[Linkpost] “The Extreme Inefficiency of RL for Frontier Models” by Toby_Ord
EA Forum Podcast (Curated & popular)2 Feb

[Linkpost] “The Extreme Inefficiency of RL for Frontier Models” by Toby_Ord

This is a link post.

The new scaling paradigm for AI reduces the amount of information a model can learn from per hour of training by a factor of 1,000 to 1,000,000. I explore what this means and its implications for scaling.

The last year has seen a massive shift in how leading AI models are trained. 2018–2023 was the era of pre-training scaling. LLMs were primarily trained by next-token prediction (also known as pre-training). Much of OpenAI's progress from GPT-1 to GPT-4, came from scaling up the amount of pre-training by a factor of 1,000,000. New capabilities were unlocked not through scientific breakthroughs, but through doing more-or-less the same thing at ever-larger scales. Everyone was talking about the success of scaling, from AI labs to venture capitalists to policy makers.

However, there's been markedly little progress in scaling up this kind of training since (GPT-4.5 added one more factor of 10, but was then quietly retired). Instead, there has been a shift to taking one of these pre-trained models and further training it with large amounts of Reinforcement Learning (RL). This has produced models like OpenAI's o1, o3, and GPT-5, with dramatic improvements in reasoning (such as solving [...]

---

First published:
February 2nd, 2026

Source:
https://forum.effectivealtruism.org/posts/64iwgmMvGSTBHPdHg/the-extreme-inefficiency-of-rl-for-frontier-models

Linkpost URL:
https://www.tobyord.com/writing/inefficiency-of-reinforcement-learning

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

Episoder(250)

[Linkpost] “Starfish” by Aaron Gertler 🔸

[Linkpost] “Starfish” by Aaron Gertler 🔸

This is a link post. By Alexander Wales Thousands of starfish had washed up on the beach, and a little girl was diligently throwing them back into the water, one at a time. A man came up to the girl a...

1 Mai 4min

“Time Sensitive Urgent Animal Welfare Action” by Bentham’s Bulldog

“Time Sensitive Urgent Animal Welfare Action” by Bentham’s Bulldog

The EATS act—now called the save our bacon act—would make it illegal for states to pass animal welfare laws that apply to products produced out of state. This would gut most state level animal protect...

29 Apr 1min

“Forecasting is Way Overrated, and We Should Stop Funding It” by Marcus Abramovitch 🔸

“Forecasting is Way Overrated, and We Should Stop Funding It” by Marcus Abramovitch 🔸

Summary EA and rationalists got enamoured with forecasting and prediction markets and made them part of the culture, but this hasn’t proven very useful, yet it continues to receive substantial EA fun...

26 Apr 8min

“My lover, effective altruism” by Natalie_Cargill

“My lover, effective altruism” by Natalie_Cargill

Crossposted from Substack. This post is part of a 30-posts-in-30-days ordeal at Inkhaven. All suboptimalities are the result of that. This is part 2, here is part 1 in my EA mini series! On my way to ...

26 Apr 8min

“A Database of Near-Term Interventions for Wild Animals” by Bob Fischer

“A Database of Near-Term Interventions for Wild Animals” by Bob Fischer

The Animal Welfare Department (AWD) at Rethink Priorities supports high-impact strategies to help animals, especially where suffering is vast and largely neglected. Therefore, one of our focus areas i...

24 Apr 17min

“The AI people have been right a lot” by Dylan Matthews

“The AI people have been right a lot” by Dylan Matthews

This post was crossposted from Dylan Matthew's blog by the EA Forum team. The author may not see or reply to comments. Subtitle: Try to keep an open mind as the world gets increasingly wild.The crowd ...

20 Apr 10min

[Linkpost] “The Anthropic IPO Is Coming. We Aren’t Ready for It.” by Sophie Kim

[Linkpost] “The Anthropic IPO Is Coming. We Aren’t Ready for It.” by Sophie Kim

This is a link post. More money is coming than AI safety has ever seen. The capacity to deploy it doesn't exist yet.Image Source: Fortune.com This week, Anthropic announced Claude Mythos Preview– a mo...

17 Apr 15min

“AI Safety’s Biggest Talent Gap Isn’t Researchers. It’s Generalists.” by Topaz, Agustín Covarrubias 🔸, Alexandra Bates, Parv Mahajan, Kairos

“AI Safety’s Biggest Talent Gap Isn’t Researchers. It’s Generalists.” by Topaz, Agustín Covarrubias 🔸, Alexandra Bates, Parv Mahajan, Kairos

This post was cross posted to LessWrong TL;DR: One of the largest talent gaps in AI safety is competent generalists: program managers, fieldbuilders, operators, org leaders, chiefs of staff, founders....

16 Apr 13min

Populært innen Samfunn

rss-spartsklubben
giver-og-gjengen-vg
aftenpodden
alt-fortalt
aftenpodden-usa
konspirasjonspodden
popradet
rss-nesten-hele-uka-med-lepperod
rss-henlagt-andy-larsgaard
lydartikler-fra-aftenposten
intervjuet
wolfgang-wee-uncut
grenselos
rss-espen-lee-usensurert
synnve-og-vanessa
min-barneoppdragelse
rss-dette-ma-aldri-skje-igjen
rss-dannet-uten-piano
frokostshowet-pa-p5
fladseth