GPU. Its market price is somewhere between $3,040 and $50,000, depending on
1:24:55
1 hour, 24 minutes, 55 seconds
the model. So, renting it long-term from a cloud service wholesaler right now will
1:25:03
1 hour, 25 minutes, 3 seconds
cost you about two dollars an hour. So, you take $2 an hour, multiply it by 24 hours, round it up, and
1:25:08
1 hour, 25 minutes, 8 seconds
you get $50 a day to rent this one GPU. So, that means
1:25:16
1 hour, 25 minutes, 16 seconds
one GPU costs $1,500 per month. If there are 400 of them, we multiply them and realize that, so,
1:25:22
1 hour, 25 minutes, 22 seconds
I’m asking for this every month now. We launched a month ago and in this month we’ve grown to the point where, you know,
1:25:29
1 hour, 25 minutes, 29 seconds
we could pay for it ourselves, let’s say, yes, we could. I don’t know, someone says, «Why would you raise money for yourself?» No, we generally tell all the investors who come to us, «Oh,
1:25:37
1 hour, 25 minutes, 37 seconds
we want to invest in you.» We say, «There’s nowhere to invest.» Buy, buy machines, mine. You can’t even buy coins now. You can mine, so go ahead and mine. And they do.
1:25:47
1 hour, 25 minutes, 47 seconds
And VC investors come and say, «Okay, for me, this is a check of, I don’t know, $400,000, which
1:25:54
1 hour, 25 minutes, 54 seconds
I’ll invest over the next six months. Okay, how do I invest it in a startup?» That’s it. And in this regard, what
1:26:02
1 hour, 26 minutes, 2 seconds
we’re seeing is that this is simply happening naturally: the greater
1:26:08
1 hour, 26 minutes, 8 seconds
the faith, the more people who teach want to participate, and especially the security. Just
1:26:15
1 hour, 26 minutes, 15 seconds
a month has passed, and we haven’t had a single moment, and when someone had more than 30% of the hashrate. Moreover,
1:26:22
1 hour, 26 minutes, 22 seconds
people are adding, like, 100 more GPUs, 100 more GPUs. Well, the numbers have also grown, and in this
Chapter 18: What do people not understand about Bitcoin?
1:26:30
1 hour, 26 minutes, 30 seconds
regard, we think this will continue, that there won’t be a moment when someone, like
1:26:38
1 hour, 26 minutes, 38 seconds
now big, well-known miners, has suddenly allocated their GPUs, along with another 300 GPUs, and will add them in two
1:26:45
1 hour, 26 minutes, 45 seconds
steps. First 150, then another 150. So they also understand that, as they remember, it was like with Bitcoin. It’s important that no one ever has a majority. But I just need
1:26:55
1 hour, 26 minutes, 55 seconds
to be a big part of the network so I can earn more coins. What they don’t
1:27:02
1 hour, 27 minutes, 2 seconds
really understand about Bitcoin, regardless of the fact that it’s written down right there in Satoshi’s head,
1:27:14
1 hour, 27 minutes, 14 seconds
is that this concept is based on the assumption—and this is a very humanistic
1:27:21
1 hour, 27 minutes, 21 seconds
assumption—that the majority is honest. We know people who built Bitcoin equipment,
1:27:29
1 hour, 27 minutes, 29 seconds
but as soon as they started approaching 50%, we no longer do it. Yeah. And we waited until the spring breaker
1:27:36
1 hour, 27 minutes, 36 seconds
I caught up with them, because in reality, these networks are built by enthusiasts who
1:27:42
1 hour, 27 minutes, 42 seconds
want this future to happen. And what’s more, they know that this can only happen
1:27:48
1 hour, 27 minutes, 48 seconds
if security is maintained. I mean, well, now no one can do 50% anymore. But yeah, that was back then, yeah. No, no, well, they just stopped themselves, yeah, they just could have,
1:27:58
1 hour, 27 minutes, 58 seconds
but they stopped themselves. As soon as you do it, you’ve lost everything you’ve invested in it,
1:28:01
1 hour, 28 minutes, 1 second
because you’ve immediately destroyed the faith, yeah, the faith in it. That’s it. And it’s the kind of thing where you
1:28:13
1 hour, 28 minutes, 13 seconds
look at it and think, «Well, no, no, humanity isn’t like that, yeah, everyone’s so-so and so on.»
1:28:19
1 hour, 28 minutes, 19 seconds
There are millions upon millions of free GPs scattered around the world, 10,000 here,
1:28:26
1 hour, 28 minutes, 26 seconds
50,000 there, this government has 100,000 there, that one here and there, they’re scattered everywhere now
1:28:32
1 hour, 28 minutes, 32 seconds
, and most of them, well, are, like, half unused. Meanwhile, among the current miners
1:28:40
1 hour, 28 minutes, 40 seconds
of the AA network, they actually have significantly more resources, and they all still operate according to the same
1:28:47
1 hour, 28 minutes, 47 seconds
principle. They add. As soon as they see that arcs are being added, they add more. Then they see
1:28:51
1 hour, 28 minutes, 51 seconds
that… That is, in this regard, we are precisely expecting this to be a kind of natural growth. Growth doubling
1:29:03
1 hour, 29 minutes, 3 seconds
every month. But this doubling every month, we all know how this exponential works. That is, if
1:29:10
1 hour, 29 minutes, 10 seconds
everything goes as it is now, that is, our first month, well, it wasn’t double, it was quadruple. Well, it’s just that in the first month, we’ve already managed to
1:29:18
1 hour, 29 minutes, 18 seconds
discuss with many people and tell them how it works. And as soon as we launched, it connected pretty quickly. But in the second month, we expect it to double. Well, maybe even more, I don’t know. And then I expect
1:29:27
1 hour, 29 minutes, 27 seconds
this doubling to continue in the next, well, 5, 6, 7, 8, 9 months, until we reach 100,000 GPUs, most likely
1:29:35
1 hour, 29 minutes, 35 seconds
in less than a year. Yeah. It sounds provocative, but interesting. And if this whole thing
1:29:44
1 hour, 29 minutes, 44 seconds
works out, it’s, of course, a phenomenal case. If it doesn’t work, then we tried, but
1:29:51
1 hour, 29 minutes, 51 seconds
at least we tried, right? At least we’re here, you know, there’s this thing: we have money, which was involved from the very beginning, from
1:30:01
1 hour, 30 minutes, 1 second
the very beginning, since there were fewer GPUs, like Bitcoin mining works. The more GPUs,
Chapter 19: The main alternative to Bitcoin
1:30:07
1 hour, 30 minutes, 7 seconds
the more money is shared among everyone. Yeah, right. For example, in 2010, there were
1:30:14
1 hour, 30 minutes, 14 seconds
1,500 GPUs in Bitcoin City. Each earned 1,000 bitcoins in a year. Yeah. So now it’s 1,000 bitcoins in a year.
