This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
This does not quite fall under culture wars, but I thought it's still relevant given that it's a big story.
On October 6, OpenAI released a large batch of mathematical results, with more than 700 manuscripts covering hundreds of problems: Sharing AI progress in mathematics.
This has led to another round of controversy and discussion about AI making mathematicians obsolete or disrupting research programs that people have spent years working on. The Verge covers some of these reactions in “Mathematicians will need years to make sense of OpenAI’s latest drop”.
Many of the results are still being checked, but the release has called into question if mathematics can survive as a profession, or what purpose it serves if machines can outpace humans. Research programs have supposedly been gutted by these results. "Careers ruined," as the narrative goes. I think that is an exaggeration or overreaction. There are thousands of mathematicians and their interests are very wide ranging. The odds of someone's research program or problem being "solved" by AI are very small.
I don't see anything changing much. The status quo of journals and academia will remain intact even with rapid progress in AI, save for the occasional breakthrough in an unsolved problem. What is overlooked or ignored is the massive gap between cotemporary research-level math vs. LLM-produced math. A math PhD program, especially from a good school, teaches or guides future researchers towards productive areas of research that are of interest to the "math community," rather than results that are merely "personally interesting". AI cannot recreate such communities or networks.
AI is great for generating computation-laden results on-the-fly, but research-level math is more concerned with abstractions and relations between them, than crunching. I recall one example where ChatGPT’s Astra program tried to pass off a finite computational cutoff as a “theorem.” Um, you can't do that unless you prove that the bound is sufficient.
To test this, I produced some AI papers which I thought were great and novel, and the AI said so, and journals rejected them (although others are still pending at various stages of peer review). I had some knowledge in said area and could assess novelty, but still not good enough. I don't see laypeople doing better in this regard, although could be wrong. I ran my ideas through the Ai and got 12 papers out of it, which saved considerable time had I written them myself. It also found non-obvious connections and results. I am really satisfied with the results, but does this meet the bar for publication? Time will tell. I bet not.
Also, success is heavily dependent on the prompt. It's not as easy as blindly prompting something. I made a post about that https://greyenlightenment.com/2026/10/11/the-nuance-the-math-debate-is-missing/ I suspect many of these successful proofs were prompted by experts who knew what to look for, rather than just naively trying. What is touted by the media as an AI breakthrough, may involve surgically precise prompting in the background by an expert in said field.
Alluding to the "airplane meme," there is also survivorship bias: you don't see all the failed attempts with AI . Sometimes it works great, but other times hit or miss. Presumably , the "top unsolved problems" have been queried by frontier models and using large token budget, with few proofs to show for it (save for the Open AI "dump").
It would be cool to see AI truly shake -up math and breathe new life into ignored areas, but that is not going to happen, imho. But I do see AI leading to a 2-track system of decentralized preprints and GitHub repositories of AI-produced results, alongside established academia and journals. They will coexist, rather than the later being obsolete. Overall, mathematicians don't have to fear losing their jobs anytime soon (unless of course you happen to have the precise problem you are working on scooped by AI, and even then, your skills would presumably transfer to new problems.
I may write a long post of my thoughts about this, because coding LLMs have also become very impressive (and reliable) very recently. In summary:
Research has stagnated, with low-hanging fruit claimed decades ago and researchers only making novel contributions in their late 20s (even many PhDs today aren't really novel). Academia has also become burdened with bureaucracy: bloated administration, "publish or perish", and "researchers" who game the system without producing anything worthwhile (because even those who do produce must game today's system a little). It has been dying long before LLMs.
LLMs open the door to a new realm of opportunity. I expect many more groundbreaking discoveries going forward, of people who apply them in creative ways and others who follow. Like the proliferation of computers or the Internet; most would agree LLMs are at least as big a milestone. And if academics play their cards right, a resurgence of impressive discoveries is exactly what will revive their public appearance and funding.
But LLMs also close the door to most of the current realm. Most work that would be PhD-worthy before is now saturated by LLMs, who can do years of grunt work in weeks (where lower quality is unnoticed or tolerated). We also don't know how much better they'll become; I suppose in math, they already surpass Tao himself. People who've devoted years of learning and effort, now finding that easy to replicate, are feeling left behind, and some will be left behind, because they have to start over...
...well, I really hope they won't have to completely start over: I believe they should get legacy funding and support, kind of like the grant given to Tao, even if their work is completely eclipsed by LLMs - the door to financial stability and recognition artificially held open for the last cohort. Because otherwise no-one will start future long-term projects for fear of getting burned. But obviously new PhD students should not do what LLMs have trivialized, or are likely to barring true ASI...which leaves open the question what will they do, but this is a question that has always existed, and which most spend their first years figuring out anyways.
As I've said before, LLMs may lead to utopia or dystopia, but I don't think of them as good or bad, and I don't think it's worth most people's thought, because they are practically inevitable. What matters is that they are transformative, so we should figure out what they may transform, how, and prepare. I wish academia would adjust: gift-wrapping up obsolete work, leaving those researchers with their currently-allocated finance and prestige (including tenure) but expecting them to try to transition into something new...but unfortunately I slightly expect it to continue as-is, with funding for an unreasonably long time (because it's so entrenched), but that may not be long at all, and/or what it becomes to maintain the funding may make today (or 6 months ago) look like the ideal of pure scientific advancement.
Anyways, how may human-directed research, and math in particular, evolve? I'm almost certain it will stay and actually think it will grow. This is what I think is actually interesting (whereas, is any of the above unexpected?), but also more speculative and philosophical, so I have to think and plan it a bit longer.
One of the problems of academia is that the amount of material to review and understand has absolutely exploded. This might be merely annoying in, say, literature, but it's ideal in history and obviously potentially very important in medicine to be able to wrap your head around all the research that's been done ahead of you.
I think LLMs are in many ways a potential lifeline out of this dilemma that we've created for ourselves. But in a very real sense I am worried that they will work too well, sucking all of the funding away from experimentation and pushing boundaries and redirecting it into using LLMs to make new connections in already published data. This might yield incredible results, but if we don't maintain empirical testing, at some point we'll hit a dead end, and my concern is that we won't hit that dead end until after we've lost the cultural knowledge of how to do empirical testing.
This is particularly important because, due to the replication crisis, we could make vast forays into realms of theory without ever realizing that our work was built on a lie if we don't bother to verify the experiment.
However, this does suggest another potentially useful feature of LLMs for science beyond using them to correlate all the contents of the world: an "auto-replication factory" where some university has an agentic swarm review new publications. Obviously you'd need a lot more than just an Astra swarm to test the more hands-on stuff, but I bet AIs are decent at catching statistical anomalies, mistakes, fraud, and plagiarism.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link