All Segments
Language or Vision - What's Harder? (Ilya Sutskever) | AI Podcast Clips
6 segments available
Full episode with Ilya Sutskever (May 2020): https://www.youtube.com/watch?v=13CZPWmke6A Clips channel (Lex Clips): https://www.youtube.com/lexclips Main channel (Lex Fridman): https://www.youtube.com/lexfridman (more links below) Podcast full episodes playlist: https://www.youtube.com/playlist?list=PLrAXtmErZgOdP_8GztsuKi9nrraNbKKp4 Podcasts clips playlist: https://www.youtube.com/playlist?list=PLrAXtmErZgOeciFP3CBCIEElOJeitOr41 Podcast website: https://lexfridman.com/ai Podcast on Apple Podcasts (iTunes): https://apple.co/2lwqZIr Podcast on Spotify: https://spoti.fi/2nEwCF8 Podcast RSS: https://lexfridman.com/category/ai/feed/ Ilya Sutskever is the co-founder of OpenAI, is one of the most cited computer scientist in history with over 165,000 citations, and to me, is one of the most brilliant and insightful minds ever in the field of deep learning. There are very few people in this world who I would rather talk to and brainstorm with about deep learning, intelligence, and life than Ilya, on and off the mic. Subscribe to this YouTube channel or connect on: - Twitter: https://twitter.com/lexfridman - LinkedIn: https://www.linkedin.com/in/lexfridman - Facebook: https://www.facebook.com/lexfridman - Instagram: https://www.instagram.com/lexfridman - Medium: https://medium.com/@lexfridman - Support on Patreon: https://www.patreon.com/lexfridman
Segments Timeline
Unity in Machine Learning
Ilya Sutskever discusses the interconnectedness of various AI domains, including computer vision, natural language processing (NLP), and reinforcement learning. He emphasizes the underlying principles that unify these fields, suggesting that advancements in one area can benefit others due to shared foundational concepts.
"so incredibly you've contributed some of the biggest recent ideas in AI in computer vision language natural language processing reinforcement learning sort of everything in between maybe not ganz is t..."
The Future of AI Architectures
Sutskever explores the evolution of AI architectures, noting the shift from diverse models for specific tasks to a more unified approach, particularly with the rise of transformers in NLP. He speculates on the potential for a single architecture to encompass both vision and language tasks, reflecting on the historical fragmentation in AI.
"today when someone writes a paper on improving optimization of deep learning in vision it improves the different NLP applications and it improves the different reinforcement learning applications rein..."
Reinforcement Learning's Unique Challenges
In this segment, Sutskever highlights the distinct nature of reinforcement learning (RL) compared to other AI domains. He discusses the necessity for RL to operate in a dynamic environment where actions influence outcomes, and he anticipates future integrations between RL and supervised learning, envisioning a comprehensive AI system.
"languages well origins I expect I think it's I don't want to be too sure because I think on the commercial you know that is very computationally efficient RL is different RL does require slightly diff..."
Defining Hard Problems in AI
Sutskever challenges the notion of what makes a problem 'hard' in AI, suggesting that difficulty is relative to current benchmarks and human-level performance. He reflects on the complexities of language understanding versus visual perception, indicating that both remain challenging and unresolved in the near term.
"something slightly differently I'd say that reinforcement learning is neither but it naturally interfaces and integrates view the two of them do you think action is fundamentally different so yeah wha..."
The Interplay of Language and Vision
This segment delves into the relationship between language and vision, questioning where one domain ends and the other begins. Sutskever considers the implications of achieving deep understanding in either field and suggests that advancements in one may inherently benefit the other, highlighting the interconnected nature of cognitive tasks.
"think is harder so people like Noam Chomsky believe that language is fundamental to everything so it underlies everything do you think language understanding is harder than visual scene understanding ..."
Human Intelligence vs. AI
Sutskever reflects on the unique qualities of human intelligence, particularly in terms of creativity and humor. He contrasts this with AI's current capabilities, suggesting that while AI can impress with its understanding of images and language, it lacks the depth and unpredictability that human interactions provide.
"oh okay so you'd have a fundamental intuition about how hard language understanding is I think I know I changed my mind that's a language is probably going to be hard I mean it depends on how you defi..."