The United States
Joined August 2017
- Tweets10.7K
- Following894
- Followers2K
- Likes22.3K
Pinned Tweet
Over a week later, company account is still wrongfully suspended after our Starlink connection post. @premium @Support
Company account is still suspended... X team have not come back to us at all. Would really appreciate some help! @X not sure if my posts are making it to any X employees though
i think we get about five years of AI progress in 2027. 50% odds it lands somewhere between three and fifteen.
the reason is in the video: according to AI 2027, and METR, about half of AI research is already done by AI. by the end of 2027 we could expect expect 95%.
each model does more of the work on the next one, so the gap between them keeps shrinking. i'd also put ~40% on AI that beats humans at every cognitive task by the end of that next year.
a lot to fit into twelve months.
1
6
341
teapawt.org/neuralithic/status/212…
eoy 2027.
Sam McCandlish said five days ago that Anthropic would be releasing a report on the speed of internal progress, so we will finally get an update to this chart. I’m very interested to see what this number is now. METR thinks it is at 50%. AL4 is the last step before full RSI.
2
4
508
will is right, it’s only a matter of when. a rising tide lifts all boats, but some would rather stay ashore
dude i hate to break it to you but it’s totally fucking over for the human race and any & all nonartisanal (derives value from being done/created by a human) economic labor. we will soon all watch from the sidelines. it’s magnus carlsen vs stockfish 13 from here on out
2
158
TLDR: bitter lesson coming for art too.
we have reached beyond “copying” and into a new era of generalisation.
seeing some interesting Fable 5.2 robotics demos too, but still far too slow for real time
I asked Fable 5.2 to create a painting as indistinguishable as possible from a photograph. this is roughly 700,000 simulated brush strokes
1
5
690
Neuralithic encoded in accelerando. pretty cool
The Latin alphabet is not how machines think in latent space nor ideally how they communicate
Created a custom 8,192-dimensional token alphabet for machines to communicate in token space directly
Inspired by the alphabet in @Grimezsz new Accelerando MV
1st img = encoded lyrics
1
198
OpenAI just released 722 mathematical manuscripts covering 372 related result families, produced by an unreleased internal model.
It was tested on roughly 4,000 open research problems after its existing math evaluations saturated.
We’re releasing a broad range of new mathematical results produced by an internal frontier model.
We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results.
github.com/openai/math
The biggest claims in OpenAI’s math release:
A proof of the quasi-Riemann hypothesis, major cases of Hodge and Birch–Swinnerton-Dyer, and a negative answer to Hilbert’s tenth problem over rational numbers.
Other claimed conclusions:
Catalan’s constant is irrational.
Both Mahler conjectures hold in every dimension.
All interpolated free group factors are isomorphic.
It also claims counterexamples to Hadwiger’s and Kaplansky’s zero-divisor conjectures.
3
244
Here it is people.
Math post is up, github live
We’re releasing a broad range of new mathematical results produced by an internal frontier model.
We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results.
github.com/openai/math
1
330
