MIT asks arXiv to withdraw preprint of paper on AI and scientific discovery
economics.mit.edu
economics.mit.edu
Between this and the subtle reference to “former second-year PhD student” it makes sense that they’d have to make a public statement.
They do a good job of toeing the required line of privacy while also giving enough information to see what’s going on.
I wonder if the author thought they could leave the paper up and ride it into a new position while telling a story about voluntarily choosing to leave MIT. They probably didn’t expect MIT to make a public statement about the paper and turn it into a far bigger news story than it would have been if the author quietly retracted it.
Edit: this comment was only partially serious, not meant as legal advice to MIT.
Furthermore, if the author could demonstrate to arXiv that the request was fraudulent, the paper would be reinstated. The narrative would also switch to people being angry at MIT for impersonating a student to do something.
Emails are not people. You can impersonate a person, but you can't impersonate an email. If I own a company and I issue the email dick.less@privateequity.com but then have to fire him... using this email address to transfer company assets back to someone who can be responsible for them isn't fraud (for that purpose, at least). How is this not the same issue?
Although not explicitly stated, i read previous comments as using dick.less@privateequity.com to cancel his personal Netflix account. (Let's say that privateequity.com allowed personal usage of company email.)
I see a difference between accessing an email account and impersonating the previous account holder.
It's infuriating that Instagram, Facebook, etc send a "Email Verification" that has NO option to say "Nope, not me, don't want it, don't do it".
Worse, I'd like to create my own Instagram now, but cannot, because somebody else tried to use my email a decade ago and now all I get is a very very confused loop.
I'm sure these systems make sense to somebody, there's detail and nuance and practicality I'm horribly ignorant of, but they just seem insanely unprofessional to me as an outsider :-/
Why not use a different email address? Nothing about that would make it less your “own” Instagram account.
The email is attached and always has been to my personal Facebook account. But it's also now attached to not-mine instagram account. And because two are kind of independent but kind of integrated, it generates a lot of weirdness I'd like to remove by having my own properly owned and integrated and tagged Facebook and instagram accounts.
It'd take 12 seconds for a human being to figure out but of course that's not an option.
I understand it might be a bit hostile but it seems there should be some means to transfer the account. I'd try contacting them first, but if they're non responsive or the account is abandoned then do a take over. They impersonated you anyways. If necessary delete the account and recreate after the cooldown period
But also I don't want to inadvertently assume any responsibility for the past content or actions of an account thay was never mine.
> I've done it for people who used my email to sign up for Facebook and Instagram.
Maybe MIT does technically have full control and technically the legal right, but if they acted in that way there would be potentially larger consequences. Employers might start preferring other emails or contact methods which would make it harder for MIT to comply with certain regulations. While academia may have been turned into a business there are still high expectations that faculty have high freedom of expression. Every move to limit this, or even imply limiting it, is met with opposition. It jeopardizes the ability to do one's job as a researcher. So I'm not surprised if MIT would never act on "their right".
They can disable the email account but that does nothing to stop an arxiv account. Again this also has risks as academic emails are traditionally preserved. Your correspondence is on a physical piece of paper and changing institutions and affiliations is common. Almost every grad student does this, and nearly every professor was once a grad student.
There's just too many downsides to MIT making such a request. Unless the email was institutional, as in impersonating the institution ( e.g. "info@mit.edu"), or impersonating someone else then they really won't act that way and really shouldn't
1. The data in most of the plots (see the appendix) look fake. Real life data does not look that clean.
2. In May of 2022, 6 months before chatGPT put genAI in the spotlight, how does a second-year PhD student manage to convince a large materials lab firm to conduct an experiment with over 1,000 of its employees? What was the model used? It only says GANs+diffusion. Most of the technical details are just high-level general explanations of what these concepts are, nothing specific.
"Following a short pilot program, the lab began a large-scale rollout of the model in May of 2022." Anyone who has worked at a large company knows -- this just does not happen.
Could a Benford's Law analysis apply here to detect that?
https://en.wikipedia.org/wiki/When_Contact_Changes_Minds
https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&d...
