

Gary Marcus: Toward a Hybrid of Deep Learning and Symbolic AI
10 snips Oct 3, 2019
Gary Marcus, a prominent AI expert and founder of Robust.AI, shares his critical insights on the limits of deep learning. He discusses the importance of merging deep learning with symbolic AI to achieve general intelligence. Marcus also explores how human cognitive flaws can foster creativity while posing societal challenges. He emphasizes the need for ethical AI that reflects human values and critiques current systems' handling of abstract concepts. Additionally, he examines how understanding language and memory can inform the development of better AI.
AI Snips
Chapters
Books
Transcript
Episode notes
AI's Impact on Humanity
- AI's place in the food chain is already changing due to automation.
- General intelligence will further shift this as AI takes over tedious and dangerous tasks, changing how we find meaning.
Chess Match with Kasparov
- Gary Marcus lost badly to Garry Kasparov in chess, highlighting machines' computational superiority in certain domains.
- He played Deep Thought (Deep Blue's predecessor) and realized machines' dominance in such games.
Deep Learning Limitations
- Deep learning excels at specific tasks like image recognition but struggles with abstract knowledge.
- It relies on correlations, lacking true understanding of concepts like causality or how objects function.