Unlocking AI Capabilities
Dwarkesh and John discuss the potential of long horizon RL training to unlock AI capabilities, highlighting the need for coherence and predicting human-level outcomes. They explore possible bottlenecks and the role of human expertise in overcoming limitations for AI models.In this clip
From this podcast
Dwarkesh Podcast
John Schulman (OpenAI Cofounder) - Reasoning, RLHF, & Plan for 2027 AGI
Related Questions
Is reinforcement learning a turning point for large language models (LLMs) and artificial intelligence (AI)?
I have a question about the episode John Schulman (OpenAI Cofounder) - Reasoning, RLHF, & Plan for 2027 AGI and the clip Future Model Capabilities regarding creating content before or after training.