> As we examined the study’s data in planning our own studies, two features surprised us: voters’ survey responses exhibit much higher test-retest reliabilities than we have observed in any other panel survey data, and the response and reinterview rates of the panel survey were significantly higher than we expected.
> The firm also denied having the capabilities to perform many aspects of the recruitment procedures described in LaCour and Green (2014).
The data quality for that would need to be unimaginably high.
% gunzip -c arXiv-2412.17866v1.tar.gz | tar xOf - main.tex | grep '\bI have\b'
To summarize, I have established three facts. First, AI substantially increases the average rate of materials discovery. Second, it disproportionately benefits researchers with high initial productivity. Third, this heterogeneity is driven almost entirely by differences in judgment. To understand the mechanisms behind these results, I investigate the dynamics of human-AI collaboration in science.
\item Compared to other methods I have used, the AI tool generates potential materials that are more likely to possess desirable properties.
\item The AI tool generates potential materials with physical structures that are more distinct than those produced by other methods I have used.
% gunzip -c arXiv-2412.17866v1.tar.gz | tar xOf - main.tex | grep '\b I \b' | wc
25 1858 12791
%Nice to include the giants we stand on as implied coauthors.
Established tradition doesn't negate "collective delusion".
And anyone who uses the use of "I" in a paper to imply anything about its authenticity is definitely indulging in some form of a delusion. It's not the norm, but is definitely permitted in most technical fields. When I was in academia no reputable journal editor would take seriously reviewer feedback that complains about the use of I.
(In this q&a, the audience does not really question the validity of the research.)
I miss google search's Cache. As with the seminar, several other hits on MIT pages have been removed. I'm reminded of a PBS News Hour story, on free fusion energy from water in your basement (yes, really), which was memory holed shortly after. The next-ish night they seemed rather put out, protesting they had verified the story... with "a scientist".
That cassyni talk link... I've seen a lot of MIT talks (a favorite mind candy), and though Sloan talks were underrepresented, that looked... more than a little odd. MIT Q&A norms are diverse, from the subtle question you won't appreciate if you haven't already spotted the fatal flaw, to bluntness leaving the speaker in tears. I wonder if there's a seminar tape.
https://sloanreview.mit.edu/audio/feed-drop-how-ai-will-chan...
There was a question at the end that made him a little uncomfortable:
[1:00:20]
Q: Did you use academic labs only or did you use private labs?
A: (uncomfortable pause) Oh private, yeah, so like all corporate, yeah...
Q: So, no academic labs?
A: I think it's a good question (scratches head uncomfortably, seemingly trying to hide), what this would look like in an academic setting, cause like, ... the goals are driven by what product we're going make ... academia is all, like "we're looking around trying to create cool stuff"...
Buddy, I've met MIT profs whose public speaking was so horrible, it would probably take them 30 minutes to order a glass of water.
See I'm thirsty, but I can drink the water later. And my grant proposal is due at midnight tonight (12:01 on Sunday, technically), and my PhD student is texting me to say that he can't log into the cluster, and we also just got the proofs back from that paper but I guess that can wait. At some point I should fill out that reimbursement form for the conference last week but first I should get back to those undergrads who said they wanted a summer ... wait water? Oh yes sure water would be great.
These seminar-style talks in particular have a strong Goodheart bias: academic scientists judge each other on the papers they write, but the highest honors usually come in the form of invited talks. The result is that everyone is scrambling to have their students give talks.
In larger scientific collaborations it can get a bit perverse: you want to get everyone together for discussions, but the institutes will only pay for travel if you give their students a 20 minute talk. You'll often have conferences where everyone crams into a room and listens to back-to-back 20 minute lectures for a week straight (sometimes split into multiple "parallel sessions"), and the only real questions are from a few senior people.
It's a net positive, of course: there's still some time around the lectures and even in 2025 there's no good replacement for face-to-face interaction. But I often wish more people could just go to conferences to discuss their work rather than "giving a talk" on it.
