Model Intelligence Scaling
Dwarkesh and John discuss the scaling of model intelligence for longer horizon tasks, exploring the potential need for more computational resources and the possibility of phase transitions in achieving capabilities across various timescales. They delve into the similarities in mental mechanisms used for planning at different timescales, suggesting a nuanced approach to training models for diverse task durations.In this clip
From this podcast
Dwarkesh Podcast
John Schulman (OpenAI Cofounder) - Reasoning, RLHF, & Plan for 2027 AGI
Related Questions