[00:00] Artificial intelligence has eaten maths  and now it's coming for physics. And the   interesting bit is no one's talking about it.  So, let's talk about it. In maths, things have   been happening so fast I've given up trying to  keep you updated. Since we last talked about it,   [00:17] Claude Fable 5 helped find a counter example to  the Jacobian conjecture, which I never heard of,   but can now tell you that it had been open since  1939. Then OpenAI announced 10 results from its   [00:31] still unreleased Astra model, each of which either  solved or advanced a long-standing open problem.   At least that's what they say. Another group used  Chad GPT to disprove a 30-year-old conjecture that   [00:43] no one had heard of previously either. And earlier  this month, Anthropic announced that Claude had   made a major advance on the Riemann Zeta function.  So working towards the Riemann hypothesis,   [00:55] finally a topic that even I have heard of. I  think at this point most mathematicians have   accepted that large language models are now  capable of serious math. It seems likely to   me that mathematics will be entirely automated in  the near future and especially young people just   [01:12] entering the field are caught between depression  and anger. Graduate student Tasmin Chu thinks that   mathematics is in an existential crisis and  calls for a moratorium on using LLM to prove   [01:26] novel mathematical theorems. Assistant professor  Max Weinreich has written a preprint in which he   makes the case for total opposition to the use of  artificial intelligence in mathematics. Postdoc   [01:39] Alexis Marchand argues that LLMs are evil because  they contribute to the climate crisis and PhD   student Kirwin Hampshire thinks he's going insane  and is suffering a profound spiritual crisis   [01:52] due to these developments. Mathematicians have  independently derived the same existential crisis   by four different methods. Assistant Professor  Daniel Litt, meanwhile, has given a more or less   [02:05] serious talk about the end of mathematics in which  he belabours that AI will lead to the production   of mathematically correct results that no one  will read, no one will care about, and no one will   [02:17] benefit from academic business as usual. But what  about physics? If AI is now good enough to lead   maths proofs, then it should also be good enough  to do physics calculations. In fact, it is. I know   [02:29] this because I've tried. I also know that a lot  of theoretical physicists are now doing the same   thing, taking research ideas that they haven't  had time to follow through and put them into AI   [02:41] to see what comes out. And sometimes the result  is verifiably good enough for publication. Peter   Woit has looked at the number of submissions to  the arXiv preprint server in the category hep-th,   [02:54] that's high energy physics theory, and finds that  it's up about 30% in May and June compared to   previous years. One of the recent examples comes  from quantum physics. Physicists in this area have   [03:08] long wondered if there's an easier way to create  some complicated entangled states. Gerardo Adesso   at the University of Nottingham has now proved  that the answer is no. And he did so with the   [03:20] help of ChatGPT 5.6 about which he says the AI  made a central contribution to developing the   proof strategy. He then checked every step of the  final proof himself. Another recent example goes   [03:33] all the way back to James Clerk Maxwell. Imagine  a collection of electric charges in random places.   There will be some places where their electric  forces exactly cancel kind of like Lagrange points   [03:46] in gravity. Maxwell proposed a rule for how many  such places you could get from a given number of   charges. Three researchers have now shown that  Maxwell's rule is wrong and their paper says   the idea behind this construction was suggested  again by Chad GPT 5.6. Those two examples are   [04:04] still very mathematical. So here is one involving  an experiment. These two researchers were studying   how tiny silicon membranes behave if you shine on  them with a laser. For this they did measurements   [04:16] and developed a model about which they write  that it was derived with the help of Claude   Fable 5. The model has excellent agreement with  the observations. I could go on like this for   [04:28] a while. There's a paper in high energy physics  that use several large language models to carry   out the symbolic and high precision numerical  computations. Another paper in quantum physics   used claude for numerical calculations. Another  team had Claude write software for moving   [04:45] ions around inside a quantum computer and in  fusion research Chad GPT agent was put in charge   of deciding which stellarator designs to try next.  Physicists have clearly picked up on the LLM maths   [05:00] revolution, but it's happening in a completely  different way. They just organically integrate   the models as tools and move on. I also haven't  yet seen the same desperation as I have seen   [05:12] among young mathematicians. Mathematicians  see artificial intelligence and ask whether   mathematics will survive. Physicists ask whether  it can finish the calculation by Friday. A few   [05:25] years ago, I received some ugly letters from a  debt collection company. They threatened to sue   me for ordering something online and then not  paying it. Turned out that someone had stolen   [05:40] my private data. I managed to get out of this,  but since then I've become very protective of my   personal information, which is why I've signed up  to Incogn who've been sponsoring this video. Each   [05:52] time you open a website, it'll try to collect data  about who you are and where you are and what other   websites you've visited. Many firms then make  money by selling your private information to data   [06:05] brokers. Most countries have laws against that and  you can ask for your data to be removed. But doing   this takes up a lot of time. Incogn automates the  process of getting you out of those databases.   [06:18] You sign up and they'll contact the big sinners  request that your personal details be removed.   They'll keep on doing that and if you want send  you updates about the progress they're making.   [06:30] They now also have a custom removal feature  where you can submit specific websites. I'll   be using this a lot. And Cognney has really made  my life better. If that sounds like something you   [06:44] could need to use my link, Zabina or the  custom link in the info because the first   100 people to use it will get 60% off. Stay  safe. Thanks for watching. See you tomorrow.