MIT asks arXiv to withdraw preprint of paper on AI and scientific discovery

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
  %

(In this q&a, the audience does not really question the validity of the research.)

https://doi.org/10.52843/cassyni.n74lq7

Nice to include the giants we stand on as implied coauthors.

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.

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.

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.

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.

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.

There was a recent podcast that covered it with some experts that's a great listen:

https://open.spotify.com/episode/5w1gdbhmlCyTapoQ3EkMHp

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.

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?

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.

I’ve even worked in places where some ML researchers seemingly made up numbers for years on end.

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.