Submission statement: This release contains solutions to hundreds of math problems, including a partial solution to the Riemann hypothesis (no zeros for Re(s) > 7/8) and two other Millenium Prize problems. Also some major results In physics, computer science and machine learning. The average result reportedly took about three hours of computing. This shows that frontier AI is now vastly superhuman at solving open math problems. This might soon carry over to other scientific fields. If so, it could have a significant impact on economic growth and productivity.
MaximumEffort433 on
Yes, but can it tell us how to reverse entropy?
shumpitostick on
This is really insane stuff and it’s hard to overstate it. It’s like a decade’s progress in math just happened in a month. There’s so many important problems there.
sambar_samurai on
Many in math sub are in full blown crisis. Looks like this might decimate lot of phd thesis work
xilcilus on
I’m sympathetic to the mathematicians concerns around essentially losing motivations to solve difficult problems – what’s the point if a frontier lab can spend millions of dollars to accelerate the findings. Furthermore, when humans stop to try to look for novel solutions, collectively, we are going to lose out on innovation – I don’t believe that any frontier labs have solved the synthetic collapse problems.
That being said, I think people should recognize that there are certain domains where the frontier models are now excelling in – OpenAI likely didn’t cheat to solve the Navier-Stokes equation – and OpenAI (and other frontier labs) should focus on other areas of research where practical applications can benefit the general populace.
Previous_Platform718 on
>Mathematicians are struggling with whether the accelerated pace of solving math problems helps or hurts their field if understanding the solutions lags behind.
>“I think at the heart of this issue is that humans have two competing natures: a tendency to compete and a capacity to appreciate beauty,” said Kai Shaikh, a graduate student in mathematics at the University of Toronto. “To me this seems to be a case of the former attempting to strangle the latter.”
This is a good sentiment and one that should be appreciated more. I don’t know if we actually *want* AI to be heading where it is.
Effort and rarity are two sources of value that existed long before the concept of money, and AI is poised to eliminate both of those across many fields. The phrase “*I can appreciate the effort”* is, seemingly, about to be rendered invalid.
It really does intend to suck the soul out of everything. A lot of what I like, in terms of art, is meaningful because someone’s life experiences and real emotions went into understanding and conveying something powerful in a beautiful way. I kind of don’t want to imagine a future where someone says, “this is my favorite piece of art/song/equation. It was generated in 0.2 seconds, I have no idea how it was made, and there is literally no connection to anyone alive or dead.”
klayona on
This thread will probably devolve into the usual AI slop discussion, but this might well be a funeral for the old math as we knew it. If it actually took 3 hours on average for one of these results on a $200 Pro plan, anyone could churn out dozens PhD-level results in a week. And several of these are Fields medal worthy (Quasi-Riemann is earth-shaking)
Gone are the days of solving a math problem and your name living for eternity, it belongs to the machines now.
xhytdr on
I don’t understand how people can look at this and the METR report and not realize just how fundamentally transformational this technology is going to be
8 Comments
Submission statement: This release contains solutions to hundreds of math problems, including a partial solution to the Riemann hypothesis (no zeros for Re(s) > 7/8) and two other Millenium Prize problems. Also some major results In physics, computer science and machine learning. The average result reportedly took about three hours of computing. This shows that frontier AI is now vastly superhuman at solving open math problems. This might soon carry over to other scientific fields. If so, it could have a significant impact on economic growth and productivity.
Yes, but can it tell us how to reverse entropy?
This is really insane stuff and it’s hard to overstate it. It’s like a decade’s progress in math just happened in a month. There’s so many important problems there.
Many in math sub are in full blown crisis. Looks like this might decimate lot of phd thesis work
I’m sympathetic to the mathematicians concerns around essentially losing motivations to solve difficult problems – what’s the point if a frontier lab can spend millions of dollars to accelerate the findings. Furthermore, when humans stop to try to look for novel solutions, collectively, we are going to lose out on innovation – I don’t believe that any frontier labs have solved the synthetic collapse problems.
That being said, I think people should recognize that there are certain domains where the frontier models are now excelling in – OpenAI likely didn’t cheat to solve the Navier-Stokes equation – and OpenAI (and other frontier labs) should focus on other areas of research where practical applications can benefit the general populace.
>Mathematicians are struggling with whether the accelerated pace of solving math problems helps or hurts their field if understanding the solutions lags behind.
>“I think at the heart of this issue is that humans have two competing natures: a tendency to compete and a capacity to appreciate beauty,” said Kai Shaikh, a graduate student in mathematics at the University of Toronto. “To me this seems to be a case of the former attempting to strangle the latter.”
This is a good sentiment and one that should be appreciated more. I don’t know if we actually *want* AI to be heading where it is.
Effort and rarity are two sources of value that existed long before the concept of money, and AI is poised to eliminate both of those across many fields. The phrase “*I can appreciate the effort”* is, seemingly, about to be rendered invalid.
It really does intend to suck the soul out of everything. A lot of what I like, in terms of art, is meaningful because someone’s life experiences and real emotions went into understanding and conveying something powerful in a beautiful way. I kind of don’t want to imagine a future where someone says, “this is my favorite piece of art/song/equation. It was generated in 0.2 seconds, I have no idea how it was made, and there is literally no connection to anyone alive or dead.”
This thread will probably devolve into the usual AI slop discussion, but this might well be a funeral for the old math as we knew it. If it actually took 3 hours on average for one of these results on a $200 Pro plan, anyone could churn out dozens PhD-level results in a week. And several of these are Fields medal worthy (Quasi-Riemann is earth-shaking)
Gone are the days of solving a math problem and your name living for eternity, it belongs to the machines now.
I don’t understand how people can look at this and the METR report and not realize just how fundamentally transformational this technology is going to be