Eye On A.I. #250 Pedro Domingos on the Real Path to AGI
Apr 24, 2025
Pedro Domingos, a computer science professor at the University of Washington and author of 'The Master Algorithm,' discusses the elusive goal of Artificial General Intelligence (AGI). He reveals why deep learning alone won't suffice and how reasoning by analogy might unlock true creativity in machines. Domingos highlights the importance of integrating AI's five foundational paradigms and warns against the hype surrounding current technologies like GPT-4. Dive into his views on evolutionary algorithms and the quest for a unified 'master algorithm' as the key to AI's future.
AI Snips
Chapters
Books
Transcript
Episode notes
Evolutionary Algorithms Explained
- Evolutionary algorithms mimic natural selection using strings that represent solutions like circuits or robots.
- For example, evolving a radio by mutating and crossing strings can produce efficient designs after generations.
Reasoning by Analogy Foundations
- Reasoning by analogy involves solving new problems by adapting solutions from similar past cases.
- Kernel machines, like support vector machines, are primitive analogizers using similarity functions to classify data.
Master Algorithm Needs Deep Unification
- AI's foundational paradigms cycle dominance roughly per decade but have merged recently.
- The master algorithm will be a deep unification of all paradigms, not a simple toolkit combining them.
