

Rebooting AI: What's Missing, What's Next with Gary Marcus - TWIML Talk #298
Sep 10, 2019
Gary Marcus, CEO of Robust.AI and an influential voice in AI and cognitive science, dives into the gaps in today's AI landscape. He discusses the importance of integrating cognitive principles, pointing out how infants' instinctive language skills can inspire better machine learning. Marcus highlights the limitations of reinforcement learning, advocating for a foundation in knowledge to enhance AI. He also critiques the narrow focus of current AI, emphasizing the need for trust and safety in autonomous systems for future advancements.
AI Snips
Chapters
Books
Transcript
Episode notes
Cognitive Science Perspective
- Gary Marcus' background is in cognitive science, focusing on child development and language acquisition.
- He believes this perspective offers valuable insights into AI, contrasting with the current data-driven approach.
Child's Chair Climbing
- Marcus' daughter climbed through a chair, demonstrating unsupervised learning without explicit examples.
- This highlights how humans leverage background knowledge and problem-solving, unlike current AI.
Deep Learning's Limitations
- Deep learning excels at supervised learning with large datasets, but lacks true comprehension.
- Marcus advocates for AI that understands, similar to children, rather than just memorizing patterns.