Oct 10, 2025 05:20 PM
Commonplace assurances to the intelligentsia and white collar workers a few years back: "Don't worry, you're safe. Smart machines will only take over the jobs of the proles."
- - - - - - - - - - - - - - -
SAABINE HOSSENFELDER
https://youtu.be/EzAIRE1MIrk
VIDEO INTRO: Mathematics, as a field of research, is being hard hit by artificial intelligence. This is only the beginning, but it’s foreseeable that this will get much worse. Much like calculators, the devices have replaced human calculators, theorem proving machines will soon replace much of mathematics. Not all of it, but much of it. That’s what I think in any case.
Mathematicians are currently somewhat in denial about this. Let’s have a look.
Earlier this year, both Google DeepMind and OpenAI reported gold-medal level performance on this year’s mathematics Olympiad problems. What’s noteworthy here is not so much that they outperformed most human’s maths skills, but that they did it using general-purpose reasoning models, not specially trained ones. This took most people by surprise. But mathematicians are not impressed, as Scientific American reports. These questions “don’t compare to the kinds of questions professional mathematicians try to answer” wrote Emily Rhiel, a mathematician at John Hopkins university. Fields medalist Terence Tao argues that the comparison is unfair because AI has a speed advantage and generates possible proofs more like an entire group of people.
Yes so what? It’s also unfair to compare a human to a calculator. That hasn’t stopped tech from taking over the profession...
https://youtu.be/EzAIRE1MIrk
https://www.youtube-nocookie.com/embed/EzAIRE1MIrk
- - - - - - - - - - - - - - -
SAABINE HOSSENFELDER
https://youtu.be/EzAIRE1MIrk
VIDEO INTRO: Mathematics, as a field of research, is being hard hit by artificial intelligence. This is only the beginning, but it’s foreseeable that this will get much worse. Much like calculators, the devices have replaced human calculators, theorem proving machines will soon replace much of mathematics. Not all of it, but much of it. That’s what I think in any case.
Mathematicians are currently somewhat in denial about this. Let’s have a look.
Earlier this year, both Google DeepMind and OpenAI reported gold-medal level performance on this year’s mathematics Olympiad problems. What’s noteworthy here is not so much that they outperformed most human’s maths skills, but that they did it using general-purpose reasoning models, not specially trained ones. This took most people by surprise. But mathematicians are not impressed, as Scientific American reports. These questions “don’t compare to the kinds of questions professional mathematicians try to answer” wrote Emily Rhiel, a mathematician at John Hopkins university. Fields medalist Terence Tao argues that the comparison is unfair because AI has a speed advantage and generates possible proofs more like an entire group of people.
Yes so what? It’s also unfair to compare a human to a calculator. That hasn’t stopped tech from taking over the profession...
https://youtu.be/EzAIRE1MIrk