> Before injecting this cocktail into your bloodstream, would
you want to know that there is at least one human cancer
expert who understands the mechanism behind this cure?
No. I'm gonna die, my man.
This guy might have the highest IQ on the planet, but it's clear he hasn't spent much time around average people.
This argument is so silly against reality, and already, almost nobody understands the things they put in their body to any significant degree, beyond the effect produced.
Will I take a cancer cure that no human understands? Yes. And so will billions of others. Just needs to cure cancer, that's nearly the only requirement.
Slide 12 and 13 sum up my conclusions from the Oct 8 100+ reactions to 100+ solutions. Several reported OpenAI drop included answers put a wrap on problems they had been working for years. Several said that they have to completely rewrite grant requests that they had just submitted. Others described learning of the solutions problems they had been working on like losing an old friend or a lover. The general sentiment was to bemoan the loss of a field, as if it would have been better be born in the early 20th century and conclude their career arc before reaching this point. My take away is that the field needs to get it through their heads that their old problems are no longer ambitious, and that their job in the short term is to find the new frontier.
One idea Terrance Tao conjectures, which is highly doubtful, is that spamming the AI button will solve open problems without producing insightful new methods. But the OpenAI drop would seem to disprove this. The sub O(nlogn) proof for DFT for example violated very old human assumptions. Decades of work in the field was incremental progress on sub optimal method that nobody questioned hard enough. More generally, we should always be able to go back to a super-human AI and say, "Attack this problem, but don't use a method tried before."
> Using AI to find and highlight new principles, methods, or
insights, rather than merely new proofs.
I would have assumed that, by producing new proofs, the AI has either validated existing principles, or discovered new ones? Isn't that worth studying?
Are mathematicians complaining that reviewing AI's proofs is not as fun as writing your own? Try being a programmer... welcome to our world!
If AI lacks imagination and is not discovering new principles, then it's doing us a favour: it's crossing out the problems that don't need new principles. So the problems/conjectures that are still left are the more interesting ones.
I just watched Primeagen’s video on this and Tao’s point is that juniors no longer have the path of solving a proof to earn Field’s medals. He also argues that the community part is being hurt by AI discovering proofs because in the past people used to get invite to talk and collaborate. Now all that is being taken away. The community must adapt because Pandora’s box cannot be closed.
Prime also mentioned that software development is different. In Software development, the product is what you’re building towards, so the means to get there can be disrupted without the industry being cannibalized.
In math research, the process is the product. You take away the researching part and not much is left. But my question is, these math proofs OpenAI released, will math shift to actually using the proofs to change the world instead of just finding new ones?
> Before injecting this cocktail into your bloodstream, would you want to know that there is at least one human cancer researcher who understands the mechanism behind this cure.
Tao is out of his lane; lots of medicines don't have understood mechanisms. We don't even have the mechanisms behind general anesthesia nailed down; do you want to forgo it when the docs cut you open to remove your tumors?
The cancer arguement on slide 20 is pretty weak. I first need to be alive in the long term to worry about the long term effects. If I had terminal cancer I'd gladly take an AI developed 'cure'.
There are still a lot of treatments/medicines in medical science where we dont know 100% the real reason as to why it does what it does but we still prescribe them because the intended effect is what we are interested in.
I'm currently on two fairly common medicines that have, in the first paragraph when reading about them, "doctors are unsure of the specific action, but it is thought that [medicine does x to y]"
If I'm terminal with cancer, I'll inject whatever if it can cure that.
It is nice to see some sanity back in the conversation!
I'm still not sure about math-2.0 (humans+AI will make fundamentally more progress):
- AI and computer usage take a mental toll on humans and humans will overlook radical improvements.
- AI may be good at finding useless things like "P==NP, but the complexity is O(n**4242424242424242)". In other words, useless.
- Humans become formalists and lose traditional sources of inspiration. Maybe interacting with Lean should be left to specialists, but not to creative blackboard mathematicians.
- AI exposure will further intellectual conformity, more than the Internet did.
As to the last point, a lot of progress (real, not measured in publications) seems to have been made when communication was slower and there were several different schools and approaches.