1:30:22
1 hour, 30 minutes, 22 seconds
Just imagine that amount now. But back then, that money was, yeah, life-changing. And in that
1:30:29
1 hour, 30 minutes, 29 seconds
sense, we have 100 million, yeah, it’s slowly growing, yeah. And a month ago, the cost of
1:30:36
1 hour, 30 minutes, 36 seconds
mining one coin was 20 cents a month ago, and now it’s not 2 cents, it’s 10 cents. And those people
1:30:44
1 hour, 30 minutes, 44 seconds
who participated a month ago, they’re like, this price change is very noticeable. And, ah,
1:30:53
1 hour, 30 minutes, 53 seconds
in this regard, we think that the more people see this as an alternative, and at first
1:31:05
1 hour, 31 minutes, 5 seconds
there will be a small number, at first there will be a small number right now, and
1:31:11
1 hour, 31 minutes, 11 seconds
now this small number will be mostly people who believe in or would like
1:31:17
1 hour, 31 minutes, 17 seconds
this future to happen, and they rent GPUs to install nodes on them and connect them to the network. Yeah. And
1:31:25
1 hour, 31 minutes, 25 seconds
the future, of course, will look different. The future will look like 10 specialized data centers, specialized chips, like access to them for years to come, buying them for billions of dollars. That’s
1:31:36
1 hour, 31 minutes, 36 seconds
what the future will look like. Exactly like it looked in Bitcoin. And there at first there were these
1:31:41
1 hour, 31 minutes, 41 seconds
1,500 GPUs, so you could connect your home computer and earn 1,000 bitcoins. And just
1:31:47
1 hour, 31 minutes, 47 seconds
three years later, it was all in specialized chips that you can’t just buy,
1:31:54
1 hour, 31 minutes, 54 seconds
that require. And now you know how Bitcoin basically works? These large companies
1:32:00
1 hour, 32 minutes
aren’t even enticing themselves anymore; they’ve already pre-sold these coins. It’s understandable, of course. Those who
1:32:06
1 hour, 32 minutes, 6 seconds
believe in the coin’s growth, for their own sake, are taking a margin of something like 5%,
1:32:12
1 hour, 32 minutes, 12 seconds
and everything else is already pre-sold. And in many ways, think about it, for large funds, like
1:32:18
1 hour, 32 minutes, 18 seconds
the large pension funds in America, who are now entering Bitcoin. It’s like for large investors: the easiest way for them to get in with, I don’t know, half a billion dollars is not to buy from the market.
1:32:29
1 hour, 32 minutes, 29 seconds
Well, firstly, if you go and buy another half a billion dollars of Force of Bitcoin from the market, it’s like a super operation – going to such a miner and buying out everything they produce over the next
1:32:38
1 hour, 32 minutes, 38 seconds
few months. And this market is simply completely different. It’s not me who installed my machine at home;
1:32:46
1 hour, 32 minutes, 46 seconds
anyone on the market pays upfront for this mining. Big companies are tempting. Yeah. And it’s
1:32:55
1 hour, 32 minutes, 55 seconds
like paying in advance for a gold excavation. Got it. You know, I have a question that involuntarily arises. In the last episode, we discussed what you actually do.
Chapter 20: What about the rest of the projects now?
1:33:07
1 hour, 33 minutes, 7 seconds
At that time, you talked about your fund, about investing in people, and so on, and so on, and so on. Accordingly, as far as I understand, you’ve been very focused on
1:33:15
1 hour, 33 minutes, 15 seconds
this lately. Actually, what’s going on in other companies? For example, I just went to the website, looked, and I saw that you, ah, haven’t increased the number of people
1:33:23
1 hour, 33 minutes, 23 seconds
you’ve invested in, right? So, let’s take a quick snapshot, an update after a year and a half. What’s going on with the other projects? Plus, as far as I remember,
1:33:32
1 hour, 33 minutes, 32 seconds
you also have some marketing companies there, like agencies, plus you have a company that deals with, say, biotech, yes. That’s basically
1:33:40
1 hour, 33 minutes, 40 seconds
what’s happening there. Look, we have our company, which represents our future.
1:33:50
1 hour, 33 minutes, 50 seconds
Yes, that day, the future began with us doing this to ourselves and creating this holding
1:33:55
1 hour, 33 minutes, 55 seconds
company. It actually has a number of projects, some of which were very public. Yes,
1:34:00
1 hour, 34 minutes
we talked a lot publicly about our fund, which invested people. Some were very
1:34:06
1 hour, 34 minutes, 6 seconds
private. Here’s our company, which optimizes applications for large corporations. Well, they
1:34:15
1 hour, 34 minutes, 15 seconds
did conferences, but professional conferences, so to speak. Our entire marketing was aimed at selling to large corporations. And we simply
1:34:25
1 hour, 34 minutes, 25 seconds
talked very little about it. And in this regard, what kind of update? I think that back then, when we met, our
1:34:33
1 hour, 34 minutes, 33 seconds
holding company, which represented the future, was still valued at 200 million. Then we
1:34:38
1 hour, 34 minutes, 38 seconds
did another round, where we were valued at 400 million. And we attracted more investment from
1:34:47
1 hour, 34 minutes, 47 seconds
very famous names, like Venka. I’m not sure the audience is familiar with them on the VIT website. Yes. Yes.
1:34:57
1 hour, 34 minutes, 57 seconds
That’s it. And we continued investing in people. It’s a fund. And as a fund, it has a cycle: you raise
1:35:08
1 hour, 35 minutes, 8 seconds
money, yes, you allocate the money, and then wait eight years for your exits. That’s it. The founders we invested in, the ones we’ve invested in, have achieved very good
1:35:17
1 hour, 35 minutes, 17 seconds
success. The main stage of this project now
1:35:26
1 hour, 35 minutes, 26 seconds
is to bring it all to exits. So, how do we get to the point with the fund
1:35:31
1 hour, 35 minutes, 31 seconds
where we can ultimately sell this stake in the future of these founders we bought
1:35:39
1 hour, 35 minutes, 39 seconds
for significantly more money than, say, the idea of ​​reselling stakes in people, right? So, again, as investors, yes, this is an investment project. No, I understand.
1:35:49
1 hour, 35 minutes, 49 seconds
So, your idea isn’t to receive dividends from these founders, but to resell their future to other funds. The model needs to be demonstrated in order to demonstrate its power
1:35:57
1 hour, 35 minutes, 57 seconds
, because we receive dividends, right? Ah, but dividends, as you know, are always
1:36:05
1 hour, 36 minutes, 5 seconds
a very small part of a company’s value. In this sense, we’re patiently waiting for
1:36:16
1 hour, 36 minutes, 16 seconds
the market to develop. And now, literally last week, someone contacted us again. I think it was Peter Diamandis’s people. I don’t know, you know, he’s the creator of Singularity University.
1:36:30
1 hour, 36 minutes, 30 seconds
He now also has one of the most powerful podcasts on technology.
1:36:37
1 hour, 36 minutes, 37 seconds
And they’re now launching a new project where they want to use this model and,
1:36:43
1 hour, 36 minutes, 43 seconds
accordingly, partner with us in terms of using this model. And the idea that this is
1:36:51
1 hour, 36 minutes, 51 seconds
a better investment—it’s just an idea that’s alive. We continue to develop it. We’re speaking, I don’t know,
1:36:56
1 hour, 36 minutes, 56 seconds
at the beginning of the year we barely spoke to students. We continue to respond, keep our finger on the pulse. But
1:37:02
1 hour, 37 minutes, 2 seconds
the fund’s cycle is about seven to eight years, I get it. Like any other fund. That’s it, I get it. So,
1:37:09
1 hour, 37 minutes, 9 seconds
you don’t have the idea of ​​investing in another 100 people there. It would be, if we showed us, as soon as we show results, we need a creditor. That’s how it works with funds,
1:37:19
1 hour, 37 minutes, 19 seconds
What you do first, then you do the second, and then I understand it. The classic classical model. That’s it, I understand. So, you need traction, you need some real exits
1:37:29
1 hour, 37 minutes, 29 seconds
in order to, say, open a new fund and attract money from it. We’re playing the long game here, so we’re waiting very patiently and continuing to move this forward. In the world of AI, this will be
1:37:41
1 hour, 37 minutes, 41 seconds
even more important. And our Biotech company is even more long-term. We’re using, uh, the discoveries
1:37:50
1 hour, 37 minutes, 50 seconds
our parents made, because computing works inside cells, inside neurons. And,
Chapter 21: What can a quantum computer do and what is its threat?
