All Segments
Yann LeCun: Benchmarks for Human-Level Intelligence | AI Podcast Clips
7 segments available
This is a clip from a conversation with Yann LeCun from Aug 2019. New full episodes every Mon & Thu and 2-5 new clips every Tue & Fri. You can watch the full conversation here: https://www.youtube.com/watch?v=SGSOCuByo24 or listen to it here: https://lexfridman.com/yann-lecun/ and subscribe on your podcast app. You can watch other clips on the clips playlist: https://www.youtube.com/playlist?list=PLrAXtmErZgOeciFP3CBCIEElOJeitOr41 or full episodes on the full episode playlist: https://www.youtube.com/playlist?list=PLrAXtmErZgOdP_8GztsuKi9nrraNbKKp4 Note: I select clips with insights from these much longer conversation with the hope of helping make these ideas more accessible and discoverable. Ultimately, this podcast is a small side hobby for me with the goal of sharing and discussing ideas. For now, I post a few clips every Tue & Fri. I did a poll and 92% of people either liked or loved the posting of daily clips, 2% were indifferent, and 6% hated it, some suggesting that I post them on a separate YouTube channel. I hear the 6% and partially agree, so am torn about the whole thing. I tried creating a separate clips channel but the YouTube algorithm makes it very difficult for that channel to grow unless the main channel is already very popular. So for a little while, I'll keep posting clips on the main channel. I ask for your patience and to see these clips as supporting the dissemination of knowledge contained in nuanced discussion. If you enjoy it, consider subscribing, sharing, and commenting. Yann LeCun is one of the fathers of deep learning, the recent revolution in AI that has captivated the world with the possibility of what machines can learn from data. He is a professor at New York University, a Vice President & Chief AI Scientist at Facebook, co-recipient of the Turing Award for his work on deep learning. He is probably best known as the founder of convolutional neural networks, in particular their early application to optical character recognition. 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
Beware of False Promises in AI
Yann LeCun warns against those who claim to have solved artificial general intelligence (AGI) or have AI systems that mimic human brain functions. He emphasizes the importance of benchmarks and practical testing in evaluating AI systems, highlighting that even toy problems can provide valuable insights into machine reasoning and memory.
"you've written advice saying don't get fooled by people who claim to have a solution to artificial general intelligence who claim to have an AI system that worked just like the human brain or who clai..."
The Role of Benchmarks in AI Development
LeCun discusses the significance of benchmarks in AI research, explaining that while new ideas may not have established benchmarks, they are crucial for pushing the field forward. He reflects on the evolution of benchmarks in various domains, including images, audio, and natural language, and the need for standards to evaluate AI capabilities.
"sort of standard you know kind of benchmark if you want it doesn't need to be real so for example many years ago here at fair people you know justin west on board and a few others proposed the the bab..."
Interactive Environments for Testing AI
LeCun explores the concept of interactive environments for training and testing intelligent systems, particularly in robotics. He explains how traditional supervised learning paradigms may not apply when the AI's actions influence the data it encounters, emphasizing the need for new approaches in AI evaluation.
"benchmark I agree that's part of the process it's definition benchmarks as part of the process so what are your thoughts about so we have these benchmarks on around stuff we can do with images from cl..."
Rethinking General Intelligence
LeCun critiques the term 'artificial general intelligence' (AGI), arguing that human intelligence is not truly general but rather specialized. He discusses the impressive adaptability of the human brain and the limitations of AI in replicating this flexibility, suggesting that our understanding of intelligence is often constrained by our own experiences.
"decrease the exploration problem the what if the samples so that creates also a dependency between samples right you you if you move if you can only move it in in space the next sample you're going to..."
The Complexity of Visual Processing
In a thought experiment, LeCun illustrates the complexity of visual processing in humans by discussing the optical nerve's structure and its implications for learning vision. He argues that the brain's local connections are crucial for interpreting visual information, highlighting the challenges AI faces in achieving similar capabilities.
"knowledge somehow ok the knowledge persists so let me take a very specific example yes it's not an example it's more like a a quasi mathematical demonstration so you have about 1 million fibers coming..."
Understanding Entropy and Perception
LeCun delves into the concept of entropy in physical systems, drawing parallels to the limitations of human perception. He explains how our understanding of the world is shaped by the information we can comprehend, suggesting that there are vast realms of knowledge beyond our grasp, which complicates the definition of intelligence.
"yeah yes that's specialization yep it's still now really damn impressive so it's not perfect generalizations not even close no no it's it's it's it's it's not that it's not even close it's not at all ..."
Defining Intelligence Beyond Human Standards
LeCun concludes by discussing the challenges of defining intelligence, particularly in relation to human capabilities. He suggests that while AI may not fit traditional definitions of intelligence, it can still demonstrate impressive problem-solving abilities, urging a broader perspective on what constitutes intelligence in machines.
"we call that heat by the way heat heat so at least physicists call that heat or they call it entropy which is okay you have a thing full of gas right call system for gas right clothes on our coast it ..."