
Chief science writer @London_Inst. 'The Man from the Future', on the unparalleled influence of John von Neumann, available everywhere. For my substack see link.
London
Joined February 2009
- Tweets28.6K
- Following6.7K
- Followers17.1K
- Likes19.6K
Pinned Tweet
Our @Nature comment this week on the use of AI in maths and theoretical physics - and why the community should embrace it!
Authors @London_Inst & @GoogleDeepMind.
First draft 8 months ago but edited many times as the field steamed ahead!
Free-to-read link at the end of 🧵1/
Despite all the talk about approaching the singularity, almost no-one appears to know who came up with the idea. Ray Kurzweil? Nope. Vernor Vinge? Wrong again. It was Johnny.
"Von Neumann was keenly aware of the various ways in which his work might ultimately contribute to humanity’s undoing. In coining the term ‘singularity’, in conversation with Ulam, von Neumann imagined a point ‘in the history of the race beyond which human affairs, as we know them, could not continue’.
Whether that would be in a negative or positive sense remains a matter of debate: thinkers have variously speculated that an artificial superintelligence might end up fulfilling all human desires, or cosseting us like pets, or eradi- cating us altogether."
--The Man from the Future (2021).
Von Neumann talked about "the singularity" in its modern sense: progress so fast that human understanding can't keep up. It was reported by his best friend, Stan Ulam. @Ananyo on where the idea comes from:
Full Episode: youtu.be/cP9B_SILZ-E
Ananyo Bhattacharya retweeted
I have a major result to share.
I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China.
This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model.
Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves.
I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems.
I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used.
It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20.
These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs.
I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at unemploymentlabs.com/checkpo….
There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.
This was an awesome conversation...
In the early 1950s, von Neumann came home from Los Alamos, skipped dinner, slept for hours, and woke in the night worried. Not mainly about the bomb. About computers. New Second Order with @Ananyo , author of The Man from the Future
Full Episode: youtu.be/cP9B_SILZ-E
1
17
903
"It might be essential to our survival, as Sagan suggests and Musk clearly believes, or we may be doomed anyway. But all the indications are that we are going to live in space, and on Mars, and that we will be doing this in the next two decades or sooner."
If you find the idea of humans living in space or walking on Mars as electrifying as I do....
'How To Build A Space Station' by @JMorrison_Times has all your favourite megalomaniac tech plutocrats, Nazi rocket scientists, spectacular visions of the (near) future and wry witticisms galore.
In short, the most inspiring book on architecture you're ever likely to read!
How To Build A Space Station
greyhoundliterary.co.uk/book…
Works particularly well when read in parallel with 'Red Mars' by Kim Stanley Robinson--I loved the audiobook narrated by Richard Ferrone.
audible.co.uk/pd/Red-Mars-Au…
1
182
My copy's arrived! All the way from the US of A! Extraordinarily timely! The story of how maths came to run the world--and created the AI systems that might surpass us all. And, ouf course, written before mathpocalypse.
"At a time when math moves faster than our own understanding, when our creations surprise and sometimes terrify us, the answer cannot be to turn away. We’ve got to keep striving — with humility, care, and every bit of wisdom we can muster.
Striving, in this new age, means remembering that every mathematical invention begins and ends with human choices. We might never fully understand the oracles we’ve made, but we can decide what to ask of them, and why. We can remain the conscience behind the code.
The mathematician in each of us will always want to know more; the human being must insist that knowledge serve something larger than itself. That might be the work of our century: letting our power grow without losing our soul."
--@stevenstrogatz & Alex Townsend.
"‘Mathematics is more than a mechanical process’ — why I think the field will survive OpenAI’s disruption" by Ursula Martin in @Nature nature.com/articles/d41586-0…
4
157
We have @KenOno691. Live. At @London_Inst!
Registration deets below.
We're delighted to be hosting @KenOno691 on Friday 27 Nov! In his AI for Mathematical Sciences (AIMS) seminar, Prof. Ono explores how AI changes mathematical discovery and proof, and what these developments mean for rigor, creativity, and human insight.
Also looking forward to our Q&A afterwards! That’s invitation only but both events will be on the @London_Inst YouTube channel so go subscribe! m.youtube.com/@London_Instit…
4
204
Ananyo Bhattacharya retweeted
In @NautilusMag , our writer @TWHodgkinson hails the mysterious-sounding X Club, which launched a scientific revolution from Albemarle St.
But old news apparently... here's Witten and Claude in July! arxiv.org/pdf/2606.18639
1
571
Not so long ago, just proving one of these cases might have secured someone a nice paper and maybe tenure I'm told.
In a new paper, our Fellow Prof. Yang-Hui He, with Dr Ziran Liu and Prof. Shing-Tung Yau, construct a strictly positive sectional curvature metric for all 28 smooth homotopy seven-spheres using a gluing theorem. The work was assisted by ChatGPT 6 Astra and Claude Pro.
1
2
717
Ananyo Bhattacharya retweeted
Most think vibecoding at scale is slop. Here’s a counterexample.
Today we release a 16M-line formalization of the entire Stacks Project, critical infra underlying modern math, with optimized code quality and its highly interconnected structure preserved.
A mathematical city 🌆
Ananyo Bhattacharya retweeted
I'd encourage anyone interested in the frontiers of mathematical discovery to come to LIMS to see how risk-taking is engineered into their daily existence; the London Institute is a leading centre for basic science, running from Michael Faraday's old home in Mayfair
Our governor @tomwalpo on why, while Silicon Valley wilts, he’s backing the London Institute to make Faraday-grade breakthroughs.
Solid advice.
Replying to @JosephPConlon
Right now the old rules are broken and no one knows the new ones.
This is precisely the best time to do a PhD — the chance to overtake all the old folk and make big lasting contributions.
It is messy right now — aim straight at the mess. 6/9























