Wikipedia:Reference desk/Mathematics
of the Wikipedia reference desk.
Main page: Help searching Wikipedia
How can I get my question answered?
- Select the section of the desk that best fits the general topic of your question (see the navigation column to the right).
- Post your question to only one section, providing a short header that gives the topic of your question.
- Type '~~~~' (that is, four tilde characters) at the end – this signs and dates your contribution so we know who wrote what and when.
- Don't post personal contact information – it will be removed. Any answers will be provided here.
- Please be as specific as possible, and include all relevant context – the usefulness of answers may depend on the context.
- Note:
- We don't answer (and may remove) questions that require medical diagnosis or legal advice.
- We don't answer requests for opinions, predictions or debate.
- We don't do your homework for you, though we'll help you past the stuck point.
- We don't conduct original research or provide a free source of ideas, but we'll help you find information you need.
How do I answer a question?
Main page: Wikipedia:Reference desk/Guidelines
- The best answers address the question directly, and back up facts with wikilinks and links to sources. Do not edit others' comments and do not give any medical or legal advice.
May 18
[edit]Google Deepmind and the future of math
[edit]This video talks about some of the new mathematical results from Google Deepmind. Before anyone changes their major, it seems worth pointing out that these are all examples of a certain type of optimization problem and that Deepmind only improved on the best known result, it didn't prove or even claim they were the best possible, only that they were an improvement on what human intelligence was able to find in those particular cases. So I wouldn't count on the Collatz conjecture being solved tomorrow. But it does seem to indicate that the way math is done in the future will be significantly different than the way it was done in the 20th century. Thoughts? RDBury (talk) 02:07, 18 May 2025 (UTC)
- The main takeaway of the video is that Matt Parker should take a vacation more often.
- Constructing formal proofs is not an optimization problem, but in a sense it is a combinatorial puzzle, and I expect more artificial help there in the future. Formalizing published proofs (possibly uncovering seemingly obvious but unjustifiable skips) could be a nice start. Another area where an indefatigable agent can prove its worth may be finding interesting conjectures connecting different fields, similar to the discovery of the monstrous moonshine, which started as an observation of a curious coincidence.
- I do not expect the Collatz or Goldbach conjectures to be solved ever, also not with the help of a superintelligent agent. We cannot, with the knowledge we have, exclude the possibility that although true they are not members of the theory of ZFC. (Consider that a simple generalization of the Collatz problem is undecidable, so it is (non-constructively) certain that some Collatz variants are true but don't hold a ZFC club membership card.) ‑‑Lambiam 09:32, 18 May 2025 (UTC)
- Formalizing proofs and developing proof verifiers is a program which has been underway since well before the recent AI boom.--Antendren (talk) 03:45, 20 May 2025 (UTC)
- The use of generative AI in constructing formal proofs is still largely uncharted territory. The Flyspeck project to produce a formal proof of the Kepler conjecture took seven calendar years and an order of magnitude more person years. Where are the formal proofs of Fermat's Last Theorem and the Poincaré conjecture, to mention just two? I know a project for FLT has been under way for about a year now,[1] but (to the best of my knowledge) current AI systems are not yet helpful for this task. ‑‑Lambiam 06:22, 20 May 2025 (UTC)
- Formalizing proofs and developing proof verifiers is a program which has been underway since well before the recent AI boom.--Antendren (talk) 03:45, 20 May 2025 (UTC)