Who Taught the Machines Mathematics?(nisheethvishnoi.com) |
Who Taught the Machines Mathematics?(nisheethvishnoi.com) |
> I learned mathematics because others gave their time and knowledge freely, and I wanted to do the same for the next generation. The possibility that a student would go on to do something I could not do was part of the point. But I had imagined this as a continuation of a human activity. [...] I could not imagine treating this knowledge as something to be guarded.
But soon after, AlphaZero was trained purposefully without human games in the training data. The "extractive" training was sufficient to show that a certain machine learning model worked, but unnecessary to produce a machine that performs beyond the level of humans.
For coding agents I imagine it is similar: humans introduce so many subtle bugs, and they're only subtle relative to human cognitive capacity. I'm sure subtle mistakes pass peer review in mathematics too, even if it's generally more rigorous.
What I'm getting at is that labs have demonstrated that a certain model architecture is capable of producing mathematics at a (super)human level, and their extractive methodology was sufficient, but possibly not necessary. You can fight in court whether their methodology is against the law, but it would be naive not to look beyond that, which is my main gripe with the AHM statement the author refers to [1]
Fortunately the author of this article has a more nuanced view:
> I can understand the anger, but I am not sure that withdrawing from AI is the answer
To clarify and be fair, the people who actually invented geometry, or at least worked on the majority, were never interested in building things. Or at least that wasn't their primary concern.
It's more accurate to say that mathematicians were just interested in intellectual exploration, everyone else was interested in building things, and there's an uneasy alliance between the two that is about to be shattered.
Use a wire to get a straight line? Used in construction
Use a determined length of wire to compare lengths ? Used in construction
Use a determined length of wire to draw circles? Used in construction
Use 2 lengths of wire to construct parallel lines? Used in construction
Use lengths of wire to construct a perfect 90 degree angle? Used in tool-making (ie. to make a jig that you then use to measure)
The list goes on and on and on. If you've never used CAD packages, you should. The amount of geometry that goes into part design in everything from construction to "tools and construction engineering" as it's called in my country is incredible. Incredible as in they reserve 8 months in university to make sure people who already would have had pretty difficult math in high school are up to date on advanced geometry (plus teaching them how to properly extend what they know into 3d)
There's been updates since Greek times, such as using lasers for drawing straight lines instead of wires, but not as many as you'd think.
This is great because as anyone will point out, doing things this way (with correct equipment) results in accuracy across large distances (up to dozens of kilometers) that you just cannot match with analytical geometry + measurements.
Geometry was obviously invented and fine-tuned for thousands of years, for construction. That's what it was used for 4000 years ago ... and that's close to the only reason to still use (non-analytical) geometry today.
For instance, imagine a mathematical institute that hires reasonably competent mathematicians who are amazing AI prompters but lack mathematical creativity or "genius". All they produce are mathematical papers with clear proofs and explanations. Very human readable stuff. But the catch is that they are not discovering any of it. All they are doing is "translating" full AI discoveries into a format others can understand.
Is that acceptable?
FWIW, it reads very similar to the woes of the SWEs previously, in the sense that it kinda appears like people assigned value to the wrong part of the work within the discipline?
The take-away of it being it being imperative to have open weights models accessible to all, instead of cloud-hosted models owned by few is of course true regardless of those specifics, but it gets a bit muddied by that, which is a shame, as that will be weaponized to dismiss it.
My local community is in a mixture of denial and despair. Especially for the PhD students and postdocs, and faculty in terms of hiring (how do we hire?).
Given that this appears to be a largely AI-written article, I guess it's AI being shaken by AI.
I say this as someone who uses LLMs in the physical sciences (one of the only areas where I would say it is ethical to use them).
We still need math research just like we still coding or software or cpu research regardless of how many problems the AI solves.
To me it feels like a bio researcher complaining that his field has no future because AI started to cure diseases so now instead to spend 100 years to run experiments someone can do it in a day with AI(obviously it’s not the case yet).
Are they guys upset that people can “run math” without hiring a math guy?
The central problem is that the majority of people go into math because they love discovering mathematics and want to participate in the process. That itself is a huge part of the reward. That creates the community. Simply have huge numbers of people outputting math as if it were an industrial factory kills the joy in the process and so in itself destroys the community.
So in short we're seeing a dichotomy between the "problem miners" who are simply hooked on producing results and mining them and "explorers" for lack of a better term.
There's one thing you've got to keep in mind - mathematics, in reality, is mostly useless in the sense that most discoveries these days will always be useless. What is useful generally is teaching, and applying mathematical abstraction to more realistic problems, a skill that you mainly get from dicovering generally useless stuff. (Math for art's sake.) That's how math works, or at least worked before AI. Once you put AI in the mix, it messes up all the incentives....
Personally, I'm 100% anti-AI because I think there's not much point in simply discovering new things if it's not primarily a human journey. It's revolting to have solutions generated by machines like that.
To be clear, I empathize with your sentiment and revolt, despite being more of a "builder" myself in the software world. I just can't see how the explorer mindset can survive the onslaught of AI-augmented output.
At one point, the metrics were fairly aligned with the things mathematicians love, but nobody was ever judged by how much they like their work, right? The theory is that the more you like your work, the better your output will be. But motivations are multifaceted and I could see people who value careers over satisfaction winning if AI helps them.
I'm not saying I like this, just that it seems to be the most likely outcome.
Interesting, thanks for the insight. Very similar to the 'builders' vs 'coders' dichotomy that seems to crop up amongst programmers now.