> Before injecting this cocktail into your bloodstream, would you want to know that there is at least one human cancer expert who understands the mechanism behind this cure?
> Or a human mathematician who understands the mathematical model used to locate the cocktail?
If we're being fair to AI - it contains collective knowledge from all fields, which means it's probably less likely to miss something that a human would.
page 20 ("A thought experiment on alignment and understanding") is perhaps not likely to quite induce the reaction in most people that Mr. Tao expected.
Has the world gone mad? In which universe would you get a mathematical model to produce a tonic of random chemicals that optimize said mathematical model, put that through stage 3 testing, and only THEN ask if you want to inject this into yourself. My Guy, you just did clinical testing on fucking humans for phase 3, it's a little late to consider the moral ramifications of the analysis-experiment dichotomy.
The real example is that you ask ChatGPT for a novel cancer cure, and it spit out some random chemicals, probably including bleach. Do you inject that into cancer patients that have no other hope? No, you don't. You absolute charlatan.
No. I'm gonna die, my man.
This guy might have the highest IQ on the planet, but it's clear he hasn't spent much time around average people.
This argument is so silly against reality, and already, almost nobody understands the things they put in their body to any significant degree, beyond the effect produced.
Will I take a cancer cure that no human understands? Yes. And so will billions of others. Just needs to cure cancer, that's nearly the only requirement.
Tao called it a cure, see the thing I quoted.
The most common Science™ failure mode.
But since when do we call for a stop to figuring out new things? We don't want a cancer cure unless we understand it?
N people should die because...math is cool for humans? Or?
One idea Terrance Tao conjectures, which is highly doubtful, is that spamming the AI button will solve open problems without producing insightful new methods. But the OpenAI drop would seem to disprove this. The sub O(nlogn) proof for DFT for example violated very old human assumptions. Decades of work in the field was incremental progress on sub optimal method that nobody questioned hard enough. More generally, we should always be able to go back to a super-human AI and say, "Attack this problem, but don't use a method tried before."
I would have assumed that, by producing new proofs, the AI has either validated existing principles, or discovered new ones? Isn't that worth studying?
Are mathematicians complaining that reviewing AI's proofs is not as fun as writing your own? Try being a programmer... welcome to our world!
If AI lacks imagination and is not discovering new principles, then it's doing us a favour: it's crossing out the problems that don't need new principles. So the problems/conjectures that are still left are the more interesting ones.
Prime also mentioned that software development is different. In Software development, the product is what you’re building towards, so the means to get there can be disrupted without the industry being cannibalized.
In math research, the process is the product. You take away the researching part and not much is left. But my question is, these math proofs OpenAI released, will math shift to actually using the proofs to change the world instead of just finding new ones?
Tao is out of his lane; lots of medicines don't have understood mechanisms. We don't even have the mechanisms behind general anesthesia nailed down; do you want to forgo it when the docs cut you open to remove your tumors?
There are still a lot of treatments/medicines in medical science where we dont know 100% the real reason as to why it does what it does but we still prescribe them because the intended effect is what we are interested in.
If I'm terminal with cancer, I'll inject whatever if it can cure that.
I'm still not sure about math-2.0 (humans+AI will make fundamentally more progress):
- AI and computer usage take a mental toll on humans and humans will overlook radical improvements.
- AI may be good at finding useless things like "P==NP, but the complexity is O(n**4242424242424242)". In other words, useless.
- Humans become formalists and lose traditional sources of inspiration. Maybe interacting with Lean should be left to specialists, but not to creative blackboard mathematicians.
- AI exposure will further intellectual conformity, more than the Internet did.
As to the last point, a lot of progress (real, not measured in publications) seems to have been made when communication was slower and there were several different schools and approaches.
> Or a human mathematician who understands the mathematical model used to locate the cocktail?
If we're being fair to AI - it contains collective knowledge from all fields, which means it's probably less likely to miss something that a human would.
The real example is that you ask ChatGPT for a novel cancer cure, and it spit out some random chemicals, probably including bleach. Do you inject that into cancer patients that have no other hope? No, you don't. You absolute charlatan.
It turns out research math was puzzling solving and we rewarded idiot savants.