The vitriol from the gen pop towards the Masullo announcement makes me welcome total civilizational capture by the labs.
Book us all a one way ticket to the permanent underclass already, Dux Dario. We cattle yearn for it!
Aniket Das retweeted
GTA VI isn't that detailed. In real life, charge-parity symmetry violations are known to occur in the weak decays of both mesons and baryons.
Aniket Das retweeted
The future of academic mathematicians is ghost-journals and YouTube shorts.
The future of industrial mathematicians is war crimes and corporate PR.
Aniket Das retweeted
I roughly agree with the sentiment, except I don't see why this would necessarily attract large audiences outside the individual's own institutions, because an average early career mathematician is not doing work at the level of those presented at Seminaire Bourbaki. Even if their AI is doing such work, I don't think I personally (and many others) would attend such ‘courses’; it's easier to make the slop accessible to everyone so they can prompt their own AI's to produce slides, make expositions etc.
Seems like any such attempt at a course or seminar on one's own research would devolve into cynical networking seminars where if people attend it is only to vulture around to know people that can help them with job offers.
I’ve been trying to think more about what human academic work products should look like as AI becomes increasingly capable. It seems likely to me that papers will decline dramatically in value as it becomes easier to automate their production; possibly the median paper will be one that no human, not even the author, has read. It is not clear to me how many traditional institutional organs of academia (journals, metrics, etc.) should adapt to this.
A couple of models I like, which already exist in mathematics but are relatively unusual, include: courses on one’s research (with accompanying notes), or research seminars (with proceedings), à la Grothendieck’s SGA seminar. These have the benefit of keeping humans in the loop, and also have built-in mechanisms to ensure that participants are gaining some expertise.
Aniket Das retweeted
I genuinely don't get why faculty do this. Researchers are horrendous gossips, news inevitably spreads, and it looks bad to seem like you're hiding an affiliation. It's a particularly big CoI if you're claiming to be representing academia.
Just disclose it. It feels good.
Replying to @nihilunbounded
Brother a large fraction of these of these ppl don't even declare their biglab affiliations on their website (which i rly dont get why). You gotta rely on word of mouth (or xitter drama) to even get to know about it. I'll eat my hat the day they start declaring benefits.
Aniket Das retweeted
The authors take full responsibility for all contents of this manuscript, the same way they have fully read and agreed to the terms of service of every app they’ve ever installed.
Aniket Das retweeted
Dear Professor, I quite strongly liked your statement on the Proofs & Prompts 100 reactions post (screenshot attached for more visibility), which emphasizes the outsized power and control over academia that these big AI corporations hold. Given this, I am a bit taken aback by your current view. Should one not view this power as fundamentally corrosive and be as skeptical as possible when it comes anywhere near academia?
I'm not sure that normalizing this unequal tension, in which academics have no levarage and have to constantly reorient themselves to the whims of the AI labs, is the right way. It essentially leaves trillion dollar corporations, with very little care for the human cost of their automation crusade, to run rugshot through hundreds of years of academic norms and ethics that are an Enlightenment-era artifact.
Replying to @ketd47 @nihilunbounded
The frontier labs hold a lot more power, and shunning them is unlikely to be a good option for academia. Also, lately they have been more open to discussions. For example, @OpenAI has shared the prompt that led to the ETH breakthrough etc.
Aniket Das retweeted
Peter Woit on the Shepherds: "Now that they have released the Lean code, their plan to keep secret the 1147 page document explaining the proof and devote great efforts to understanding it and explicating it themselves makes no sense."
Indeed. This makes no sense!
math.columbia.edu/~woit/word…
Aniket Das retweeted
Replying to @danintheory
Who Shepherds The Shepherds?
just kidding, I'm very excited for the sheep as well.
Couldn’t have said it better. For the same reason, I find it p disgusting when papers where most of the work was done by AI bury their declaration at the very end. Comes across as super slimy and costs nothing to declare up front
Replying to @mathandcobb
For a smell test, take Alex's text and think "stolen from someone else" instead of "used AI" and see if it matters:
"So what if it was? Is it a valuable contribution to the scientific literature or not? For the purposes of hiring: does the person claiming the result actually understand deeply the subtleties of the result, and can communicate it effectively to others? That's what actually matters."
obvious cranks aside, im way more disappointed at the number of ppl who are part of the community and are making 4-5 ai assisted submissions, nearly maxxing out their quota. y'all know who you are, and thanks to subs being public, we all do too.
STOC 2027 received 1665 submissions. In a bold new experiment, all submissions (and author identities) are immediately made public. This is the most immediate and revealing posting of any conference I am aware of.
Take a look and see what you find!
stoc2027.sigact-ec.workers.d…
Aniket Das retweeted
Replying to @nihilunbounded
OpenAI were primarily concerned with advancing human knowledge -- not credit -- when they dedicated $10m+ of compute to Navier-Stokes upon hearing rumors that Levent and Buckmaster were close.
Aniket Das retweeted
Replying to @tonylfeng
But unfortunately this is not obviously a stable state because you only develop the intuition to discover new questions by struggling with the old ones, which doesn't happen if you just consult an oracle.
Aniket Das retweeted
Masterful maneuver by OpenAI. Simply offload all legal responsibility onto the user.
Replying to @seanhn
im verified on daybreak and my sessions are mostly other peoples unpatched bugs
if thats true theyd be training on stuff im not even allowed to share
Aniket Das retweeted
My previous post started to stir up AI hype, which was not the intention, so I deleted it. Suffice it to say that the world is changing fast, and people should be preparing carefully and thoughtfully.
Regarding mathematics specifically, it has been a cornerstone of human intellectual flourishing, and I certainly wish for that to continue. I hope there will be spaces for exploring mathematics in new ways, but also spaces for preserving the old artisanal ways, and that people think about how to sustain such environments.
