1:38:00
1 hour, 38 minutes
uh, we’re preparing the entire technological base in order to build the quantum computer of the future
1:38:06
1 hour, 38 minutes, 6 seconds
based on this principle. Wait, first, to conduct experiments, first we prepare for a long time to think through and understand what experiments to conduct, because they will be expensive,
1:38:16
1 hour, 38 minutes, 16 seconds
long-term experiments. That is, quantum computers are already, well, I think there are practically no
1:38:21
1 hour, 38 minutes, 21 seconds
people who understand that this will work. Ah, yes, it’s already kind of obvious that this is the future of,
1:38:28
1 hour, 38 minutes, 28 seconds
uh, computing. Ah, but everyone understands that this is beyond, uh, 20 years now. Ah, I mean,
1:38:38
1 hour, 38 minutes, 38 seconds
forward, 20 years. Yeah. Listen, by the way, forgive me for the preview about quantum computers. In general , regarding the crypto market and Bitcoin, yes, one of the key risks,
1:38:47
1 hour, 38 minutes, 47 seconds
let’s call it that, is the emergence of a quantum computer that can decrypt it. But, as far as I understand, I honestly haven’t delved deeply into this topic, that in general, the technological
1:38:55
1 hour, 38 minutes, 55 seconds
design is such that even if we arrive, well, or when we arrive at a quantum computer, it won’t be like this. We wake up and it’s here tomorrow. So, it will be a step. And at each of these
1:39:06
1 hour, 39 minutes, 6 seconds
steps, technologies will change, encryption levels will be supplemented, and by the time the computer itself appears, the encryption model itself will have completely changed. Moreover,
1:39:15
1 hour, 39 minutes, 15 seconds
Bitcoin doesn’t… No, moreover, Bitcoin seems like it’s a monolith that never changes, yes,
1:39:22
1 hour, 39 minutes, 22 seconds
but the Bitcoin code, that’s what I’m saying, is that by that point it will have changed 10 times already, and that
1:39:28
1 hour, 39 minutes, 28 seconds
the model will have adapted. Of course. And you can’t change the past because
1:39:33
1 hour, 39 minutes, 33 seconds
you’ll remain, and it will already be written down, and everyone will understand it. The only
1:39:40
1 hour, 39 minutes, 40 seconds
thing is, the keys to the cipher will certainly change, and that means, for example, someone could potentially log into your account. That’s the threat of a quantum computer for Bitcoin. Yes, I’m
1:39:51
1 hour, 39 minutes, 51 seconds
saying that someone could potentially decrypt it, yes, decrypt it, and try it, but I understand that this will also change. That is, the model will fully adapt afterward. But
1:39:59
1 hour, 39 minutes, 59 seconds
how do these distributed networks work? Adaptation doesn’t actually happen immediately, because you don’t have a single server that you update everything on. But as soon as there’s consensus
1:40:09
1 hour, 40 minutes, 9 seconds
among all the participants that an update needs to be made, these updates happen. Bitcoin
1:40:13
1 hour, 40 minutes, 13 seconds
updates happened just, well, a year ago, and they introduced an additional [music] type of encryption for,
1:40:21
1 hour, 40 minutes, 21 seconds
uh, to make these shared signatures. Yeah. This feature wasn’t in Bitcoin
1:40:27
1 hour, 40 minutes, 27 seconds
at the beginning. Ah, that is, as soon as there’s very clear proof that such an update is necessary,
1:40:38
1 hour, 40 minutes, 38 seconds
Yeah. The community itself accepts it. So, you’re really right
1:40:42
1 hour, 40 minutes, 42 seconds
that it’s more of a fairy tale that some computer will change everything in an instant. What will actually
1:40:51
1 hour, 40 minutes, 51 seconds
happen? It’s like you’re chatting today and you hope no one reads it.
1:40:56
1 hour, 40 minutes, 56 seconds
But they will. But they will. Come on, come on. I said it wrong. You hope no one
1:41:02
1 hour, 41 minutes, 2 seconds
reads it? But they do. No, no, wait, they can’t read it today. If you encrypt well,
1:41:07
1 hour, 41 minutes, 7 seconds
they can’t read it. Well, even if… But when you encrypt, you can still
1:41:12
1 hour, 41 minutes, 12 seconds
record it all. I understand. Look, I have no illusions whatsoever about
1:41:19
1 hour, 41 minutes, 19 seconds
everything we discuss and rewrite, that it will all be read someday. Yes, it definitely will, because it’s being recorded today, which means it will be read in the future. And as for bitcoin security,
1:41:29
1 hour, 41 minutes, 29 seconds
it will definitely be updated. There’s already aaa m quantum resistance encryption. It’s
1:41:36
1 hour, 41 minutes, 36 seconds
just that there’s no point in updating it now, because a quantum computer is still far from being able to…
1:41:43
1 hour, 41 minutes, 43 seconds
Be practically applicable. For Bitcoin, this will also mean that the hardware will need to be replaced, yes,
1:41:50
1 hour, 41 minutes, 50 seconds
let’s say. But wait a second, Bitcoin is exactly the kind of ecosystem in which hardware changes very
1:41:57
1 hour, 41 minutes, 57 seconds
quickly. Yeah. Like we, as was said, if over 15 years, productivity has increased 300,000-
1:42:05
1 hour, 42 minutes, 5 seconds
fold. Yeah. You can just try to imagine how many times it changed every year and
1:42:09
1 hour, 42 minutes, 9 seconds
how many years it takes for your current, old equipment to consume
1:42:18
1 hour, 42 minutes, 18 seconds
more electricity than the Bitcoin you produced with it is worth, you know? Yeah. And you’re out,
1:42:25
1 hour, 42 minutes, 25 seconds
so you throw out this old equipment and install new equipment in its place. Anyone who asks the question: «What should I do with my old Bitcoin equipment?» Throw it out. Well,
1:42:35
1 hour, 42 minutes, 35 seconds
those who understand how this market works, they entered into this consciously. They know
1:42:41
1 hour, 42 minutes, 41 seconds
how it works. So, when the equipment needs to be modified for quantum-secure
1:42:46
1 hour, 42 minutes, 46 seconds
encryption, it won’t be difficult at all. Oh, okay. One more topic, look,
1:42:53
1 hour, 42 minutes, 53 seconds
let’s continue. You and I, ah, last time we recorded, we discussed problems in the economy in general, yes, and we discussed the pension fund model, the fact that young people
Chapter 22: Rising unemployment and two scenarios for the future with AI. Robots to help people?
1:43:02
1 hour, 43 minutes, 2 seconds
don’t get paid, they have no motivation to start companies, and all that stuff. This is getting even worse now. Now it’s even worse, yeah, even worse. That’s exactly what I’m getting at. That
1:43:12
1 hour, 43 minutes, 12 seconds
is, a year and a half has passed since then, this thing is being cultivated even more, and at the same time , we have this so-called threat from artificial intelligence. Look, so
1:43:22
1 hour, 43 minutes, 22 seconds
here, I also want to try to figure out where we’ll finally
1:43:28
1 hour, 43 minutes, 28 seconds
turn this whole story. That is, there is an idea that this singularity point will arrive, where it will become at
1:43:34
1 hour, 43 minutes, 34 seconds
least as smart as we are, and perhaps even smarter. Well, that is, as soon as this point is reached, it will begin to rapidly, yes, there will be an acceleration in development. So. And, uh, you and I also talked about
1:43:45
1 hour, 43 minutes, 45 seconds
rising unemployment, which will only grow and so on, and so on, and so on, because people have no benefit in doing anything and going somewhere to work. In the current
1:43:54
1 hour, 43 minutes, 54 seconds
world paradigm, unemployment will grow even more. A startup in San Francisco, like with cars
1:43:59
1 hour, 43 minutes, 59 seconds
in Waymo, right? 10 million people, bam, just instantly disappear. And so I tried to instantly reach
1:44:07
1 hour, 44 minutes, 7 seconds
10 million people in a couple of years. This is instantaneous in our time, instantly in a couple of years. So, and it’s clear that if we go back and take a kind of excursion into history, we will see
1:44:16
1 hour, 44 minutes, 16 seconds
that when, for example, the technological revolution was happening and certain things were happening, then, in general, yes, a person at a machine tool is no longer needed. Millions of people were left without work,
1:44:24
1 hour, 44 minutes, 24 seconds
automated machines arrived, but these people went to learn how to repair these machines, I don’t know, or program programs to operate them, and so on.
