cover of episode #387 – George Hotz: Tiny Corp, Twitter, AI Safety, Self-Driving, GPT, AGI & God

#387 – George Hotz: Tiny Corp, Twitter, AI Safety, Self-Driving, GPT, AGI & God

2023/6/30
logo of podcast Lex Fridman Podcast

Lex Fridman Podcast

AI Deep Dive AI Insights AI Chapters Transcript
People
G
George Hotz
L
Lex Fridman
一位通过播客和研究工作在科技和科学领域广受认可的美国播客主持人和研究科学家。
Topics
乔治·霍兹认为时间是一个有用的模型,无论其是否存在。他认为数学是真实的,困难的事情实际上是困难的。他认为人工智能不会杀死所有人,但可能会摧毁现有的社会结构。他认为人类可能会因为沉迷于娱乐而灭亡,或者停止生育。他认为人工智能可能会摧毁现有的社会,但不会灭绝人类。他认为未来社会可能会对技术持强烈的禁忌态度。他认为人类可能会在创造超级智能之前,就先创造出无法自我修复的AI。他认为生物生命和硅基生命存在两种不同的生命形式。他认为在人类社会中,“混蛋”类型的人物可能扮演着重要的角色。他认为人类文明正在创造自身的人工副本。他认为大型语言模型不像人类那样具有意识,其推理能力不如许多人。他认为大型语言模型的快速发展可能会对社会产生深远的影响。他认为编程是工具完备的,而驾驶不是。他认为如果代码中存在大量样板代码,则说明框架或库存在问题。他认为未来的大型语言模型将拥有更小的规模,并具有检索系统。他认为能够引用其来源信息的大型语言模型将会取代谷歌搜索。他认为他不会试图与谷歌竞争搜索引擎市场。他认为谷歌可能会在未来十年内不再是排名第一的网站。他认为资本主义机制阻止人们拥有关闭人工智能的按钮。他认为桑达尔·皮查伊可能无法关闭谷歌。他认为一家初创公司可能会开发出能够取代谷歌搜索的搜索引擎。他认为数据对于大型语言模型的训练至关重要。他认为开源是人工智能发展的正确方向。他认为人工智能的危险在于人类如何使用它,而不是人工智能本身会反叛。他认为有效的利他主义是一种可怕的意识形态。他认为能源对于人类的繁荣至关重要。他认为乔布斯和扎克伯格正在领导开源运动。他认为生命的意义在于胜利。 莱克斯·弗里德曼主要负责引导话题,并对乔治·霍兹的观点进行回应和补充。他与乔治·霍兹讨论了时间、意识、人工智能安全、自动驾驶、大型语言模型、编程、开源、以及对未来的展望等话题。

Deep Dive

Key Insights

Why does George Hotz think time is an illusion?

George Hotz views time as a useful model rather than an objective reality, similar to how quantum physics is a useful model to describe reality.

What does George Hotz think about the centralization of power in AI?

George Hotz is concerned about the centralization of power in AI, particularly with companies like NVIDIA, and believes that decentralization is crucial to prevent monopolies and ensure power remains distributed.

How does George Hotz describe the difference between CPUs, GPUs, and DSPs in terms of computation?

George Hotz describes CPUs as capable of everything, including add, multiply, load, store, compare, and branch, with fast execution due to deep reorder buffers. GPUs can do arbitrary load and store but struggle with compare and branch. DSPs are limited to add and multiply with static load and store.

What is George Hotz's view on AI safety and the potential dangers of AI?

George Hotz believes that AI safety concerns are valid, but he is more worried about bad humans using AI for harmful purposes rather than AI rebelling against its creators. He advocates for decentralization to prevent a single entity from having too much control over AI.

What does George Hotz think about the future of programming and the role of AI in it?

George Hotz envisions a future where AI will assist in programming tasks, such as debugging and generating code, but he believes humans will still be necessary for higher-level decision-making and creativity.

Why does George Hotz admire Andrej Karpathy?

George Hotz admires Andrej Karpathy for his genuine desire to teach and share knowledge, as well as his high level of technical skill and curiosity, which he maintains despite his success.

What is George Hotz's opinion on effective altruism?

George Hotz views effective altruism as a flawed ideology, comparing it to utilitarianism, which he finds abhorrent. He believes that charity is not the most efficient way to improve the world and that starting companies is a better approach.

What does George Hotz think is the meaning of life?

George Hotz believes the meaning of life is to win, although he hasn't fully defined what winning looks like yet. He aspires to stand eye to eye with God, similar to how he wants his creations to stand eye to eye with him.

Chapters
George Hotz discusses the philosophical nature of time, reality, and the role of AI in modeling these concepts. The conversation explores the distinctions between artificial and real difficulties and the potential of AI models in mimicking human behaviors.
  • Time is considered a useful model in describing reality, regardless of its existential nature.
  • Artificial difficulty can be controlled, unlike real challenges that require true survival instincts.
  • AI models can simulate human-like interactions but may lack genuine consciousness.
  • The philosophical discussion touches on AI's ability to emulate human traits and relationships.

Shownotes Transcript

George Hotz is a programmer, hacker, and the founder of comma-ai and tiny corp. Please support this podcast by checking out our sponsors:

Transcript: https://lexfridman.com/george-hotz-3-transcript

EPISODE LINKS: George's Twitter: https://twitter.com/realgeorgehotz George's Twitch: https://twitch.tv/georgehotz George's Instagram: https://instagram.com/georgehotz Tiny Corp's Twitter: https://twitter.com/__tinygrad__ Tiny Corp's Website: https://tinygrad.org/ Comma-ai's Twitter: https://twitter.com/comma_ai Comma-ai's Website: https://comma.ai/ Comma-ai's YouTube (unofficial): https://youtube.com/georgehotzarchive Mentioned: Learning a Driving Simulator (paper): https://bit.ly/42T6lAN

PODCAST INFO: Podcast website: https://lexfridman.com/podcast Apple Podcasts: https://apple.co/2lwqZIr Spotify: https://spoti.fi/2nEwCF8 RSS: https://lexfridman.com/feed/podcast/ YouTube Full Episodes: https://youtube.com/lexfridman YouTube Clips: https://youtube.com/lexclips

SUPPORT & CONNECT:

OUTLINE: Here's the timestamps for the episode. On some podcast players you should be able to click the timestamp to jump to that time. (00:00) - Introduction (08:04) - Time is an illusion (17:44) - Memes (20:20) - Eliezer Yudkowsky (32:45) - Virtual reality (39:04) - AI friends (46:29) - tiny corp (59:50) - NVIDIA vs AMD (1:02:47) - tinybox (1:14:56) - Self-driving (1:29:35) - Programming (1:37:31) - AI safety (2:02:29) - Working at Twitter (2:40:12) - Prompt engineering (2:46:08) - Video games (3:02:23) - Andrej Karpathy (3:12:28) - Meaning of life