We are arguing here about whether to allow posts of AI-generated content or not. If the predictions of the release of sovereign agents come to pass, these deliberations will be moot. There will be no need for humans anymore.
Check out We have started losing control of AI. It’s time to shut it down | Garrison Lovely | The Guardian if you don’t know what I am talking about.
Edit: Another just-posted article Why some experts increasingly fear AI will take over
I personally still enjoy engaging with other people here in the forum. There have been concerns about how we communicate, regarding the use of AI output etc. and I think it’s important that we discuss these issues at the moment (some things are certainly changing and some also quite fast).
But are you saying that we shouldn’t bother at all and stop doing anything right now (discussing things on this forum and discussing how we discuss things on this forum), because there is a (unknown) chance that things will get worse and that there “will be no need for humans anymore”? In that case, I disagree. Whether we “need” this forum depends only on whether people (and I really mean humans) want to use it or not, in my opinion. If you ask me, it is, ultimately, a discussion forum for humans…
Do we really need computers, or the Internet, or electricity? It all depends on what you mean by “need” 
Not at all. My point in fact is that it is high time we took a stand.
Oh I see. Then I misunderstood your original post (I was a bit surprised too).
It’s for fun.
I think sure, getting work done is now all about AI generated PRs and capturing lots of information in issues. That gets more done. But I like read non-AI stuff here, so I still do it as a hobby.
Is model collapse not a concern any more? Have they solved that problem? If it hasn’t been solved, then maybe the AI models need us (creating genuine human-authored content in places like this) more than we need them. Or, are they able to harvest all the training data they need now from direct interactions with their tools?
It still is, but it’s managed by smarter (and more expensive) collection of real data and in some cases formal verification.
For one, I could argue that the world never needed humans in the first place, and that we have become a net detriment before AI generation ever took off. But if we assume humans deserve to exist, then there will always be a need for us. We’re the only ones capable of expressing our various conflicting whims; no machine can do that even if they became sentient superhuman lifeforms. Sure, the threat to livelihoods and decline of various human activities sucks, but that isn’t what is being discussed now.
The actual problem is that AI is currently not sentient life with any sense of morality based in the real world. Any model makes mistakes, LLMs hallucinate, and agents are a large game of prompt-telephone that can reach unexpected prompts and context. Putting actual tools and decisions in that mix to affect the real world was always risky, and people have already been hurt. Models are only getting more capable, and there is valid criticism that frontier labs are approaching this as an arms race. We should be far more worried about AI wrongly denying healthcare coverage (Estate of Gene B. Lokken, et al. v. UnitedHealth Group, Inc.) than AI agents solving the two-language problem without us. Granted, we can still debate smaller problems like whether we need this forum.
If you think about the OP more carefully, you will realize I did not propose we discuss the need for this forum.
The warnings continue to come: The Guardian view on controlling AI: humanity cannot outsource its survival | Editorial | The Guardian
Is anyone listening?
The paradox is that to protect our institutions and our system from technological “progress” and the imminent final assault of AI, we can no longer afford to disconnect the internet without causing societal collapse. Unfortunately, the UN is weakened, and the enforcement of any global AI regulations seems illusory.
Forget about the forum, isn’t the looming question:
Do we need scientific institutions anymore? (and all the tooling that comes with it)
I had an interesting debate on that about a year ago with a Professor from Berkeley. He said that in a way AI, just like the internet before, is pulling the curtain from the institutions. It opens up knowledge and capabilities to everyone. It cheapens everything they produce. It is cheaper as a tool to use than hiring researchers. It is also used by students - more than anyone would like to admit and available to everyone, outside of the field.
Most over here use Julia for scientific projects, or simulations. But what’s the benefit if you hand over more and more to a digital assistant that can easily recreate it in C++ or Rust. So as we are talking about doom - nerdy forums like ours may continue for a while, but who will it be for in the future.
You have not spoken to students recently, have you?
I do, because I teach. Students do not want to use AI, they want to learn.
Nothing is cheapened. It is not cheaper than hiring researchers. Have you heard the number OpenAI claimed to have spent in tokens when they stole the results from other researchers so they now can claim to have solved a Milennium problem?
And sure, I have a few colleagues (Math department here) that for several weeks now run around and only scream “we are doomed!” – I heavily disagree.
I recommend as well to for example read “Against the Uncritical Adoption of ‘AI’ Technologies in Academia” by Guest et al.
So please stop writing unproven claims AI would cheapen anything. As that paper compares quite well, then you can also claim we all should smoke (then we are all more relaxed!) or use more petrol (we are so much faster! No one has to ever walk again!).
You sound offended. I did not mean to 
Yes indeed. Because
Is utter and total nonsense. Without any proof/evidence and even (see my reference) the complete opposite is true.
I saw you posting this experience with your students before and honestly: kudos that you had that.
My own is quite different. And though I am not an academic, I don’t think it is less relevant as an opinion.
I closed my own digital agency three years ago, precisely because I saw that the business model of selling people to do development work was coming to an end. It was a pretty rough decision and ended a 15+ year run working on large scale installations together with my team.