1:44:34
1 hour, 44 minutes, 34 seconds
In the current paradigm, when AI does create gigantic potential unemployment, where
1:44:40
1 hour, 44 minutes, 40 seconds
, in your opinion, will these people convert? Let’s try to simply draw a picture. Do you want to give a different answer or what? Let’s ask you, am I right in understanding that you’re somehow in sync,
1:44:48
1 hour, 44 minutes, 48 seconds
right? In my mind, what answer have you already read, what answer does he want to give? I heard him and understood and understood that
1:44:55
1 hour, 44 minutes, 55 seconds
he… So there are two points, and everything depends very much on the speed at which AI develops.
1:45:05
1 hour, 45 minutes, 5 seconds
No one knows the answer to this question, so we need to analyze two scenarios. And in both scenarios
1:45:13
1 hour, 45 minutes, 13 seconds
, what we’re talking about will play a key role in whether this will be very bad for everyone or great. Yeah.
1:45:20
1 hour, 45 minutes, 20 seconds
One is that AI is developing not so quickly, and not instantly. And I, for example, say, a decade. Well,
1:45:31
1 hour, 45 minutes, 31 seconds
well, 20 years is what it’s called. And it dramatically increases the productivity of every person who
1:45:37
1 hour, 45 minutes, 37 seconds
can afford to use it. Yeah. And in this world, in this world, the question of
1:45:46
1 hour, 45 minutes, 46 seconds
whether you have access to it, whether you can afford it, will determine
1:45:51
1 hour, 45 minutes, 51 seconds
whether you have a job or don’t have a job, whether you’re homeless or something, and in such a world, and again,
1:46:00
1 hour, 46 minutes
there’s a very high risk that this tool will be in a small number of hands, and it will be given out
1:46:10
1 hour, 46 minutes, 10 seconds
proportionally to your social rating. Your social rating in some way. A and a and in
1:46:17
1 hour, 46 minutes, 17 seconds
In this regard, for example, there’s a programmer who has access to this and a programmer who doesn’t, and you’re simply in two different leagues. In two
1:46:26
1 hour, 46 minutes, 26 seconds
different leagues, you’re simply second. How easy is it to justify that? Well, imagine if it’s so dangerous that a person with this thing can do anything in a day, then how would people be licensed to access this thing?
1:46:39
1 hour, 46 minutes, 39 seconds
Is it even socially safe for society to be given access to this thing that can
1:46:45
1 hour, 46 minutes, 45 seconds
program a new operating system in a day or not? It’s actually good to imagine
1:46:51
1 hour, 46 minutes, 51 seconds
this future. Yeah. And it would be good to imagine not that crazy future in which this superintelligence, which is everything, can already enslave all of humanity,
1:47:00
1 hour, 47 minutes
but simply imagine a future in which there is artificial intelligence.
1:47:05
1 hour, 47 minutes, 5 seconds
This superintelligence, which can, which may still not have its own will. Yeah, it just does what you say, like the J-chat. You wrote to him: «Do this.» And he does it.
1:47:14
1 hour, 47 minutes, 14 seconds
And like a magic wand, he can do everything. Well, that is, he has all the knowledge of humanity,
1:47:19
1 hour, 47 minutes, 19 seconds
the ability to think rationally, and add to that the arms of a robot, with which
1:47:28
1 hour, 47 minutes, 28 seconds
you can physically do anything with any level of precision. Maybe one
1:47:35
1 hour, 47 minutes, 35 seconds
that can perform eye surgery will cost a million, but one that can
1:47:40
1 hour, 47 minutes, 40 seconds
pour me a glass of water will cost 10,000 bucks. Hey, maybe even less, because now the number of competitors producing them is simply catastrophically large.
1:47:50
1 hour, 47 minutes, 50 seconds
We have friends of ours who have been working on robots for, well, 15 years, I don’t know how many. It’s just that it was very early. I have a friend who also does this, and, well, you know,
1:47:59
1 hour, 47 minutes, 59 seconds
Tyoma Skolova, who makes a manoid. Tyomykov made real working prototypes in seven months, yes, yes. Working prototypes in seven months. Here’s the current speed, which our friends of ours
1:48:09
1 hour, 48 minutes, 9 seconds
were making robots a long time ago, so it was too early, so they had several waves of attempts, lost again, lost again, went to work for someone else, went back. Now he
1:48:19
1 hour, 48 minutes, 19 seconds
has a database of 500 companies producing robots in his own database. The competition
1:48:24
1 hour, 48 minutes, 24 seconds
will be insane. Well, and it will be insane if the brains of these systems are equally accessible,
1:48:33
1 hour, 48 minutes, 33 seconds
right? And if the brains fly away, imagine everyone makes phones, everyone makes phones, but
1:48:39
1 hour, 48 minutes, 39 seconds
only Apple has access to the internet, right? Who will win this market? Apple. And now, if the internet
1:48:45
1 hour, 48 minutes, 45 seconds
is accessible to everyone, then everyone will sell. This is the difference between these two futures. In the past, it seems like the same thing
1:48:51
1 hour, 48 minutes, 51 seconds
is happening: this AI, which can do everything, but in one world, it can do everything, and it is
1:48:58
1 hour, 48 minutes, 58 seconds
equally accessible to everyone. And then this unemployment, this tremors, they will, we should say, this world.
1:49:06
1 hour, 49 minutes, 6 seconds
If a robot costs $5,000, then in this world today, in today’s world, there are a billion people, at least a billion people, who can afford to go and buy one right now.
1:49:17
1 hour, 49 minutes, 17 seconds
That is, let’s say, I just come home and don’t waste time cooking . I pressed the button, I said: «Make me an omelet.» Zab, forget about food. It’s going
1:49:26
1 hour, 49 minutes, 26 seconds
to work at the factory for you. It’s your robot. It’s your robot. It’s going to control the intersection for you. It
1:49:33
1 hour, 49 minutes, 33 seconds
‘s going to walk the dog for you. It’s for you, imagine that it works 24 hours. Okay,
1:49:39
1 hour, 49 minutes, 39 seconds
fine. Not 24, 20 hours, four hours on charge. So it’s charging for four hours and there’s some kind of myth that
1:49:44
1 hour, 49 minutes, 44 seconds
another robot will fix it. At that moment, another robot will fix it. At that moment, it works 20 hours, divided
1:49:49
1 hour, 49 minutes, 49 seconds
into sectors, roughly, by a third, roughly. There, for 7 hours, it works at your home, doing everything
1:49:56
1 hour, 49 minutes, 56 seconds
for your home. For 7 hours, it works for you at your job, and the other 7 hours, it
1:50:05
1 hour, 50 minutes, 5 seconds
works for social purposes, like some kind of cooperative thing, like building roads, factories, building electricity,
1:50:11
1 hour, 50 minutes, 11 seconds
laying what you’re talking about now. It doesn’t really affect unemployment. No, it changes the essence of unemployment
1:50:17
1 hour, 50 minutes, 17 seconds
. You no longer need to work. You no longer need to work because your robot works for you. But this version is only valid if the brains
1:50:28
1 hour, 50 minutes, 28 seconds
They don’t control it. If they control the brain, it doesn’t look like that. So, you have a robot from Google,
1:50:34
1 hour, 50 minutes, 34 seconds
which they only make for you if you… It costs a million dollars, it costs a million, they rent it out to you and squeeze as much money out of you as you’re willing to pay for it. So, you
1:50:44
1 hour, 50 minutes, 44 seconds
know, this is, uh, interesting, uh, there are now several models of self-driving
1:50:49
1 hour, 50 minutes, 49 seconds
cars. There’s Weima, yes, you can’t buy anything there at all. Everything, it’s just rented out,
1:50:55
1 hour, 50 minutes, 55 seconds
and there’s a very clear path there. They’re going to kill Uber, Lif, and all the other companies now and can set any price, and then they’ll set any price they want. I’m sorry, I’ll remember
1:51:05
1 hour, 51 minutes, 5 seconds
the thought, but Uber, they had the idea that they would only become profitable when they appeared, but they simply won’t. Because they don’t have the real deal, yes, they just couldn’t
1:51:15
1 hour, 51 minutes, 15 seconds
build it that way, yes, and accordingly, they simply couldn’t build them. Why? Because Google sued them for allegedly having Google files there, although the supposedly
1:51:25
1 hour, 51 minutes, 25 seconds
proven files were a presentation from a weekend meeting that didn’t actually mention anything about technology,
1:51:31
1 hour, 51 minutes, 31 seconds
but Google sued them and forced them to stop doing it. And in the end,
1:51:37
1 hour, 51 minutes, 37 seconds
they own this Weima thing, right? There’s this model, and then there’s this supposed alternative
1:51:44
1 hour, 51 minutes, 44 seconds
from Elon Musk. He’s telling the same story as Dania,
1:51:49
1 hour, 51 minutes, 49 seconds
that every car is a machine that will be controlled by this network, this robotics, and that will earn money for you. Yes, the only thing he doesn’t mention is that you don’t own this AI.