I don’t expect a TED talk but we’re still talking about MIT here. I’ve seen 8 year olds more articulated. I guess where I am from being called in front of the class and having to present or talk about the homework reading is common, so perhaps why it’s seen as exotic in US to be able tie words together without saying “like” after every other word, or slump and touch the hair every 10 seconds.
I wonder about this too, but I think any academic institute asks a lot of PhD students. 90% of it isn't about giving a good public talk. Especially at PhD level it's much more about actually gathering a blob of data, distilling it into a (still nonlinear) structure, and then, finally, serializing it into a paper draft. In many cases the talk is something you do at the end as a formality.
This doesn't get any simpler just because you're at an institute with a fancy name. Your hypothetical 8 year old has one chance to get a cookie and had better be pretty articulate about it. This MIT-branded academic has a million other things going for them and can afford to slack off a bit on the presentation skills.
Nah, they also can explain how potential and kinetic energy works, talk about how many types of stars are out there and so on. Not hypothetical at all. They do like cookies, too!
> This MIT-branded academic has a million other things going for them and can afford to slack off a bit on the presentation skills.
Well, I posit in this case their 1st out of 1 million other worries was to sound credible, because they may be asked about their methodology. Staying consistent while making things up does take considerable amount of effort and the speech will suffer. Listen to the segment I point out and see how they act. They sort of pretended they didn't hear the question at first.
The question is more about how good of a cheater he is. Not very good as keeping a story straight takes up a considerable amount of effort and speech suffers.
Very true point. I've been wondering why academics "suffer" so much from a system that they themselves created and are actively running (unpaid, all volunteers-based). Conferences are organized by academics for academics. Grants are subitted by and evalated by academics. Journals' editorial boards are staffed by academics to review the work of their peers. Even metrics of merit are defined by scholars of scientrometrics, also academics, to rate the works of their peers. Yet we have a system where peer review has a high element of arbitrariness, lacks minimal professional standards, conferences organizers take too much of their valuable time to do a job (rotational, even!) that is often mediocre, and authors donate their papers for free to commercial publishers from which their institutions then buy these same papers back for a lot of money. After a quick analysis of these entrenched systems in my first months of doctoral studies, I questioned the intelligence of people who first created such a system and then keep complaining about it, yet they make no move to change anything.
Let's invent a new meeting format where people basically travel to a nice place with few distractions in order to discuss their research informally, no talks.
In my field (computer science), it's what workshops once were before they became mini-conferences with three minutes question time (for all the questions from the whole audience after one talk, not per person asking) after talks.
PS [edit]: I once saw two older professors discussing something on the corridor floor of a conference while talks were going on inside the various rooms. They were sitting on the floor, both held pens and there were some loose papers scattered on the floor. This was right were people coming out of talks would have had to walk over them. I had skipped that session, so I asked them what they were doing. They said "Oh, we're writing a paper. We only meet twice a yeara some conference, that's when we need to get most of our important work done." At the time I found it funny, but with the benefit of hindsight isn't it a sad state of affairs?
There was a recent podcast that covered it with some experts that's a great listen:
The word has many, many uses: filler/pause, oral punctuation, discourse marker, hedging, qualifier. It also serves an important social function, in that it can reduce perceived severity or seriousness. Young women seem to use it assure peers that they are sweet and not threatening.
I hate it. It's not uncommon to hear it more than four or five times in a single sentence.
The implied expectations are odious: eloquence is a faux pas; directness is rude; a fifth-grade vocabulary is welcoming.
then you be all like, "but! but! that's not what i mean", and then they be like, "ok boomer" and then they be sharing tiktoks of yo red face on red dit(!) and img2img you into Wojak memes
--
Couldn't resist. I feel like you feel, and I'm saddened by the fact that there seems to be no way to hold any quality standards in modern society. Yes, standards of writing and speech and culture are arbitrary, but it's valuable to have them; meanwhile, we live in times where you can't require anything of people, because there's always someone ready to argue the expectations are unfair or malicious, and there's no culturally valid counterargument to that.
