scripod.com
Thoughts on AI progress (Dec 2025)

Overview

Shownote

Highlights

Transcript

Chapters

Pins

Thoughts on AI progress (Dec 2025)

Dwarkesh Podcast

2025/12/23
Dwarkesh Podcast

Dwarkesh Podcast

2025/12/23
OverviewShownoteHighlightsTranscriptChaptersPins

Shownote

Read the essay here. Timestamps 00:00:00 What are we scaling? 00:03:11 The value of human labor 00:05:04 Economic diffusion lag is cope00:06:34 Goal-post shifting is justified 00:08:23 RL scaling 00:09:18 Broadly deployed intelligence explosion Ge...

Highlights

This podcast examines the current trajectory of AI development, focusing on the challenges and misconceptions surrounding scalability, economic impact, and the path to artificial general intelligence. It critiques common narratives about AI adoption and progress, particularly in reinforcement learning and real-world deployment.
00:00
If models were human-like learners, current RL approaches would be pointless
05:06
If AI had human-like capabilities, it would diffuse quickly because it’s easier to integrate than human employees.
06:36
The previous definition of AGI may have been too narrow
08:25
A million-fold compute scale-up may be needed for GPT-level RL gains

Chapters

What are we scaling?
00:00
The value of human labor
03:11
Economic diffusion lag is cope
05:04
Goal-post shifting is justified
06:34
RL scaling
08:23
Broadly deployed intelligence explosion
09:18

Transcript

Dwarkesh Patel: I'm confused why some people have super short timelines, yet, at the same time, are bullish on scaling up reinforcement learning atop LLMs. If we're actually close to a human-like learner, then this whole approach of training on verifiable ...