Some random things that may be of interest:
- Ethan Siegel has a discussion of “vibe physics“, people convincing themselves that they can solve fundamental scientific problems by chatting with an LLM. For a story about one billionaire doing this, see here.
LLMs should be much better than the usual crackpots at generating worthless papers about theoretical physics, likely should be able to generate papers not easy to distinguish from a lot of what is on the arXiv. I’m wondering how much of this has already happened.
In math, Daniel Litt has noticed a bunch of recent LLM-generated worthless papers on the Hodge conjecture. As examples, he points to these, four papers posted during the past month. Unfortunately the arXiv does not seem to now have an effective way to protect itself against these things getting posted, or to get them removed once identified (Daniel identified them publicly two weeks ago, no indication anything will be done about this).
- Also on the arXiv is an article by George Lusztig which goes over some history, with this summary
By publishing this document I aim to rectify the historical narrative for the benefit of the mathematical community and of the general public and to ensure that proper attribution and academic integrity is upheld by all.
I trust that all readers -including Kashiwara- will recognize these established facts:(a) The canonical basis was first defined in my work [L90] and Kashiwara’s subsequent contribution built directly on this foundation.
(b) The crystal basis is not solely Kashiwara’s discovery.And everyone who knows the history would suggest Kashiwara to publicly acknowledge (a) and (b), to correct all false and misleading information once and for all.
- Another one has been added to the list of Leinweber Institutes for Theoretical Physics, discussed here. It’s the new Leinweber Institute for Theoretical Physics at Stanford, previously called the Stanford Institute for Theoretical Physics.
- UCSB has announced that they’ve digitized Joe Polchinski’s papers.
The link they give doesn’t appear to work and I don’t know of any other way to access this archive. The link now works. - The Chinese each year are now organizing a conference that covers mathematics and theoretical physics on a truly massive scale, called the International Congress of Basic Science. You can keep busy by watching 390 talks on Youtube.
- At Strings 2025 earlier this year there was not yet a plan for a Strings 2026. The Chinese have also taken this on, Strings 2026 will be in Shanghai.
- For a podcast worth watching, see Curt Jaimungal’s interview with Nikita Nekrasov.
- For another one, there’s Sean Carroll talking to David Tong. I especially recommend the part around 52 minutes in, where Tong advertises a crucial hole in our understanding of the Standard Model: the non-perturbative formulation of the chiral gauge theory of the electroweak sector, in particular the lack of a viable lattice formulation.
- Last month there was the Open Symposium on the European Strategy for Particle Physics in Venice. Crucial numbers are in this report: 8-9 billion to build a linear collider, a big new ring (FCC) would be 15 billion for an initial lepton machine, another 19 for a higher energy proton machine (these are rough numbers, think of as dollars, euros or swiss francs). The FCC project has been the leading proposal, but the crucial question is whether such a thing is financially viable.
Update: Daniel Litt has also written about this on his blog. There’s a comment there from “knzhou” saying
“This is also happening in hep-ph, which now has an average of 1-2 nonsensical papers per day.”
Update: Terry Tao’s NSF grant at UCLA has been suspended (along with 279 others), because UCLA is “antisemitic” since there were anti-genocide protests there last year. Unclear to me in this case which pro-genocide forces are collaborating with the dictator to shutdown Terry Tao, and why they are doing it.
UCSB appears to intend “digitized” to mean “created an online index” rather than the ordinary understanding “made available online”.
X,
Yes, but is there a working link to that index?
Looks like the link has now been fixed.
For the colliders, rather than asking whether they are financially viable, I think physicists should seriously consider what the chances are of not discovering anything new. Imagine the fallout from newspaper headlines “Particle theorists spend $30 billion on new supercollider, discover nothing.”
Recall that the LHC was guaranteed to discover something new (either the Higgs or the lack of a Higgs, which would have been much more exciting). Will this be true of the next collider?
Peter Shor,
No guarantees this time. There’s a lot to be said about the scientific case for these colliders and that’s what most of the talks were about. This is part of a process leading to a report next January. That will give the best judgment of the experimental HEP community about which if any of these to try and build. That will be the time for a public discussion about whether the best plan physicists can come up with is worth the money it would take to implement. I don’t think there’s much point in arguing now against a plan that doesn’t yet exist.
I was giving those numbers just because they’re crucial, but rarely emphasized, buried deep in the materials for this conference. From my understanding of CERN finances and prospects for getting contributions from others, I don’t see how they can afford the 15 billion for a first stage (lepton) FCC ring. But, that’s been the leading proposal so far.
To argue about whether to build a collider, the question of cost is crucial. If you could build the FCCee for 1 billion, something that would easily fit into the current CERN budget (about 1.5 billion/year), it would be worth doing, since you would at least be testing the Standard Model in new ways. Whether it would be worth 15 billion is a different question, but one that might be moot if there’s no way to get that sum.
We’ll see what emerges in January.
Are you aware of the recent feuds involving popular Youtuber “Professor Dave” (who is a regular debunker of pseudoscience), and critics of the HEP orthodoxy – first Eric Weinstein, and now Sabine Hossenfelder as well? He’s portraying them as agents of an oligarchic master plan by Peter Thiel, intended to sow doubt regarding mainstream institutions of knowledge.
