EIEpisodeIndex
From podcastDwarkesh PodcastFrom guestIs RL

Episode #84

Is RL + LLMs enough for AGI? — Sholto Douglas & Trenton Bricken

Listen now | Scaling reinforcement learning, tracing circuits, and the path to fully autonomous agents

May 22, 20252h 24m13 chaptersIs RL

People

Is RL
More from Dwarkesh PodcastRead transcript
What this episode covers

Listen now | Scaling reinforcement learning, tracing circuits, and the path to fully autonomous agents

Where to start

Why are LLMs ‘baby AGI’ but not AlphaZero?

Why are LLMs ‘baby AGI’ but not AlphaZero?

Start at 1:37:42

Is continual learning a key bottleneck?

Is continual learning a key bottleneck?

Start at 16:27

Inference compute will bottleneck AGI

Inference compute will bottleneck AGI

Start at 1:18:55

People and topics

People

Is RL
Key takeaways
  • How far can RL scale?
  • Is continual learning a key bottleneck?
  • Model self-awareness
  • Taste and slop
  • How soon to fully autonomous agents?
All moments
Is RL + LLMs enough for AGI? — Sholto Douglas & Trenton Bricken podcast chapters, timestamps & summary | EpisodeIndex