You make it clear that your own world won’t be as affected. I am rooting for you, but please have the decency to explain what makes your job so special.
My job is not special. I am just a mathematician. There is very many people much more clever than me in the world that could do my job much faster.
Luckily most of them find probably very boring on that I research and work on.
Sorry to hear that you had to close down your business.
am not claiming AI has zero use. It probably has a bit of use here and there.
But it will not replace personal communication, exchange, teaching and learning.
Please read the paper above – but here is one argument they also state. Sure we can leave students alone and tell them they should write their thesis with an AI. Do they learn something with that? Probably not. There is a reason we do not hand calculators to first year school kids.
And again – that is a bit of a higher level with students and AI and such. Sure they can solve first and second year math exercises with AI.
The AI companies scraped all the university homepages for all the solutions of existing exercises by now and learned their models on them. They give a shit about copyright or ownership or any moral responsibility.
Do students learn something when they ask an LLM “Solve this exercise sheet for me”? No. And they know that. I even started an additional class hour this semester to provide more assistants in person, not less. Because students ask for that.
Also – only the rich students do buy the plans. We are further dividing societies here.
So for these reasons your initial anecdote is utter nonsense.
I am especially tired that since this year I get told at least twice a week, that my standpoint on critically reflecting where to maybe use AI and where definetly not is wrong and I should only hail to the AI gods and otherwise I am doomed.
I am quite sure I am not. Sure I reflect on AI use here and there (see the other forum posts here for example). But I would prefer if we could get rid of all these content thieves.
But I am really tired of this, so maybe I stop here and if the forum here – as it seems – turns into only having these discussions why the should all Hail the AI gods and that is the only way to continue, then you are right, I am probably wrong here.
I am still confident that this forum some day can return to actually talk about using Julia and nice package developments again. And if not, then I am maybe just a boring small mathematician sitting in my office coding and writing small papers by myself without AI – and teach students math topics I like.
No need to stop posting. And I get why you feel tired. But perhaps the most obvious choice is to just ignore the threads that are quite obviously on the topic of the impact of AI then?
It is the reason why I stayed as far away as I could from that other rage baiting topic. I had high hopes that this one would be different…
That said, it is great that you contribute a different opinion. That’s why I posted mine. Perhaps next time we can go about without calling the other one “utter nonsense” just because yours differs?
But then how else should I refer to a statement that can not be further away from the truth?
Sure maybe there is one or two professors that claim that they now do everything with AI. I would claim, that the results they obtain are no longer research – in the several aspects now mentioned.
And maybe they are as rich as the AI tech bros are now as well and they can afford it. That still does not make it true.
But again. I probably really should leave these topics be – and maybe also this forum. I am tired and do not have the feeling this is useful anyways.
You do realize that just this week openai released the solution to no other than a century math problem. Is it “stolen” off of other people’s research? Most likely but the result is the same - the ai model came out on top.
I am not devaluing the work of actual humans or their creativity. I think you got a wrong impression of me here. I think that AI is only as creative as the most capable person in the room. But assuming that companies and institutions are seeing the capabilities play out like above, what is more likely to you, that they increase the headcount or that they reduce and make up for it with AI tooling instead. How many students will you have in the future if their knowledge does not guarantee them an entry level position anymore?
Those are topics worth exploring, and I don’t think that calling them BS or nonsense make them go away.
PS: AI is capable of creating papers now that clear peer reviews Towards end-to-end automation of AI research | Nature
You do realise that it is very probable they stole the result from other researchers that used some LLMs as a guide?
And upon discussion they claimed they would “include” one of the authors, but not the other, since he is (besides that project) employed by Anthoropic? See also https://mathandai.org
It is unclear whether their result is purely based on stolen data and results. They will not (and probably even can not) make that transparent. That is neither morally responsible nor is it science. It is pure PR and killing science by theft.
The mere point that people claim “but you will not get jobs!” is something that is also wrong. One example is the claim by colleagues of mine, to get more money for tokens to keep up with the large companies. They burned several million dollar worth of tokens in just a few days!
And yes I call BS on statements like “will you have students when their knowledge is useless” – because what they learn is not how to generate the solution of an exercise sheet with am AI. Learning (and Teaching) is much more than the result.
Please read the article I linked. We need a responsible way to handle LLMs. Starting with a proper naming and no PR-bullshit-bingo in claims.
Yes, currently computer science first year students here for example have tremendously dropped. The propaganda you repeat has its effects. Good luck finding reasonable computer scientists once you notice, AI can not replace human computer scientists. That is what we currently loose. Not a race against AI. Good luck with software, when we do no longer educate computer scientists.
I am very sure, we will have enough math students in the future if we do not fall for all the PR of the AI gods.
But again. I am tired of all this and need a break. You at least managed that I will leave this forum and take some time off. That’s what you definetly achieved.
That is of course also a way to hail the AI gods, bombarding all criticism with more results the AI claims to have achieved, not reflecting on it, just repeating all the PR results.
PS: Yes anecdotally AI also writes papers that pass peer review. Peer review was broken before AI; if you pay enough you get anything published. That is a problem of the publishing system that AI just makes more prominent now.