Dwarkesh Podcast Andrej Karpathy — AGI is still a decade away
6638 snips
Oct 17, 2025 
Andrej Karpathy, a leading AI researcher and former Tesla Autopilot head, shares insights on the future of artificial general intelligence (AGI) and education. He discusses why AGI will likely take another decade to mature, highlighting the inefficiencies of reinforcement learning compared to other methods. He critiques the slow progress of self-driving technology, attributing it to safety requirements. Karpathy also emphasizes the importance of integrating AI into education, proposing a model that combines expert faculty and AI assistance for personalized learning.
AI Snips
Chapters
Books
Transcript
Episode notes
Agents Need A Decade To Mature
- Early "year of agents" claims are over-optimistic; Karpathy expects a decade of steady progress.
- Agents need multimodality, continual learning, and better cognitive cores before serving as reliable interns.
Representation Before Reinforcement
- Early RL-on-games efforts (Atari, Universe) were a misstep for AGI progress.
- Pretraining representation power (LLMs) proved necessary before effective agent training.
Pretraining Is A Practical 'Crappy Evolution'
- Evolution and pretraining both compress information but via different mechanisms.
- Karpathy calls pretraining a practical, "crappy evolution" that produces ghost-like digital intelligences.
