-4
$\begingroup$

I am a mathematician with a PhD from 2016 and a long-term user of this website, though I've not been active of late.

I am wondering, should Mathoverflow take a stand against Stackexchange for their use of AI? I believe AI is fundamentally damaging in the long-term to society in many ways, some of which are:

  1. AI removes human connection from the equation of information sharing. Sharing ideas and information is a fundamental cohesive force in our society and sharing information has always been about two things: the information itself, and developing relationships with other people. AI removes the latter, and creates a world where information sharing is anonymous. Of course, Stackexchange's initial foray into AI isn't a complete replacement, but it is an augmentation that will significantly distort human interaction in a negative way and eventually lead to a greater degree of anonymity.
  2. AI on a greater societal level is poised to take over almost every sort of creative job, including professor. Previous technologies have never been so poised to take over so many creative jobs. Of course, it's unlikely that AI will be able to do mathematics research any time soon (though it may do so in the future), but as we all know, mathematics is primarily a service department for other sciences and soon that function will be eliminated.
  3. In the long term, AI taking away jobs such as math professor, artist, and writer, will cause a mass loss of purpose in life and thus a mass wave of depression and lack of direction for humanity. Jobs will be increasingly meaningless, existing merely to further developing more advanced AI systems. The original motivations of research were curiosity and improving the human condition, but we are reaching a tipping point where those values will be stripped from us.
  4. With great power comes great responsibility, as the saying goes. We hardly have the ability to handle basic power with responsibility (c.f. fossil fuels, the internet). AI is simply too powerful to be used with responsibility by humanity.

In closing, as mathematicians, we have a special viewpoint about the world. We can envision complex systems and far-reaching consequences. And, if we look at such consequences, we must conclude that AI is not just about some minor improvements to search and answers on Stackexchange (the short-term), but about the long-term societal effects of making the world devoid of human connections because removing such connections is the most efficient optimum for big-tech companies to concentrate the wealth of the world into the hands of the few.

Thus, I believe that Mathoverflow should take a stand against AI and do everything possible to remove Mathoverflow from the Stackexchange network. Of course, this might have some short-term pains, but it would send a serious message that society at large should be more wary of what is going on.