1:52:01
1 hour, 52 minutes, 1 second
Even when you buy another car, you can’t transfer this AI to the new car.
1:52:06
1 hour, 52 minutes, 6 seconds
So, as soon as your car breaks down, that’s it, they won’t sell cars, they’ll just
1:52:11
1 hour, 52 minutes, 11 seconds
lease them out. And in this regard, he’s now, he’s even saying that robot taxis will be
1:52:17
1 hour, 52 minutes, 17 seconds
separate, and he has a two-trillion-dollar project to fly to Mars, so he will
1:52:22
1 hour, 52 minutes, 22 seconds
feel absolutely committed to humanity. I charge humanity the entire price for the taxi service
1:52:29
1 hour, 52 minutes, 29 seconds
I need to build rockets and fly to Mars. That is, if we don’t have it , the end justifies the means, even morally justified. And if you don’t have it, now he’s
1:52:39
1 hour, 52 minutes, 39 seconds
in power, so he has all the levers to do it. All the levers are there. Therefore, there will be a shadow of an alternative, when the brain of this system is open source, open code that is available to everyone,
1:52:51
1 hour, 52 minutes, 51 seconds
which is calculated by a decentralized network, then this is the future. Then unemployment in reality has the same meaning, the same significance, as it has today. So. A future
1:53:01
1 hour, 53 minutes, 1 second
where everyone has a robot capable of doing any job is a future where, even if for some reason we don’t get paid, we’ll all get together and build everything
1:53:11
1 hour, 53 minutes, 11 seconds
we need ourselves, because our robots will do it for us. They know how to be a doctor, how to be a teacher, how to build a house, how to cook. The very fabric of the economy is different,
1:53:20
1 hour, 53 minutes, 20 seconds
I understand. But it seems to me that, like any global change in society and the economy, it will always begin with a peak, right? That is, at first, it will be like these scenarios
1:53:29
1 hour, 53 minutes, 29 seconds
where it’s inaccessible to everyone, and so on, and then everything will gradually penetrate into the slump. Maybe, maybe, maybe not. Well, look, let’s assume a scenario where it
1:53:37
1 hour, 53 minutes, 37 seconds
happens the way we’re used to, where at first it’s inaccessible to everyone, there’s massive social stratification, rising unemployment, and so on. What, in your opinion, will happen
1:53:44
1 hour, 53 minutes, 44 seconds
to people and the economy at that point ? What will they do? I think we ‘ll see things a little differently in different countries. Yeah. And in the United States, it’ll all be heavily
1:53:54
1 hour, 53 minutes, 54 seconds
regulated. And the income of these companies, which will be distributed as some kind of benefit
1:54:00
1 hour, 54 minutes
, will only care about the citizens of the United States, America, while extracting
1:54:06
1 hour, 54 minutes, 6 seconds
resources from the rest of the world. If you want ours, pay up. And inside the United States, there’s a kind of socialism and
1:54:13
1 hour, 54 minutes, 13 seconds
reality in that regard. And so, when we fly around the world and meet with ministers from different
1:54:22
1 hour, 54 minutes, 22 seconds
countries, they all understand this, they all understand that this is a question of sovereignty. If they don’t have
1:54:28
1 hour, 54 minutes, 28 seconds
This resource, then what will happen in 10 years? Well, this also has a very strong impact on defense,
1:54:34
1 hour, 54 minutes, 34 seconds
because if you have robots, say, and another country doesn’t, the issue is closed. The issue is closed. Fractional, robots, it’s all sort of determined by AI, but it’s all about sovereignty.
1:54:47
1 hour, 54 minutes, 47 seconds
If there’s no alternative, then there will be countries where, well, it’s
1:54:52
1 hour, 54 minutes, 52 seconds
good to distribute it somehow. And in general, in America,
1:54:58
1 hour, 54 minutes, 58 seconds
unemployment rates are quite high even today. When everyone was put on COVID-19, they were paying an extra 10,000, I think, right? Those were crazy people, they were paying crazy amounts of money there. And people weren’t earning that much.
1:55:08
1 hour, 55 minutes, 8 seconds
I have, I think I told you about that once. My guys, uh, in New York, we have a big company that does, like, catering, and they do all the big events. They have thousands of people there, and
1:55:18
1 hour, 55 minutes, 18 seconds
after COVID, people simply said, «No, I don’t want to work anymore, my benefits are more than I was earning with you, so to speak.» And there was a huge labor shortage after that. And
1:55:26
1 hour, 55 minutes, 26 seconds
the effects of it are still lingering. And so there will be countries where all this won’t be so depressing.
1:55:35
1 hour, 55 minutes, 35 seconds
And there will be countries where pension funds have invested in Open AI XAI. So, for example, we know such countries, yes. Well, even Saudi Arabia or the Emirates, they
1:55:44
1 hour, 55 minutes, 44 seconds
invest a lot. They understand that the oil will run out, but this, on the contrary,
1:55:49
1 hour, 55 minutes, 49 seconds
will remain. And they hope through this investment to get money back into this fund and distribute it among their citizens. Well, there will be countries where there will be significant…
1:56:02
1 hour, 56 minutes, 2 seconds
well, if you travel through Africa, people there are still using these same rakes to get their own food, and there is one
1:56:10
1 hour, 56 minutes, 10 seconds
car for every 10 villages. And it could actually look the same as
1:56:20
1 hour, 56 minutes, 20 seconds
these upheavals did in previous periods. And it will be enough, well, it will be very difficult for people to adapt. People
1:56:29
1 hour, 56 minutes, 29 seconds
have never actually encountered change at such a speed,
1:56:36
1 hour, 56 minutes, 36 seconds
and so it will even be difficult for them. Well, I mean, imagine, my idea, I can imagine very easily, so I can draw up the whole chain, how this could happen, with regulations,
1:56:45
1 hour, 56 minutes, 45 seconds
with restrictions, and so on. That in 30 years, right now, one robot could be produced for every
1:56:52
1 hour, 56 minutes, 52 seconds
person on planet Earth? Yeah. I mean, it’s hard to imagine. It’s unclear how this is… Look,
1:56:59
1 hour, 56 minutes, 59 seconds
how much time has passed, like the iPhone? That’s it. No, that’s not what I mean. I mean, it’s hard to imagine what the world will look like at that moment. Yes. I mean, what they will produce, you can imagine.
