4 ms·
[flagged]
by ldjkfkdsjnv 1y ago
[flagged]
- zeroonetwothree 1y agoHow did transformers move forward pure math?
- ldjkfkdsjnv 1y agothere is a straight-faced argument that transformers are already a mathematical research tool and even a source of new theorems. and if true, surely more revolutionary than some ultra complicated corner of math
- dullcrisp 1y agoBig if true.
- ldjkfkdsjnv 1y agovery big
- markgall 1y agoDoes there exist a person who would make this argument straight-faced? I am a professional mathematician and have yet to hear of anyone coaxing an even slightly interesting new theorem out of AI. I think the day is clearly coming but it's not here.
- ldjkfkdsjnv 1y agofair enough, i suppose im a believer that the seeds are planted, the day is soon. and i must say, it seems more worhtwhile trying to figure out how to finetune an llm/implement reinforcement learning that could do some form of pure math, than it is to try and do new pure math by hand
- cycrutchfield 1y agohttps://en.wikipedia.org/wiki/Dunning%E2%80%93Kruger_effect https://en.wikipedia.org/wiki/Dunning%E2%80%93Kruger_effect
- ldjkfkdsjnv 1y agoonly low iq people talk about the dunning kruger effect
- cycrutchfield 1y agoOnly low iq people talk confidently about fields of study they barely comprehend.
- wizzwizz4 1y agoI do research in this field. LLMs can be used as (ridiculously inefficient) implementations of some search algorithms that we haven't yet identified and implemented in software ourselves, but which can be inferred from a statistical analysis of the literature. Sometimes those search algorithms generalise to new areas, but more often than not, they flail. The primary advantage of a language model is that it's one big algorithm: when one subcomponent would flail, another (more "confident") subcomponent becomes dominant; but that doesn't solve the case where none of the subcomponents are competent and confident in equal measure. In short: to the extent it's useful, it's a research dead-end. Any potential improvements we understand are better implemented as actual search algorithms. You've probably seen that thing where ChatGPT cracked Enigma[0]. It used several orders of magnitude more computational power than a Bombe (even given Moore's Law, still thousands of times more electrical power), and still took two dozen times longer. You would literally be better off doing brute-force search with a German dictionary. Thus is it with mathematics: a brute-force search is usually cheaper and better than trying to use a language model. Terry Tao is one of the staunchest knowledgeable advocates of GPT models in mathematical research, and afaik he doesn't even bother trying to use the models for proof search. It's like trying to build a house with a box of shoes: sure, the shoe is technically more versatile because you can't use a hammer for tightening bolts (the shoe's sole has enough friction to do this) or foot protection (the shoe is the right shape for this) or electrical isolation (the bottom surface of the shoe is largely rubber), but please just use a hammer if you want to manipulate nails. [0]: https://www.techradar.com/news/we-watched-an-ai-crack-the-enigma-code-in-just-over-ten-minutes https://www.techradar.com/news/we-watched-an-ai-crack-the-en... – and I know that's not the original ChatGPT®, but I am not rewarding this company for such a wasteful and pointless publicity stunt.
- cma 1y agoShould Berners-Lee get a fields medal if the web helped math? How about a farmer that is proven to have fed the most mathemeticians?
- moralestapia 1y agoYou're right/wrong. It goes for big breakthroughs, but not all of them, depends mainly on "networking", like Nobels.
- sno129 1y agoMy academic work was very close to June Huh's; my Ph.D. thesis was directly inspired by his Fields Medal winning work. His accomplishments have undoubtedly moved the field forward and connected various ostensibly disparate areas of math, not to mention he is one of the clearest writers and speakers in all of mathematics. There are very few people in pure math that care about transformers; they have had practically zero impact on the sort of research mathematics that the Fields Medal is concerned with.
- dang 1y ago"Please don't post shallow dismissals, especially of other people's work. A good critical comment teaches us something." "Don't be curmudgeonly." https://news.ycombinator.com/newsguidelines.html https://news.ycombinator.com/newsguidelines.html
- acheron 1y agoIs "Don't be curmudgeonly" a new addition? I don't remember that one before.
- ldjkfkdsjnv 1y agoDang, with all due respect, this comment was really not that bad. You are too heavy handed
- defrost 1y agoAs a late observer, equally respectfully, it very likely wasn't dang's hand that downvoted and or flagged your comment. You're responding to an after action review comment by moderator as to what might have caused the flag.
- dang 1y agoI know it sucks to get scolded by a mod, and it's not my wish to make anyone feel bad.