A Man Behind AlphaGo Thinks AI Is Taking the Wrong Path
8 hours ago
- #Superintelligence
- #Artificial Intelligence
- #Reinforcement Learning
- David Silver co-developed AlphaGo, which achieved superhuman proficiency in Go through self-learning.
- He founded Ineffable Intelligence, focusing on reinforcement learning to create 'superlearners' beyond human intelligence.
- Silver criticizes the LLM-based approach to AI superintelligence as limited by human data and advocates for self-learning systems.
- Ineffable Intelligence raised $1.1 billion in seed funding, with Silver pledging to donate all personal equity to charity.
- Silver plans to use simulations to scale AI learning from confined environments like Go to real-world complexity.
- He emphasizes safety and alignment, suggesting simulations allow observation of AI behavior to ensure benevolence.
- Reinforcement learning, central to Silver's approach, has historical roots and recent successes in AI systems like chatbots.
- Silver's reputation as a top researcher and humble collaborator aids in recruiting talent for his mission.