1:57:07
1 hour, 57 minutes, 7 seconds
But what this world will look like, that’s very difficult to imagine. And I’m kind of painting the picture
1:57:12
1 hour, 57 minutes, 12 seconds
we’re painting, that given the choice of what we’re starting, well, betting on decentralization and
1:57:21
1 hour, 57 minutes, 21 seconds
open source, yes, it’s a completely different world. This is a world in which, uh, humanity’s main problem
1:57:28
1 hour, 57 minutes, 28 seconds
will be the meaning of life, choice. Choice. What else do I want my robot to do for me today? Hm, it’s
1:57:37
1 hour, 57 minutes, 37 seconds
a difficult philosophical question, because if we take it again, we can dig quite deeply into history, it’s clear that this particular technological progress
1:57:47
1 hour, 57 minutes, 47 seconds
most likely didn’t exist, at least not that we know of. Ah, and there
1:57:52
1 hour, 57 minutes, 52 seconds
wasn’t such a direct shift. But if we simply look at the structure of people’s psyches, yes, people in power, including
1:57:58
1 hour, 57 minutes, 58 seconds
states as such, they always favor a different outcome, for power in the hands of a single person,
1:58:05
1 hour, 58 minutes, 5 seconds
for total dominance, and so on, and so on, and so on. And here’s your model, in which everything
1:58:11
1 hour, 58 minutes, 11 seconds
will be equal for everyone, open source, and people will think about the meaning of life. Honestly, I find it really hard to believe and even imagine right now, because it contradicts the way
1:58:21
1 hour, 58 minutes, 21 seconds
the world works. Sasha, look, here’s the thing. The whole thing right now is that they’re all outsiders in
1:58:28
1 hour, 58 minutes, 28 seconds
this sense. So we’re talking about the US and China, which, like China, maybe,
1:58:34
1 hour, 58 minutes, 34 seconds
yes, everyone else, and the only thing left is elections. And the only thing left for the rest is elections. So, this completely changes the rules of the game. They’ll all be for us to happen. They’ll
1:58:44
1 hour, 58 minutes, 44 seconds
support it with all their might, so that it… I’m not even talking just about you, I’m talking about the whole idea. I’m not even talking about maybe 100 more people besides you who will try to create it. I’m just talking about…
1:58:52
1 hour, 58 minutes, 52 seconds
That this is what changes the course of history, let’s call it that. And there are already examples. And in the late nineties,
1:59:01
1 hour, 59 minutes, 1 second
all servers were controlled by various proprietary operating systems. These were Microsoft, Sun,
1:59:12
1 hour, 59 minutes, 12 seconds
Microsystems, Oracle. You mean before a personal computer appeared in every home? No, wait, wait. When a personal computer
1:59:20
1 hour, 59 minutes, 20 seconds
appeared in every home by 2021, that’s it, we’re already in 2000, 2001, everyone already has it.
1:59:30
1 hour, 59 minutes, 30 seconds
Why? Because they figured that everyone had enough private computers that the total number of private computers would be greater than the total number of corporate computers. It was all
1:59:40
1 hour, 59 minutes, 40 seconds
controlled by corporations. But even in 2001, Linux was still less than 1% of the server market. Of the
1:59:49
1 hour, 59 minutes, 49 seconds
server market. Right. And today it’s 75th, 75%. Oh, that is, in fact, we already have it in history
1:59:56
1 hour, 59 minutes, 56 seconds
, and it’s just like Linux Linux, people don’t understand what Linux is. Everyone, everyone, what?
2:00:03
2 hours, 3 seconds
All Android phones are Linux. But I wanted to say something else. All the protocols
2:00:09
2 hours, 9 seconds
that run the internet, all the protocols that run the web, the web, languages ​​in general, uh,
2:00:15
2 hours, 15 seconds
servers, web servers, mail servers—all of this, all of this is open source. It was so successful precisely
2:00:24
2 hours, 24 seconds
because a very small group of people back then came up with Lee Linus Torvalds in 1991, who came up with the idea that an alternative to closed operating
2:00:34
2 hours, 34 seconds
systems was needed and decided to gather a group of enthusiasts to write Linux. And it’s hard to imagine now
2:00:42
2 hours, 42 seconds
that exactly the same situation existed back then. Five American corporations controlled all
2:00:50
2 hours, 50 seconds
the world’s servers, and all home computers were running them. And now 90% of all phones run open source Linux.
2:00:57
2 hours, 57 seconds
Yeah, right. 75% of all the servers we use run open source Linux. The entire internet is open
2:01:05
2 hours, 1 minute, 5 seconds
source. Someone might say, «Well, now all this is supported by corporations, that means,
2:01:10
2 hours, 1 minute, 10 seconds
no, well, there are a lot of these corporations, and they are sort of different, centralization, but but
2:01:17
2 hours, 1 minute, 17 seconds
the second change happened. Why, for example, Cloud? Why are we now talking about Cloud GPU? Yeah. In order to
2:01:24
2 hours, 1 minute, 24 seconds
support AI. Why then didn’t cloud computing also become
2:01:29
2 hours, 1 minute, 29 seconds
decentralized, because in fact there was no Bitcoin framework? Yeah. Cloud, Amazon,
2:01:35
2 hours, 1 minute, 35 seconds
Google, GCP, Microsoft’s openwork. This is the beginning, this is 2008. This is what laid the foundation for this. That
2:01:45
2 hours, 1 minute, 45 seconds
is, open source had these enthusiasts, but there was no model to make money,
2:01:52
2 hours, 1 minute, 52 seconds
to finance the infrastructure, to finance physical hardware. And Bitcoin appeared in parallel with
2:01:58
2 hours, 1 minute, 58 seconds
Cloud. And during Cloud, yeah, during Cloud’s development, Bitcoin built an infrastructure larger
2:02:05
2 hours, 2 minutes, 5 seconds
than Cloud. Just for some coin, some incomprehensible coin, which used to It was unclear,
2:02:12
2 hours, 2 minutes, 12 seconds
and now it’s become 10% of gold. Look, wait, I have a question. But from a purely technical standpoint, why
2:02:18
2 hours, 2 minutes, 18 seconds
can’t we, uh, stretch the Bitcoin processing power to something like that and it will stay that way? Just shift slowly, right? No,
2:02:27
2 hours, 2 minutes, 27 seconds
well, you’ll just have to remove the hardware. Hardware, it’s not suitable for its number. Ah, that is, the current hardware is not suitable. It is suitable, listen, so what is hardware? What is jelly? Yeah, wait,
2:02:35
2 hours, 2 minutes, 35 seconds
wait, wait. The building is built, it’s hardware. I understand, of course. The power plant buildings are built.
2:02:44
2 hours, 2 minutes, 44 seconds
I understand that this is a huge infrastructure. This is a huge infrastructure. Only the chips there are replaced every couple of years. And the next change in the next couple of years will lead to the fact
2:02:54
2 hours, 2 minutes, 54 seconds
that 10, 20, 30% of Bitcoin’s processing power and data centers will be occupied not by Bitcoin machines,
2:03:04
2 hours, 3 minutes, 4 seconds
but by AI machines. These are the first changes that we, for example, made. That is, the first thing we made in the protocol is The fact is that in Bitcoin, the computation is a so-called
2:03:13
2 hours, 3 minutes, 13 seconds
hash function. It doesn’t matter whether our listeners understand this, but this specific function
2:03:19
2 hours, 3 minutes, 19 seconds
isn’t like a computation, it’s different. And the computations are all very standardized,
2:03:26
2 hours, 3 minutes, 26 seconds
they’re all very similar. Umbrella multiplication. It’s specific
2:03:33
2 hours, 3 minutes, 33 seconds
mathematics. That’s it. And so, if you build a protocol, because the hardware that does this kind of computation wins, then the same innovation will occur.
2:03:44
2 hours, 3 minutes, 44 seconds
It’s just different computations, they’re slightly different. Even now, among those people who worked on ASX before, there are specialized chips called ASX for Bitcoin, and among them…
Chapter 23: Bitcoin has its own role
2:03:58
2 hours, 3 minutes, 58 seconds
Dozens of companies have been working on new Transformer architectures and designs for the past year or two.