I’ve even worked in places where some ML researchers seemingly made up numbers for years on end.
A cursory review of the first paragraph of the abstract of his single author paper should've set off alarms:
"AI-assisted researchers discover 44% more materials, resulting in a 39% increase in patent filings and a 17% rise in downstream product innovation".
Anyone with rudimentary familiarity with industrial materials science research would have suspected those double digit numbers - even single digit improvements are extremely rare.
https://cassyni.com/events/MiPYGu3qzKP5MQFWNUn9Tb
In retrospect, there seems to be a tell that when he's lying he won't look at the screen/camera: his eyes go up, left, right, anywhere but forward. What I find scary is that this practice of extemporaneous fabrication may be a well-ingrained habit at this point that isn't limited to the scientific realm of the author's life.
In this case perhaps that's a tell, but there are a lot of people who don't like looking people in the eyes for other reasons, too, just so you're aware.
I struggle to make and maintain eye contact, because bad childhood trauma. Eye contact was perceived as aggression and disrespect by the man who lived in my house, and would lead to beatings. There are a lot of us that struggle with that, for many different reasons.
“It’s fantastic,” said Acemoglu. “I was floored,” said Autor.
https://www.wsj.com/economy/will-ai-help-hurt-workers-income...
Clearly, Lahart is a journalistic tool and charlatan, so he's @ the right place:
"Justin Lahart is an economics reporter based in New York. Previously, Justin
was a Heard on the Street columnist and wrote the Ahead of the Tape column."https://www.wipo.int/amc/en/domains/decisions/pdf/2025/d2025...
Detailed (meta)science-based reporting here from another person who guessed Corning: https://thebsdetector.substack.com/p/ai-materials-and-fraud-...
>It looks eerily similar to the distribution in this [pharma] preprint. This distribution might make sense for drugs, but makes very little intuitive sense for a broad range of materials, with the figure of merit derived directly from the atomic positions in the crystal structure. This is the kind of mistake that someone with no domain expertise in materials science might make.
"This work was supported by the George and Obie Shultz Fund and the National Science Foundation Graduate Research Fellowship Program under Grant no 2141064. IRB approval for the survey was granted by MIT's Committee on the Use of Humans as Experimental Subjects under ID E-5842" https://couhes.mit.edu/policies-procedures/review-and-approv...
So the PhD student might have been kicked out. But what about the people who "championed it". If they worked with the student, surely they might have figured out the mythical lab full of 1000s material scientists might not exist, it might exist but they never actually used any AI tool.
It doesn't excuse the fraud of the junior person but it makes you think how many senior-level people out there are riding similar fraudulent waves, doing zero checks on their own junior people's work.
Basically, science is quite vulnerable to malicious exploiters. Part of this is because society isn't funding science anywhere near sufficiently to do a priori in-depth checks. You claim you got data on hundreds of measurable thingies in a certain way (from surveying people to scanning the web to whatever)? If it's not blatantly obviously a lie, it'll probably be accepted. Which is inevitable: at one point, you're going to have to accept the data as genuine. If there's no obvious red flags, you'd only waste time on further checking data - you'd need to do a real deep dive (expensive time-wise) to come up with circumstantial evidence that may still be explainable in a benign manner. For scientists, it is almost always more profitable to spend such time investments on furthering their own scientific efforts.
So yes, there are various ways in which someone willing, dedicated and sufficiently skilled can "Nigerian-Prince" the scientific process. Thankfully, the skill to do so typically requires intimate knowledge of the scientific process and how to conduct research -- this cheating is not easily accessible to outside bullshitters (yet).
Why not?
b) I don't see how calling them out by name in this forum might be helpful
c) I don't know the whole story - maybe some of these people actually raised concerns or wanted to but felt powerless; causing people, search engines and LLMs to associate them with this case of scientific fraud seems potentially bad
d) maybe I'll have the chance to discuss this paper with them on social media or in person - then it would not be good idea to publicly shame them now.
e) anyone can quickly find out the names by reading the paper (https://arxiv.org/pdf/2412.17866)
Edit: It was the WSJ article: https://www.wsj.com/tech/ai/mit-says-it-no-longer-stands-beh...
