transcript
Screenshot from Thomas Dietterich on X: “Attention @arxiv authors: Our Code of Conduct states that by signing your name as an author of a paper, each author takes full responsibility for all its contents, irrespective of how the contents were generated.”
With a reply from James Miller: “So this means you expect every author to check every citation and make sure that every citation is real and accurate? What if it’s beyond the ability of one of the authors to verify one of the citations because that citation is in a language he doesn’t know or concerns technical material he doesn’t understand but another author on the paper does?”
You must log in or # to comment.
Yes and fucking get to work
So this means you expect every author to check every citation and make sure that every citation is real and accurate? What if it’s beyond the ability of one of the authors to verify one of the citations because that citation is in a language he doesn’t know or concerns technical material he doesn’t understand but another author on the paper does?"
Why would you cite a work you hadn’t read and/or don’t understand?
It’s quite common in interdisciplinary papers that some of the authors cover a part of the paper in which they are expert while others cover another part.
It is uncommon in those cases for all authors to read all referenced papers.
I mean in that case then the question is “do i trust my other authors”
Indeed, you should trust the other authors but that is the whole point of the response in the original post.
I know this is a meme post and I apologise to the OP for the incoming rant.
When someone proposes, implements or enforces a clearly sensible rule, and someone else brings weird corner case scenarios up, always ask yourself if there’s a conflict of interests. Sometimes there’s none, but often there is — undeclared and disguised as concerns about something else.
That’s the case here. Check James Miller’s XCancel profile and you’ll see it.
excepts from his profile; emphasis mine in all cases.
[Description] Smith College economics professor. PhD Chicago. JD Stanford. AI safety, game theory. Stroke survivor, hoping to make it to the singularity.
[pinned tweet]Nerd score: How many have you considered? // Cryonics, multiverse, Boltzmann brain, AI utopia, quantum immortality, Roko’s basilisk, gray goo, paperclip maximizer, great filter, ethics in infinite universe, acausal trade, longevity escape velocity, simulation and zoo hypothesis.
[tweet]Suppose AIs become superhuman at math within two years, as far beyond humans in math as they already are in chess. What practical breakthroughs might follow, perhaps room temperature superconductors?
[retweeted from another]I would be mortified to have a typo—never mind a hallucinated citation—in a paper. But you see from twitter threads that some people think having a tidy bibliography is the definition of good research. They’ve got a 6th grade report-in-clear-plastic-binder view of the process
[tweet]I tell my students that they have to use AI to help write papers. I give them guidance on how they can effectively do this. I think I’m giving them more practical help than if I forced them to write without AI. This past semester, one student even asked me if she had to use AI and I said yes.
et cetera.
Note how he’s too invested into large “language models” to admit they’re often a source of misinformation. To the point he’s telling his students the equivalent of “scientific paper, toilet paper, same thing lol, just add shit lmao”.
He’s clearly ignorant on why references are such a big deal in science. When you write a paper, you must be able to tell people where you got the info from, otherwise the whole thing devolves into “trust me” = “I think you’re gullible filth”. With “trust me” there’s no knowledge being shared, just a bunch of bullshitters repeating the (often incorrect) assumptions of each other.
And odds are he gives no flying fucks about either “concern” he raised. Specially because the solution for both issues is simple, as long as you care about science instead of “me publish paper lol lmao”:
- language barriers: your research should not rely on things you do not understand. So work with a translation of the work, translate it yourself, or don’t quote it.
- co-author adding references LLMs made up: why are you co-authoring a paper with a gullible muppet who uses LLM output as source of [mis]info???
/rant
We might never reach singularity. I think our universe is a simulation built to study how it happened in base reality - or what kinds of singularities aliens build. The event itself is probably too expensive to render. Our universe will cut to black.
I feel a little bad dunking on a random guy, but this is very stupid
Usually I’d also feel bad dunking on a random. However, when that random does a disservice to the scientific community, I think it becomes fair game.
Specially in the light of the ongoing replication crisis. There are multiple reasons scientists are having a hard time reproducing published results, but a lot of them boils down to “someone skipped proper procedures” (like he encourages people to). Peer review is supposed to catch this, but when a person who can enforce those proper procedures says “we’ll enforce them”, suddenly the same random makes up reasons against the policy.
It is always morally correct to dunk on AI boosters
Wish I could Wayback his profile and see when he had the Rick and Morty PFP
Don’t tell that guy about free translation tools. Or about trusting your colleagues.
I’m a strong advocate for trusting your colleagues (and verifying). No disagreement here.
You should know, however, that free translation tools are AI. It’s emphatically exactly the same technology that gave us slop cannons in the first place. The birth of the transformer architecture came from trying to do improvements in machine translation.
That might be too tangential or pedantic, given I think we share the sentiment about this dude, but it feels worth mentioning.
Or, like, not citing what you don’t understand
What a fucking imbecile LOL
If you can’t check all your sources and/or can’t trust your co-authors, maybe science just isn’t for you?
Amazing to watch the foss owners fight to retain control of the code by turning down essentially free labor adding features and fixes to their project. And here too.
Wrong thread?
Relating the two.