2:04:07
2 hours, 4 minutes, 7 seconds
Oh, and some of them, who are just now releasing chips,
2:04:12
2 hours, 4 minutes, 12 seconds
will release chips next year, are making some bold claims—how do you say «bolt» in Russian? Like, some kind of firm,
2:04:18
2 hours, 4 minutes, 18 seconds
some kind of harsh claim—that they’re 10 to 80 times more efficient at performing LLM operations
2:04:27
2 hours, 4 minutes, 27 seconds
than Invidiva’s top-end chips, which are already 10 times higher. And that’s Bitcoin speed. If we increase it by 20% every year
2:04:36
2 hours, 4 minutes, 36 seconds
, that’ll be 20 times—wait, not 20 times—that’ll be
2:04:43
2 hours, 4 minutes, 43 seconds
300,000 times over 15 years. Yeah. Yeah. Yeah. Yeah, it’s coming. It has this paradigm shift, a shift that
2:04:52
2 hours, 4 minutes, 52 seconds
doesn’t weaken Bitcoin’s position. Well, look, we think Bitcoin has its role. This
2:04:56
2 hours, 4 minutes, 56 seconds
gold, it can’t be moved. Yeah. So, it’s just the base. Even if Bitcoin,
2:05:04
2 hours, 5 minutes, 4 seconds
all of it, were simply transferred to another blockchain consensus that doesn’t require all
2:05:13
2 hours, 5 minutes, 13 seconds
that electricity, Bitcoin wouldn’t disappear. Yeah. It’s all gold now. But the fact
2:05:19
2 hours, 5 minutes, 19 seconds
that there’s gold floating around somewhere on stroids, and we could soon reach it, has n’t yet brought down the price of gold on planet Earth, despite the fact that there’s more gold there than on the
2:05:29
2 hours, 5 minutes, 29 seconds
entire planet Earth. There definitely is. And we’ve even already seen it in telescopes and know for sure that we’ll get there. But they’re not polling that right now. Yeah. Well, that’s understandable, because that’s a certain belief
2:05:39
2 hours, 5 minutes, 39 seconds
of Skersity at the moment. The number of Bitcoins won’t increase, even if they were transferred, say, to, I don’t know, the Salan network. That’s it. Rather, what will happen is that
2:05:49
2 hours, 5 minutes, 49 seconds
another protocol will emerge that doesn’t claim to be gold, but rather claims to be
2:05:55
2 hours, 5 minutes, 55 seconds
the AI ​​fuel of the future. And this might not be our protocol. It might
2:06:04
2 hours, 6 minutes, 4 seconds
not be our protocol. But if we look at all the current alternatives to our protocol, yes, we
2:06:12
2 hours, 6 minutes, 12 seconds
would advise them all to reconsider the very essence of what they’re doing and restructure around Proof of War, because Bitcoin was built on this model.
2:06:24
2 hours, 6 minutes, 24 seconds
The more work the infrastructure does, the more. And here we directly need infrastructure.
2:06:29
2 hours, 6 minutes, 29 seconds
And all the other projects are now in blockchain. The most famous is this one, 60% of all issued Bitter coins
2:06:38
2 hours, 6 minutes, 38 seconds
go to validators and subnet owners. It has a very complex structure. It’s a real mess. There, you’re
2:06:45
2 hours, 6 minutes, 45 seconds
the one who invested the capital, not the one who provided the best equipment. It’s a pain. And in this regard, they
2:06:52
2 hours, 6 minutes, 52 seconds
don’t lead to this cycle where it makes sense to expand, update, and add equipment, and that’s clear. I understand, of course, the difference is very clear. And in this regard,
2:07:01
2 hours, 7 minutes, 1 second
yes, there are already existing decentralized projects, but they’re more like an idea,
2:07:11
2 hours, 7 minutes, 11 seconds
rather than interesting conceptual projects, but they can’t get the same flywheel going.
2:07:20
2 hours, 7 minutes, 20 seconds
And we’re currently assembling a group; this will be our third trip. There’s this country called
2:07:32
2 hours, 7 minutes, 32 seconds
Bhutan. Yeah. And there’s a question about taking a step back, right? It’s well-known there, not very well-known,
2:07:40
2 hours, 7 minutes, 40 seconds
but it’s well-known in the crypto industry because it’s the third-largest country in bitcoin mining,
2:07:47
2 hours, 7 minutes, 47 seconds
in terms of state ownership of bitcoins. Because they saw this opportunity; they have a
2:07:55
2 hours, 7 minutes, 55 seconds
lot of electricity, hydroelectricity, the country had a lot, and they realized that they were kind of squeezed between China and India. It’s hard for them to sell to anyone other than China and India. These are poor
2:08:05
2 hours, 8 minutes, 5 seconds
countries that are buying up their electricity cheaply. But can this electricity be sold to
2:08:11
2 hours, 8 minutes, 11 seconds
the Bitcoin community? Mine Bitcoin. They signed a contract with Bitcoin, brought
2:08:18
2 hours, 8 minutes, 18 seconds
these specialists to Bhutan, and made it for free among themselves. Who gets the Bitcoin, and with almost
2:08:24
2 hours, 8 minutes, 24 seconds
free, but huge amounts of electricity, hydroelectricity, yes,
2:08:31
2 hours, 8 minutes, 31 seconds
they mined about $3 billion in Bitcoin and Litecoin, and for a very small country with a small population. What are you talking about? That
2:08:36
2 hours, 8 minutes, 36 seconds
the state is actually no longer just Bhutan, well, the whole
2:08:42
2 hours, 8 minutes, 42 seconds
world has already figured out how Bitcoin has fulfilled its main goal: it has educated the entire world on how
2:08:52
2 hours, 8 minutes, 52 seconds
decentralized blockchain technologies work. So we’re gathering a group of people there, accordingly . We gathered it in February for my birthday. It coincides with my birthday there.
2:09:05
2 hours, 9 minutes, 5 seconds
The king, so there was a very pleasant group there, where we brought together Silicon
2:09:12
2 hours, 9 minutes, 12 seconds
and crypto entrepreneurs who were lured by Bitcoin to see
2:09:18
2 hours, 9 minutes, 18 seconds
the full scale of all these hydroelectric power plants, everything that’s being built there. And the main
2:09:25
2 hours, 9 minutes, 25 seconds
topic there is also what the next stage of this process is, when will it be AI computing? Yeah.
2:09:39
2 hours, 9 minutes, 39 seconds
In order to address the very disparate ones, our next cooperation with Bhutan will be aimed at bringing together AI manufacturers,
2:09:50
2 hours, 9 minutes, 50 seconds
future AI chips, ASICs for Transformers, in order to negotiate this protocol. That
2:09:58
2 hours, 9 minutes, 58 seconds
is, the government is already moving in this direction quietly, without panic. Some are already
2:10:06
2 hours, 10 minutes, 6 seconds
panicking, but overall they are already moving. And this will, in fact, be quite interesting and beautiful
2:10:12
2 hours, 10 minutes, 12 seconds
international cooperation at some point. These are precisely the countries that have no options. Countries that have no options. And we will be surprised that almost all the large,
2:10:23
2 hours, 10 minutes, 23 seconds
strange countries don’t have these options. Exactly. And just like this is happening right now, in the next 2-5
2:10:31
2 hours, 10 minutes, 31 seconds
years, this shift will happen, because one sequence, at first you really think about, oh, this has too low a probability, rather, the probability of this
2:10:41
2 hours, 10 minutes, 41 seconds
scenario, where everything is concentrated, is higher. But as soon as you really imagine this future, it becomes so scary that you’re like: “Okay, I’ll at least put a little bit of my resources here.” And
2:10:52
2 hours, 10 minutes, 52 seconds
that’s enough. That’s it. Ah, but what’s stopping me from putting 1% of my resources into this thing? And this
2:10:59
2 hours, 10 minutes, 59 seconds
1% is already enough so that, ah, growth will then go up by 2, 5, 10%, 20%. And an alternative
2:11:07
2 hours, 11 minutes, 7 seconds
can happen. So, if I were to very briefly summarize in three paragraphs where we will end up
Chapter 24: Is a world of great upheaval and abundance waiting for us?
2:11:15
2 hours, 11 minutes, 15 seconds
in the long term—I don’t know, it’s difficult to outline any horizon, in the long term of 5-10 years—
2:11:21
2 hours, 11 minutes, 21 seconds
what kind of world will this be, in your opinion? This will be a world of great upheaval
2:11:29
2 hours, 11 minutes, 29 seconds
, in that the world will change dramatically, and many people could suffer greatly from
2:11:36
2 hours, 11 minutes, 36 seconds
their inability to adapt. At the same time, this is an incredible opportunity to
2:11:49
2 hours, 11 minutes, 49 seconds
improve the general well-being of the masses, the well-being of each of us, when things that are currently only available to the world’s richest people will become available to us.
