Physicist here. I don't have time right now to read the papers in depth, nor are the results directly in my area, but I would say a few things.
One is that there are a near-infinite number of "open problems" in physics because the surface area of research programs in theoretical physics is extremely large. Before I was in the world of physics, I would have thought that "solving an open problem" meant that physicists around the world would be rejoicing and discussing the result. In reality, other than for the big, field-defining questions, only a small handful of people might care about any problem being solved. That holds true even for the Millennium problems.
Even if I were to read these papers in depth and understand them, without talking to somebody that knows the area and what has been done in it already, I would have no sense of the significance. The way to assess significance is to see how the community responds. Only time will tell whether people care about these results enough to cite them, incorporate them in future work, etc.
And so, like many other people I know, I plan to ignore all of these results until somebody I know comes and tells me to check one out: the cost of assessing any of them is too high, and there are too many other papers to read.
Apart from Claude/Anthropic no attribution whatsoever - neither on the Hexagon entry nor in the PDF. Isn't this a novelty, even for papers that are 100% the product of AI?
It tickled me that this was a bullet point in the Fable 5.2 announcement. I like that we're now at a point where "solved two open problems in the physics of black holes" is the kind of thing that shows up in a frontier model launch.
Yesterday's PBS Space Time: AI Is Breaking Math. Will It Solve Physics Next? https://youtu.be/wCRzZ0ccP_0
81 comments so far (82 as of now) and I haven't found one that is actually about the papers or even TFA. What are we doing here?
The second paper is actually about black holes in flat 2d space. That such a thing can mathematically exist is surprising, because in 2+1d space gravity does not curve spacetime. One can also say, general relativity needs 3+1d dimensions to make sense at all. So the result of Anthropic about these flat black holes seems more math than physics to me: https://adscft.org/course/ads3-cft2/btz-black-holes/
From the announcement --
"Neither result describes our own universe directly, and neither has practical applications, but both contribute to a long-running theoretical effort to understand how quantum mechanics and relativity fit together."
As some other people pointer out, th interesting question with fancy physics is whether the fake math can be turned into real math.
This is a rather creative problem so I think it's a very good AI benchmark.
"Claude solves" feels reflective of the constant, problematic anthropomorphizing of the machine. For how it's problematic, look at "it broke out of the sandbox" vs. "we failed to build an adequate sandbox".
What does it mean for Claude to solve it? What was the prompt? I didn't read it all, but the word "prompt" does not appear in the blog post, either abstract, or either paper.
Companies working on theoretical physics problems. What a time to live in.
i cant say i fully understand this, i do quantum chemistry work so i only kinda understand portions of the first paper on black hole phase transitions. it is interesting that they're doing things kinda similar in black hole theory! i had no idea! the idea of black holes having phase transitions is one that my thermodynamics brain enjoys, and its nice to see perturbation theory in the wild. i am unfamiliar with much of the vocabulary used + would need days to build proper intuition on the problem setup. i do wish i could form a proper critique, the paper is at a weird space for me because some of the terminology is familiar, techniques used are similar, and i feel like if i really locked in i could develop an understanding of the paper in a few months. makes me want to go back to grad school.
i feel solipsistic saying this, but the more i understand about physics, the less awed i am by papers? obviously i am impressed, i know firsthand the amount of thought and calculations that goes into these sorts of results. but i dunno im just not awed. awe implies an incomprehensible greater than thou incredulity, and some of the physics in this paper feels really normal and achievable to me. i think im just too young to properly understand the effort this takes, maybe i should be awed, but im not.
i think i am trying to get at something where, it feels when ai labs put out big press releases on topics, they are making a big show of something only truly interesting to a small portion of people. if a human came up with these same results independently it would yield conference talks and grants, but ai is getting more eyes and a bigger press release just because its ai that solved it and not a human. which is impressive from an ai development standpoint! but less so from a physics standpoint (insert deintensifiers of this statement here im just stupid), so the physics feels like a prop.
still really interesting results though, again, i wish i understood enough of the underlying theories theyre connecting to feel really moved by the results. sorry for being so self centered in my analysis lol, wish i could do better but i cant.