[0] E.g. arxiv/0812.0848: "This paper has been withdrawn by the author due to a crucial definition error of Triebel space".
Peer review is not well equipped to catch fraud and deliberate deception. A half-competent fraud will result in data that looks reasonable at first glance, and peer reviewers aren't in the business of trying to replicate studies and results.
Instead, peer review is better at catching papers that either have internal quality problems (e.g. proofs or arguments that don't prove what they claim to prove) or are missing links to a crucial part of the literature (e.g. claiming an already-known result as novel). Here, the value of peer review is more ambiguous. It certainly improves the quality of the paper, but it also delays its publication by a few months.
The machine learning literature gets around this by having almost everything available in preprint with peer-reviewed conferences acting as post-facto gatekeepers, but that just reintroduces the problem of non-peer-reviewed research being seen and cited.
> you will never be made aware of the disputes surrounding the research
ArXiv is not peer review, so it's as confiable as Wordpress or Medium or Blogspot or X/Tweeter or ... The main difference is that the post is in PDF instead of HTML. There is an invitation system to avoid very stupid cases, but it's very week.
I remember a weird cryptography "breakthrough", and they published it in arXiv, and the first 5 pages were an explanation of the rule of 9 for divisibility and then a dubious fast factorization algorithm. [The preprint was linked in https://news.ycombinator.com/item?id=20666501 and I replied to it (s/module/modulo/g)]
Explaining why research seems sus (what the comment does) is some form (strength proportional to the explanation) of signal.
I pointed out one issue with the GGP. That's not contrarian and not about everything.
> it is somehow NOT nice
Who said that? How could you possibly get that out of my short comment? Are you just making things up? Then what you say also becomes just noise, not a signal.
> existence of negative comments is "not a signal" when it clearly is
It's clearly not and I demonstrated why. What is your argument against my reasoning?
A fundamental of information theory (or decision theory) is that if something gives you the same response every time, it's noise and not signal. If the weather forecaster always predicts it will rain, they aren't 'correct' when it rains; their predictions are just noise. If your coworker criticizes everything you do, their criticisms aren't signals about your work, they are just (annoying) noise. It's no more meaningful than water dripping on a drum.
It can't be the same every time when the entire comment consists of specifics only concerning one this paper.
Artificial Intelligence, Scientific Discovery, and Product Innovation [pdf] - https://news.ycombinator.com/item?id=42115310 - Nov 2024 (47 comments)
Edit: Since the paper has been cited, others may still need to reference the paper to determine if it materially affects a paper citing it. If the paper is removed it’s just a void.
[1]: https://info.arxiv.org/help/withdraw.html#:~:text=Previous%2...
[2]: https://economics.mit.edu/news/assuring-accurate-research-re...
> The paper was championed by MIT economists Daron Acemoglu, who won the 2024 economics Nobel, and David Autor. The two said they were approached in January by a computer scientist with experience in materials science who questioned how the technology worked, and how a lab that he wasn’t aware of had experienced gains in innovation. Unable to resolve those concerns, they brought it to the attention of MIT, which began conducting a review.
Seems pretty serious if they kicked him out.
Lay low for a year, work on some start-up-ish looking project, then use his middle name to get hired at one of the many AI startups? (only half joking)...
Also, there are companies who will see that win at any cost mentality as a positive trait.
I'm betting whoever it is, is okay now.
They may have thought they could jump into an industry job, including the paper and all of its good press coverage on their resume. Only the author can retract an arXiv paper, not their academic institution. It wouldn't be hard to come up with a story that they decided to leave the academic world and go into industry early.
MIT coming out and calling for the paper's retraction certainly hampers that plan. They could leave it up and hope that some future employer is so enamored with their resume that nobody does a Google search about it, but eventually one of their coworkers is going to notice.
He probably will never be someone of significance, but he also will probably be able to have a standard middle class life.
That high-level Apple employee was probably a manager and oversaw hiring people.