$\endgroup$
13
  • 1
    $\begingroup$ You start out "wondering, should Mathoverflow take a stand against Stackexchange for their use of AI?" but you seem to have convinced yourself ("I believe that Mathoverflow should take a stand against AI"). But more to the point, I don't know what you are referring to when you say Stackexchange is using AI. Maybe you could elaborate on that. $\endgroup$ Jul 27, 2023 at 13:32
  • 2
    $\begingroup$ I think that if MathOverflow takes a stand against anything, it should be people who are so sure of their own value judgments that they feel fine about co-opting large diverse organizations into endorsing them. $\endgroup$ Jul 27, 2023 at 13:46
  • 1
    $\begingroup$ @StevenLandsburg ... I guess the answer is ... no then. $\endgroup$
    – user1437
    Jul 27, 2023 at 13:50
  • $\begingroup$ Does this answer your question? At what point should MO recognise it needs to leave Stack Exchange Inc? $\endgroup$
    – Alex M.
    Jul 27, 2023 at 17:11
  • $\begingroup$ If we ban AI here, how are we supposed to ask questions and give answers on MO that will make AI safer? $\endgroup$ Jul 27, 2023 at 19:28
  • 6
    $\begingroup$ AI is coming. Much like genetic engineering, clean energy, and quantum computing. Nothing short of a global catastrophe is going to stop that. Pretending that we can prevent AI coming into our lives is tantamount to burning textile machines in England. If you have a vested interest in this being safer, be a policy maker, an advocate, and a researcher on the frontier of it all. $\endgroup$
    – Asaf Karagila Mod
    Jul 27, 2023 at 21:56
  • 1
    $\begingroup$ In view of the "ratchet effect" caused by the fact that one can upvote comments but not downvote them: my personal stance differs from Asaf's, but nevertheless my answer to Jason's question would be "no". $\endgroup$
    – Yemon Choi
    Jul 28, 2023 at 20:58
  • 1
    $\begingroup$ It is not practical for MO to get rid of AI since the search bar that we use to search for questions uses AI. It will not be practical for MO to dictate what kinds of AI are allowable and which kinds are not allowable since that will be very complicated and divisive. $\endgroup$ Jul 29, 2023 at 15:32
  • $\begingroup$ If someone wants to make sure that AI is safer, one should donate one's disposable income to AI safety research or conduct some AI safety research by one's self (most mathematicians should be able to figure out some AI safety research). This is better than imposing rules on others. $\endgroup$ Jul 30, 2023 at 12:35
  • 1
    $\begingroup$ @JosephVanName ... I wasn't trying to impose a rule on anyone. I was wondering if there was a strong consensus in the community about this issue. This post told me there was not, and so I really don't care about trying to impose anything on the mathematical "community". I am happy to go elsewhere. $\endgroup$
    – user1437
    Jul 30, 2023 at 16:07
  • $\begingroup$ @JasonPolak An attempt to ban AI from another community will be counterproductive as well. Since you are concerned about AI, you should either stay offline and off the grid, or you can work on AI safety. Right now, many people have trouble figuring out what trained neural networks are actually doing, so we need to develop and investigate more AI interpretability tools. $\endgroup$ Aug 3, 2023 at 12:17
  • 1
    $\begingroup$ @JosephVanName Thanks for your advice, I am working on AI safety by writing about it and informing people of its dangers. You may not agree with this line of action, but I believe (for a variety of reasons that wouldn't fit in this box), but I believe that course of action is the only morally correct one. Simply playing the prisoner's dilemma as you suggest I believe is not the right way to go and I will never in my life contribute one iota of understanding to AI and how they work, so that someone else can build a better AI. Suffice it to say, I disagree with everything you say on every level. $\endgroup$
    – user1437
    Aug 4, 2023 at 13:54
  • $\begingroup$ @JasonPolak To continue the discussion, we should go to the AI safety chat room that I created. chat.stackexchange.com/rooms/147565/ai-safety So you want people to know about the dangers of AI, but you want them to not understand AI? $\endgroup$ Aug 4, 2023 at 15:12

1 Answer 1

-2
$\begingroup$

From the voting for this question and from the comments to this question, we can conclude that the MO community is not in favor of Jason's proposal. MO should be a place that is welcoming to professional mathematicians, and this means that MO should not have an official stance on issues that will make MO less welcoming to professional mathematicians.

It is not practical for MO to take a stand against AI in general or even for particular uses of AI even if we want to take that stand. MO has a search bar for searching for questions and answers, and all search engines use AI in one form of another. Even the process of doing basic research when searching for an answer uses AI since one will need to interact with a search engine outside MO. One can often use gradient ascent/descent in order to search for counterexamples or figure out constants (especially in linear algebra), and one can use simulated annealing and evolutionary computation to produce counterexamples to conjectures. On this site, we are also able to ask and answer questions directly about AI, and we are able to have discussions about AI too.

When trying to mitigate the risks imposed by AI, one has to carefully consider which strategies one should employ. Taking a stand against AI is unlikely to be effective since AI will develop and progress regardless of whether we want it to or not. A stand against AI development will simply allow some groups to progress unimpeded while other groups will be stifled, and this effort will result in inequality and less AI safety. If one wants to slow down AI progress, then one should have worked on AI safety earlier, and one now needs to work harder on AI safety. While we may not be able to stop or slow down the development of AI, we can certainly steer the development of AI into a safer or more friendly direction. The MO community can do things individually and collectively to improve AI safety. The individuals here can read about AI safety so they know what others think about the risks of AI. After reading about AI safety, we may be willing and able to perform mathematical research that will improve AI safety. The community here can ask and answer questions about AI safety. The community can have a high level discussion about AI safety. But to have questions, answers, and discussions about AI safety here in the first place, AI cannot be banned from MO.

I have created an AI safety chat room here on MO in case anyone is interested.

If you are interested in AI safety, make sure to take care of yourself.

$\endgroup$

You must log in to answer this question.