2:12:03
2 hours, 12 minutes, 3 seconds
Even those who are already outside the market today, unable to break into it at all, those who
2:12:09
2 hours, 12 minutes, 9 seconds
dig the earth and have no other alternative. Even for them, this world will suddenly become a world
2:12:15
2 hours, 12 minutes, 15 seconds
in which they will live as well as people in developed countries. And this can be seen a little
2:12:21
2 hours, 12 minutes, 21 seconds
by their phones. There was a time when rich people had special phones
2:12:27
2 hours, 12 minutes, 27 seconds
and so on, with diamonds and gold, and now they have the same phone. And overall, it’s not much
2:12:32
2 hours, 12 minutes, 32 seconds
different from the Android that people have, yes, the cheapest Android, which people in
2:12:37
2 hours, 12 minutes, 37 seconds
Indonesia have, and they earn much less money. And the same thing will happen with a robot
2:12:43
2 hours, 12 minutes, 43 seconds
. It can bring the same thing, only a robot and AI change your ability
2:12:49
2 hours, 12 minutes, 49 seconds
to earn money, change your ability to do work, create things, build you your
2:12:54
2 hours, 12 minutes, 54 seconds
own home, and so on. And as for accessibility, well, yes, maybe you, a
2:13:00
2 hours, 13 minutes
rich person, will have a robot for 100,000, and it can do cool things very presciently, absolutely. And they will have the same thing, only for 5,000 dollars for one per village. And well, it will
2:13:11
2 hours, 13 minutes, 11 seconds
do it for them in such a way that it’s the same thing, but not stylish. Their life will be radically an order of magnitude better
2:13:16
2 hours, 13 minutes, 16 seconds
than today. So, these two things are there. And, ah, we actually have a concern that the world is still
2:13:24
2 hours, 13 minutes, 24 seconds
the future, you know, as they say, the future is already here, but it’s unevenly distributed, then, that anyway, what we’re talking about will be uneven at the beginning, at least to a certain extent.
2:13:35
2 hours, 13 minutes, 35 seconds
In some places there will be major upheavals, and in others we’ll see this, aa, growth in a country where people will
2:13:42
2 hours, 13 minutes, 42 seconds
think more about this, Bhutan and so on, yes, already in the early stages, in fact,
2:13:50
2 hours, 13 minutes, 50 seconds
abundance will grow significantly. Ah, but at the same time, of course, each of us today needs to spend at least 5% of our attention
2:14:00
2 hours, 14 minutes
on this issue. What if a world emerges in which AI can produce everything I can?
2:14:08
2 hours, 14 minutes, 8 seconds
Produce better? Better. Cool. A quick blitz, a short question, a short answer. Let’s take it one at a time. If AI becomes smarter than humans, will that be a victory or a defeat? A victory, of course. Excellent.
Chapter 25: Blitz poll
2:14:21
2 hours, 14 minutes, 21 seconds
If you had access to the future, literally for 10 minutes, what would be the first thing you would watch?
2:14:27
2 hours, 14 minutes, 27 seconds
I would look, well, in general, at how we live, how far we have come, how far we have come. But
2:14:36
2 hours, 14 minutes, 36 seconds
today, I can’t imagine what our life will be like if it doesn’t last, we will see the stars of the galaxy and so
2:14:43
2 hours, 14 minutes, 43 seconds
on. I really want to see when humanity finally got there and whether it will ever get there. And that’s why usually, when people ask me: «What year in the past would you like to travel to?» I say:
2:14:53
2 hours, 14 minutes, 53 seconds
«No, I want to go 10,000 times into the future and see what’s there.» Cool. And what discovery of the last
2:14:58
2 hours, 14 minutes, 58 seconds
two years really changed your thinking? Oh, in the last two years, yes. I came across the company
2:15:07
2 hours, 15 minutes, 7 seconds
Kify. Yeah. And I saw, uh, how AI, we will influence biochemistry, chemistry, the physics of new materials.
2:15:24
2 hours, 15 minutes, 24 seconds
AI will literally change everything in the next two years. New drugs, new materials, new
2:15:29
2 hours, 15 minutes, 29 seconds
batteries at insane speed. It’s all kind of already coming. It’s all already happened. It’s just
2:15:36
2 hours, 15 minutes, 36 seconds
a question of what it will be applied to in the next two years. Well, in the past two years, it’s literally happened. That is, leading chemical companies, leading pharmaceutical companies
2:15:47
2 hours, 15 minutes, 47 seconds
have already begun ordering their new molecules from AI companies. Cool. The main skill of the entrepreneur
2:15:54
2 hours, 15 minutes, 54 seconds
of the future. And in general, I think that this is the main skill today, and that is
2:16:04
2 hours, 16 minutes, 4 seconds
the ability to understand responsibility, no, the ability to exist in uncertainty. Yeah. We
2:16:13
2 hours, 16 minutes, 13 seconds
all really don’t like living in uncertainty. There are a small number of people who, no matter how unpleasant it is, do it. It’s like pain. Some people,
2:16:23
2 hours, 16 minutes, 23 seconds
it seems, tolerate uncertainty more than others, while others can tolerate it. It’s when you have to make a decision
2:16:29
2 hours, 16 minutes, 29 seconds
in circumstances of absolute uncertainty, when the possible outcomes of these events differ so greatly,
2:16:36
2 hours, 16 minutes, 36 seconds
and you still have to navigate them. It’s about responsibility; someone has to take responsibility for making these decisions, because you’re supposed to make it in this
2:16:46
2 hours, 16 minutes, 46 seconds
uncertainty. If everything goes badly, then you’re in for it. If everything goes well, you
2:16:50
2 hours, 16 minutes, 50 seconds
‘ll earn some money and thank them. Well, if everything goes badly, then, as you already mentioned, well, what do you need to do? How do you
2:16:58
2 hours, 16 minutes, 58 seconds
position yourself? How do you think about it? When someone offers you $100 million
2:17:03
2 hours, 17 minutes, 3 seconds
Теперь, и вы говорите: «Ну, мы можем построить другое будущее для человечества». Это сложно.
2:17:10
2 часа, 17 минут, 10 секунд
Выбор. И я думаю, что способность существовать в этой неопределенности и ориентироваться в сценариях, которые
2:17:19
2 часа, 17 минут, 19 секунд
совершенно непредсказуемые, и как они будут разворачиваться, является одним из самых важных
2:17:26
2 часа, 17 минут, 26 секунд
часть предпринимательства. Круто. Большое спасибо. Супер интересно. Я не знаю, кажется, что мы снова разговаривали 2 часа подряд. Этот разговор оказал глубокое влияние на меня с точки зрения моего
2:17:36
2 часа, 17 минут, 36 секунд
Общее восприятие реальности и мои собственные подходы. Это, я уверен, окажет глубокое влияние, потому что то, что мы обсуждали сегодня, так важно, и
2:17:44
2 часа, 17 минут, 44 секунды
Я думаю, что это то, что касается многих людей прямо сейчас. И в целом, я и многие люди сегодня строим свои
2:17:50
2 часа, 17 минут, 50 секунд
Восприятие через призму того, как изменится мир. Это, наверное, самый важный вопрос, который есть у любого здравомыслящего человека прямо сейчас. Так что большое спасибо. Я
2:17:58
2 часа, 17 минут, 58 секунд
Как всегда, был взрыв. Я очень счастлива. Отлично, друзья, за то, что следили за нами до конца. Если вы нашли это полезным, если вам понравилось, обязательно подпишитесь на канал и дайте ему лайк. Самое главное, поделиться
2:18:08
2 часа, 18 минут, 8 секунд
Ваши выводы, идеи и идеи, которые вы почерпнули сегодня в комментариях. Поделиться. Всегда интересно читать и смотреть. Большое спасибо, ребята. Вот и все, [ __ ] здорово.