I would tell myself every day, "I wouldn't hire me."
It's not self-defeating.
It's not being a victim.
I wouldn't let it stop me from trying.
It's being accurate about what kind of company you'd want to build yourself, and the internal state of a lot of hiring managers. And with a true model of the world you can make better decisions.
E.g. Kaavya Viswanathan, who plagiarized a high-profile bestseller while a freshman at Harvard, went on to Georgetown Law and a job as an attorney as a blue-chip law firm.
And there are many other examples ...
I strongly recommend people not investigate this unless they think 4chan is quaint. As in, if the reason you are not using X is because of the outrage at Elon and the "typical user" of X, then maybe use xcancel instead.
in the space of working on cool shit, he's good people. see: moldbug and urbit, for a similar case.
So i guess i misread!
Twitter's issues long predate Musk : for instance Mastodon was created as a reaction to the 2012 APIpocalypse.
That's a paradox. Maybe it's just me but if you need more than 3 "posts" to say something, put it in a blog and link to it from Twitter. That was excruciating to try and get through (which I failed at doing.)
non-scientific studies can't be replicated
As a CS PhD who has worked with many economists, my understanding is that the culture sees it as diluting credit, and so might put people in the acknowledgements where computer science folks would add them as authors.
This is how I see this: They published a false paper, knowing they would eventually be caught. But the boost that the paper had in AI marketcap and market bullishness and hype, far outweighs the consequences of this apology(not really) statement.
So they got a lot of money for hyping the AI, and now they pay off a small amount of it as backlash. still a huge net income. Pharma companies do this all the time. this is how they make money.
Reminder that if you link articles for your argument, doesn't necessarily mean that you are right. There are substantial amounts of false and slightly distorted papers and articles out there, even from the most valid publishers.
Most of the citations are to other papers on preprint servers and I guess a lot of people were working on similar themes and added it as a reference
It sounds like "we don't like it and won't tell you why, we're hiding behind MIT policy and vague notions of privacy".
MIT should just demonstrate in a paper what the shortcomings are and print it, adding it to the citation tree of the original.
Looking very briefly at the paper and speculating wildly, I could imagine that the company who were subject of it - or their staff - might not appreciate it and have put pressure on MIT??
Solid amount of Streisand Effect going on here -- lots of attention has been bought to the paper (and that is everything after all!).
The apparent issue is that the data appears to have been entirely fabricated and is a lie. The author appears to simply be a fraud
FERPA is federal law. It is quite likely that MIT is legally bound to not release some pieces of evidence which are crucial in this case (hypothetically, for example: that the student's educational record is inconsistent with claims made in the paper).
(Incidentally, I don't think misplaced trust in preprints is much of an academic issue, people that are experts in their field can easily judge quality for themselves. It's laypeople taking them at face value that's the problem.)
This case seems a counterexample though. Top economists Acemoglu and Autour went on record approving of the paper. The main flaws were only obvious to material scientists, and since this was submitted to an economics journal it's possible that if it hadn't been so hyped it would have slipped through peer review by economists
*Note that this creates a new version lacking any download which also becomes the default but any previous ones are still available.
Compare the failure rates for condoms.
Better title:
MIT disavows heavily-discussed economics preprint paper about Artificial Intelligence and Scientific Discovery.
(Since press release titles about negative news tend to studiously avoid saying anything, we tend to classify them in the "misleading" bucket of https://news.ycombinator.com/newsguidelines.html, which justifies rewriting them.)
[0]: https://info.arxiv.org/help/withdraw.html: "Articles that have been announced and made public cannot be completely removed. A withdrawal creates a new version of the paper marked as withdrawn."
Whether MIT is right or wrong, the arrogance displayed is staggering. The only thing more shocking is that obviously this behavior works for them and they are used to people obeying them without question because they are MIT.
I think MIT is trying to protect its reputation as a would-be place of fraud-free research, unlike Harvard.
just post a correction notice on arXiv. let others decide if there is merit to it or not.
silencing is so anti-science. shame on MIT.