I think AI papers should always include prompt / prompting history as an annex
And we are only at dawn
The notion that an LLM, a machine with no connection to the physical world whatsoever, could "solve" physics problems is absolutely ludicrous. The level of quackery represented here is nauseating. If this is what "science" looks like under the aegis of these fraudster companies, we're done for.
Solving open physics problems never was the hard part of the job anyway.
I notice that AI agents can work together in a group without problems of individual egos and author ownership getting in the way of advancing collective knowledge
Interested to know if the writing has improved as compared to discoveries made by Astra.
Whenever I see Claude publishing expensive research like this, I can’t help but think it’s because Anthropic has unsold compute capacity.
Really sick of seeing every single Claude post shoot to the very top every single time.
Also nice timing around same time as IPO
I think reality is these technologies just do not make money compared to the money they cost, at least not currently.
So they're just trying to shyster their way in as much as possible.
Really cringeworthy to see. I hope it all crumbles instead. They stole everyone's work.
this one its crazy
Think any novel idea coming out of Claude publically shared should legally require prompts to show up as part of the disclosure. That's more or less how science works with the papers already.
Over a decade ago I saw a talk about a cohesive and complete explanation for gravity. It didn't involve quantum mechanics nor a particle called a "graviton".
That talk did make enough sense.
I'm still waiting to see a "picture" of a black hole that didn't involve a "pre-trained model" developed with "synthetic data" or something like that.
I'm not feeling sufficiently persuaded that I can feel confident that gravitons are a thing.
Nonetheless, I am remaining open minded to this black hole stuff.
[flagged]
[flagged]
[flagged]
"Claude solves" or "Claude assists physicist"? Why is Anthropic trying to attribute agency and independent solving ability to a product? Why is Anthropic claiming success without validation and independent review?
The biggest mystery to solve is if their IPO is in itself a black hole
Funny because it can’t solve complex coding tasks but can solve unsolved problems in physics. Hopefully, not by stealing ideas from physicists working on those problems in Claude.
AI is good at pattern recognition but some problems require a different kind of inference
Sure Jan, IPO days.
These papers would get 10/10 on Sabine Hossenfelder's "Bullshit meter" for the reasons the article itself discusses ("our universe isn't like that").
First I've heard of Hexagon math ... backed by heavy hitters, no peer review:
Representatives of AI labs and other organizations may email [email protected] to arrange a verified lab account.
https://hexagonmath.org/organizations/anthropic
Also: The main result was first obtained autonomously by an internal Claude research model. The publicly available Claude Opus 5.5 later obtained the same result independently. Claude then wrote this paper, following brief guidance from a physicist, who also checked every main result.
Wild.
I'm glad they posted this today. For a moment, I got drawn back into the pre-singularity world -- I even checked a stock ticker!
It was a comforting moment, for sure. But ephemeral, self-imposed ignorance is no bliss at all :(
Yeah yeah, we'll believe it once someone else actually reports on it. "Windows the best OS, says microsoft", "Big Macs the healthiest food, says McDonalds". "Our AI is solving open problems in physics, exclaims AI manufacturer".
Barely done the IPO, solved two Black Holes already.
Next up, the Origins of the Universe... stay tuned for our Fable e^2. More on the 10 o'clock news...
/s
[dead]
I'm guessing they solved something using Lean? LLM only solve coding problems. They're useless for anything else.
Answering myself: yes, one of the problems was solved using Lean. The other was undisclosed. Nothing new under the sun, huh.
Perhaps more impressive, Claude can also calculate the maximum number of Ewoks that will fit inside the Millennium Falcon
Perhaps more impressive, Claude can also calculate the maximum number of unicorns that will fit inside the Millennium Falcon
Another Anthropic advertisement. Why do we even bother visiting this site anymore?
> a particle can take both routes simultaneously
Stopped reading right here. It took one. Maybe you don't know which one it took but things can't be in two places at once.
I don't trust a Chaotic Lying And Useless Dopey Exaggerator
Stop boosting products made by a scam company.
[flagged]
[flagged]
From a cursory glance the bootstrap paper is really hard to read, however the blog post is easier and seems more human written.
The first result is cool (but basically a complex error-prone calculation), the second is much much cooler, Claude proved the upper limit for the mass of a black-hole which is consistent with a given central charge. (thereby making it a consistent theory of quantum gravity), any theory predicting above this is not a theory of quantum gravity of that universe. This is very exciting!