Mitchell Porter,
I’ve briefly looked at this stuff. “Professor Dave” is a Youtube influencer with an undergrad degree in chemistry and about 4 million followers. That anyone would pay the slightest attention to what he has to say about something he knows nothing about (the state of research in fundamental physics) is just one more indication of how bad the problem of epistemic collapse has become…
These 11 papers on the Math arXiv from May/June are also all AI-generated nonsense: https://arxiv.org/search/math?query=Zheng%2C+Dongzhe&searchtype=author
AI pollution of scientific literature will certainly become a serious problem.
I came across two papers that claim to have solved the millennium problem concerning Yang-Mills, which very strongly suggest AI authorship:
— D.C. Jacobsen: arXiv:2506.00284
— Logan Nye: https://doi.org/10.3390/ijt2010002
The second one has actually been published in a journal unknown to me (International Journal of Topology) .
I think in the future the arXiv and serious journals should at least require a statement of the authors about their use of AI, just like there are statements about conflicts of interest often required.
Given that BS LLM generated papers are likely, if nothing is done, to become both more numerous and harder to identify, it would probably be a good idea for relevant fields to create some kind of code of conduct and norms about what constitutes responsible usage of these technologies to help with paper creation, including full disclosure of exactly how and for what AIs were used… with non or false disclosure to be treated like plagiarism or p-hacking or other species of academic malfeasance.
To add more to my comment about AI-generated hep-ph papers: I primarily think about new kinds of precision experiments to search for BSM physics, particularly involving ultralight bosonic fields. It doesn’t require very complicated QFT, but it’s nontrivial for two reasons. First, one has to start from a Lagrangian and understand what it implies for a macroscopic detector full of interacting electrons, nuclei, and photons; calculating some free particle S-matrix element is rarely enough, and often leads to completely wrong estimates of signal strength, e.g. by dropping shielding, screening, binding, decoherence, reciprocal process, finite-size, or finite-time effects. Second, new experiments often require pushing equipment to work in new regimes, where novel noise sources can emerge. One really has to understand the physics of each detector to anticipate and estimate these noise sources.
For these reasons, it has always been easy to write a paper that vastly overestimates the strength of a proposed experimental approach. Moreover, peer review at top journals like PRL has always been unreliable, because the subject is interdisciplinary. (The referee may be an experimentalist who can’t catch that the theoretical calculation of the signal is wrong, or a theorist who doesn’t know about noise sources that are being ignored.) AI-assisted papers make this problem worse since they let many people mass-produce papers that look reasonable at first glance, but are missing the more careful thought that would catch issues with the signal or noise. I don’t envy the editors of PRL.
In the short term, this will just speed up the siloing of the field, where established groups avoid reading anything that wasn’t produced by a small number of peers. The long term threat is that governments could spend significant money on experiments that can’t possibly work. This is “fortunately” not an issue right now in the US, since even the experiments we _know_ can work are being defunded, but that’s another issue…
Has also happened in astrophysics. I’ve notified arXiv of this paper:
https://arxiv.org/abs/2504.19807
In addition to LLM-generated text, it includes at least one LLM-hallucinated reference.
The paper has now been withdrawn “due to serious violation of arXiv policies for acceptable submissions”. Of the three original authors, only one is listed in the withdrawal, so the issue may have been AI-generation not only of content but also of authors.
Angela Collier, whose “String Theorists Lied to Us” video you recommended a while ago, has a pretty nice video on vibe physics if you’re at all interested!
Terence Tao’s response: https://mathstodon.xyz/@tao/114956840959338146
abc,
Thanks. Tao does an excellent job of explaining the implications for mathematics research of this funding cutoff. I do think though he makes a mistake by not dealing with the underlying reason this is happening at UCLA and elsewhere (the campaign to stop criticism of Israeli genocide by tarring it as antisemitism and the Columbia capitulation encouraging wider use of the tactic of illegal taking of research funds).
This is not something happening for unknown reasons or done by unknown people.
Peter –
“That anyone would pay the slightest attention to what he has to say about something he knows nothing about (the state of research in fundamental physics) is just one more indication of how bad the problem of epistemic collapse has become”
He generally has physicist guests on to provide the technical details behind his claims. In fact, he has right now released another anti-Sabine video, in which Michael Peskin and Nick Warner (and four other physicists) defend the state of the field…
Mitchell Porter,
There’s plenty to reasonably critique about Sabine Hossenfelder’s arguments and how she is putting them forward. But she’s a serious person with a serious background in the field making serious arguments (sometimes lately expressed in an over the top manner…). Yes, “Professor Dave” has gotten some other physicists to appear on his show arguing against her. But he’s not trying to host a serious discussion, it’s just an intellectually dishonest propaganda campaign worse than anything she has ever done.
The description of the video starts off
“At this point, everyone who watches my content is well aware that Sabine is a disgusting fraud peddling propaganda for fascist oligarchs.”
Whatever “Professor Dave” tells you, the state of parts of theoretical physics is not healthy and has not been healthy for a while. Just about any serious person in the field understands this, it’s not “propaganda” from “fascist oligarchs”. It’s also a complicated subject that you will learn nothing about by watching “Professor Dave” youtube videos. I wasted too much of my life trying to have serious arguments over these issues with people, but at the worst of it, none of that was as idiotic as this kind of worthless video.
I’ll stick by my claim that “Professor Dave” is a poster boy for epistemic collapse.