Dwarkesh Patel
@dwarkesh_sp
Episode dropping tomorrow with @paulfchristiano.
He thinks there’s a 1 in 3 chance of extreme superintelligence within a year (absent any slowing down).
And he'll explain what labs and governments should do right now given that possibility.
The avalanche that hit math is coming for AI development.
This got totally buried cause it came out the same day as Navier Stokes.
But I think this is super interesting evidence about what the true major source of algorithmic progress is.
Dwarkesh PatelX
View on XPretraining progress seems to be coming mostly from data improvements.
@who_is_jerbear and I pretrained combinations of year-representative open model recipes and data corpuses across 2019 to 2025 at various small scales.
Data improvements contributed 3.24x as many compute multipliers as model improvements did (12.0x vs 3.7x).
And the gains stack independently - a better dataset helps every architecture about equally, and vice versa.
Here are full results, plus what we think this means for the future of AI progress:
https://www.dwarkesh.com/p/pretraining-progress-is-mostly-data
At Blood on the Clocktower, Fable 5.2 is still worse than GPT-6 Astra.
Though much better than Opus 5.5 or Fable 5.1
Dwarkesh PatelX
I spent 20,000 dollars in API credits having different AI models play each other at Blood on the Clocktower (a Mafia or Werewolf type social deception game).
The results were fascinating. Models have gotten much better over the course of this year at creative on-the-spot scheming and bluffing, and also at reasoning as a team to break apart these underground conspiracies.
I go through some of the most standout games these AIs played in this video. The new models are getting pretty good. If my friends pulled off some of these moves while we were playing, I’d give them a fist bump.
Fwiw, OpenAI’s newer models are much better than Anthropic’s at this game.
You can see the full benchmark at the website, and I’ve also made it so that you and your friends can play again the AIs: http://clocktowerbench.com/
View on X
I spent 20,000 dollars in API credits having different AI models play each other at Blood on the Clocktower (a Mafia or Werewolf type social deception game).
The best game I watched the AIs play:
Dwarkesh PatelX
I spent 20,000 dollars in API credits having different AI models play each other at Blood on the Clocktower (a Mafia or Werewolf type social deception game).
The results were fascinating. Models have gotten much better over the course of this year at creative on-the-spot scheming and bluffing, and also at reasoning as a team to break apart these underground conspiracies.
I go through some of the most standout games these AIs played in this video. The new models are getting pretty good. If my friends pulled off some of these moves while we were playing, I’d give them a fist bump.
Fwiw, OpenAI’s newer models are much better than Anthropic’s at this game.
You can see the full benchmark at the website, and I’ve also made it so that you and your friends can play again the AIs: http://clocktowerbench.com/
View on X
Robert Caro on loneliness:
I recently got to interview Robert Caro, who is my favorite author of all time.
This gave me the excuse to reread his biographies of Lyndon Johnson and Robert Moses, and to compile my prep notes into this video essay.
These guys pulled off some truly incredible schemes.
Finally happened.
Dropping tomorrow.
Dwarkesh PatelX
View on XWithin the next few months, I'm going to be releasing an episode with by far my biggest guest yet
If you're interesting in sponsoring it, shoot me an email at sponsor@dwarkeshpatel.com
My semis tutor @willmichaels0 and I made a bunch of flashcards about semiconductor manufacturing.
How a modern transistor works:
Pretraining progress seems to be coming mostly from data improvements.
Within the next few months, I'm going to be releasing an episode with by far my biggest guest yet
If you're interesting in sponsoring it, shoot me an email at sponsor@dwarkeshpatel.com