Чатбот знает физику получше многих людей (вынос коммента из-под замка в Фейсбуке)

[posic]

All people are supposed to do something at which they are better than computers. In particular, the not-so-smart people are not supposed to do physics. There are lots of things to apply one's hands and mind at instead. I, for one, do not fear being surpassed by chatbots at whatever I am doing in mathematics. I know full well that what I am doing is original, nontrivial and cannot be extracted from the previously existing corps of literature by a computer algorithm. Other original thinkers in whatever areas of knowledge should not fear that, either.

Comments

[posic]

Who surpasses whom depends on the judgement of an observer. If today's average peer reviewer of prestigious math. research journals is to be taken as the observer, then I wouldn't reject out of hand the possibility that GPT-4 has already surpassed Einstein and Scholze. I don't care about this kind of clueless observers. I am quite confident that, in the eyes of a genuinely qualified observer -- not to be confused with an observer the Editorial Board of Inventiones Mathematicae considers to be qualified -- I will never be surpassed by a computer. But in any event I do not see why I should worry about this (nonexistent) eventuality. I have only one life and want to do something nobody would have done without me (ever, or at least anytime soon). There is no point in being worse than a computer. If computers can do better at math. than I can (though they can't), the only reasonable conclusion is that I should be doing something else.

[posic]

I don't believe that a chatbot can possibly prove the Riemann Hypothesis. For me, that's more of a matter of my (high) opinion about the Riemann hypothesis than my (low) opinion about artificial intelligence. Of course, no proof of the Riemann hypothesis, whether done by a human (highly unlikely in my lifetime, I believe) or by a chatbot (all but impossible) would automatically make my books and papers useless or unnecessary. In this sense, there is no acute direct competition between me and any provers of the Riemann hypothesis, computer-assisted or not. Though there may be some degree of indirect competition, as I am still entertaining the idea that a century after my death someone might prove the Riemann hypothesis using some future concepts that will have grown out from the definitions I have been working on.

[posic]

Register my vote against regulating the industry. I am a libertarian. I can full well compete as a mathematician against mathematicians or software developers armed with GPT-999, no problems with that.

[posic]

In my view, human-played chess became a much less meaningful activity after computers started playing chess considerably better than humans. I would argue that it became much less interesting for people to play or observe people playing chess on the top competitive level after that point. Yes, many people always played chess as amateurs and can continue doing so now, there is nothing wrong with that. But there used to be a huge difference between Kasparov playing Karpov and me playing my brother. I would argue that now there is much less of such a difference. All human-played chess is now amateurish as compared to what computers do. This is reflected in the fact that preventing computer-assisted cheating became crucial for all human-played competitive chess.

Likewise, there is nothing wrong with amateur mathematics per se. Occasionally an amateur mathematician may prove a theorem that is more interesting than some professional work. But what I am doing is not amateur mathematics. What I am doing in mathematics is supposed to be really important. It is supposed to have lasting positive impact in the world. That is what I expect of myself. Otherwise I would feel bad about myself doing it. People armed with computers reducing the significance of what I am doing in mathematics to the generally amateurish level would be a huge blow to my vision of myself.

Imagine that a computer can write a research paper or monograph on a given topic in homological algebra which in almost all respects surpasses mine (and for a fraction of time, too). Imagine that I am reduced to defending my work by saying that it was not computer-assisted -- the way a present-day chess player might defend his suboptimal moves by saying that his play was not computer-assisted and should be only compared with other humans' play. A book produced by computer expresses a better approach and deeper understanding with clearer vision, stronger results and greater potential for applications than my book does, rendering my book essentially useless given that the computer's book is available. But I defend my work by arguing that it should be only compared with what other people can do without using chatbots. I would see no point for me to continue writing my books and papers under such circumstances. But I do not believe for a second that it can happen.

[posic]

https://posic.livejournal.com/2775103.html