Lex Fridman Podcast Melanie Mitchell: Concepts, Analogies, Common Sense & Future of AI
22 snips
Dec 28, 2019 
In a thought-provoking discussion, Melanie Mitchell, a professor of computer science, dives into the complex world of artificial intelligence and common sense. She highlights the significance of analogy-making in both human cognition and AI, explaining how simple rules can lead to complex behaviors. The conversation also touches on the challenges of achieving human-like understanding in machines and the existential risks posed by advanced AI. With insights from her book, she explores the future implications of AI technology on society.
AI Snips
Chapters
Books
Transcript
Episode notes
Generative Models and Perception
- Active, generative models are a promising approach in AI, simulating mental models.
- These models generate expectations and predictions, guiding perception and understanding.
DeepMind's Breakout and Conceptual Gaps
- DeepMind's Atari Breakout system, despite superhuman performance, lacked core concepts.
- It couldn't adapt when the paddle moved slightly, highlighting limitations of pure learning.
Challenges of Autonomous Driving
- Autonomous driving is harder than anticipated due to the open-ended nature of the real world.
- Edge cases and unpredictable events pose significant challenges for current systems.
