In less than 4 years we will see Fields Medal quality work from completely autonomous models. Quanta is just trying to help prepare everyone for the reality that at each cognitive task they will have a brilliant AI assistant.
For outsiders, there is usually a lot more nuance existing in the fields of study than people an AI hypsters give credit for. Even in fields where everything *seems* cut-and-dry (like you would think "novelty of materials" is an easy metric RIGHT?...) . The reality tends to more nuanced and subtle. That is why experts exist. So even if ChatGPT-8 can do "field's medal level of work" as claimed by Sam Altman (or the new head of product) that may not mean anything ... Because that is not a metric that can be reduced to a single number or benchmark ... just like most things cannot be reduced to easy benchmarks that can be trained on ... reality doesn't live on your computer.
For perspective, here is a sentence from the description of the Leibniz Prize, established by the (Edward) Fredkin Foundation in the 1980s, "to be awarded "for the proof of a 'substantial' theorem in which the computer played a major role," and administered by the American Mathematical Society, according to the description at https://www.ams.org/prizes-awards/paview.cgi?parent_id=5
"The quality of the results should not only make the paper a natural candidate for publication in one of the better mathematical journals, but a candidate for one of the established AMS prizes (e.g., Cole, Veblen) or even a Fields Medal."
The prize was never awarded and "support… has been withdrawn." For additional perspective, the original challenge of this kind was published in the 1958 article by Newell and Simon entitled "Heuristic Problem Solving: The Next Advance in Operations Research" and accessible at https://www.jstor.org/stable/167397. There the prediction was that within 10 years
"a digital computer will discover and prove an important new mathematical theorem."
By "completely autonomous" do you mean that it came up with the questions on its own? Or that it solved someone else's "Fields Medal quality" question? If the latter, then I'm doing my part to hasten your scenario with my Benchmark questions, numbers 3-10 of which will be appearing here soon. Question 10, and arguably also question 9, would qualify the humanoid that solves it for a Fields Medal, in my opinion.
ps: I see you quote David Nobel saying “the history of modern technology in America is of a piece with that of the rise of corporate capitalism.” He was a good friend back when he worked at Pomona. I suppose you know he got fired from MIT and effectively banned from US universities for pushing this bolshie line. Thats why he ended up in Canada - no US university would employ him. I assume you have tenure MH so can't be threatened with that fate.
The representation of AI as something that just "happens spontaneously" reminds me of how medieval Europeans believed that snakes and worms and insects generated spontaneously out of the mud. What is the "mud" from which AI emerges??
This is getting deeper all the time. I was surprised yesterday to get an email from Zentralblatt Math (for whom I review on occasion), that they are getting a lot of inappropriate AI-generated reviews. Why would a reviewer do that?
From the grave, Simons continues to influence mathematics the way he wanted, embracing private, closed, elitist things (many of these LLM experiments are prohibitevly expensive for everyone but staff of corporate labs) over public, open, non-profit, etc. Think of the way they severely affected computer algebra research in US by funding Magma licences for all US campuses about 10 years ago.
Whether or not Quanta's editors are communing with Simons with the help of mediums or Ouija boards, I think they can take responsibility for their own editorial decisions.
I'd be interested to hear more about the effect of Magma licenses on computer algebra research, since I wasn't personally affected.
In less than 4 years we will see Fields Medal quality work from completely autonomous models. Quanta is just trying to help prepare everyone for the reality that at each cognitive task they will have a brilliant AI assistant.
Also, I am going to just leave this here:
https://thebsdetector.substack.com/p/ai-materials-and-fraud-oh-my
For outsiders, there is usually a lot more nuance existing in the fields of study than people an AI hypsters give credit for. Even in fields where everything *seems* cut-and-dry (like you would think "novelty of materials" is an easy metric RIGHT?...) . The reality tends to more nuanced and subtle. That is why experts exist. So even if ChatGPT-8 can do "field's medal level of work" as claimed by Sam Altman (or the new head of product) that may not mean anything ... Because that is not a metric that can be reduced to a single number or benchmark ... just like most things cannot be reduced to easy benchmarks that can be trained on ... reality doesn't live on your computer.
For perspective, here is a sentence from the description of the Leibniz Prize, established by the (Edward) Fredkin Foundation in the 1980s, "to be awarded "for the proof of a 'substantial' theorem in which the computer played a major role," and administered by the American Mathematical Society, according to the description at https://www.ams.org/prizes-awards/paview.cgi?parent_id=5
"The quality of the results should not only make the paper a natural candidate for publication in one of the better mathematical journals, but a candidate for one of the established AMS prizes (e.g., Cole, Veblen) or even a Fields Medal."
The prize was never awarded and "support… has been withdrawn." For additional perspective, the original challenge of this kind was published in the 1958 article by Newell and Simon entitled "Heuristic Problem Solving: The Next Advance in Operations Research" and accessible at https://www.jstor.org/stable/167397. There the prediction was that within 10 years
"a digital computer will discover and prove an important new mathematical theorem."
1968 has come and gone and we are still waiting.
Don't worry just 5 more years! Maye 2...
By misrepresenting the authors they cite? I guess their fervent belief in the robot-god justifies any other ethical concerns.
By "completely autonomous" do you mean that it came up with the questions on its own? Or that it solved someone else's "Fields Medal quality" question? If the latter, then I'm doing my part to hasten your scenario with my Benchmark questions, numbers 3-10 of which will be appearing here soon. Question 10, and arguably also question 9, would qualify the humanoid that solves it for a Fields Medal, in my opinion.
ps: I see you quote David Nobel saying “the history of modern technology in America is of a piece with that of the rise of corporate capitalism.” He was a good friend back when he worked at Pomona. I suppose you know he got fired from MIT and effectively banned from US universities for pushing this bolshie line. Thats why he ended up in Canada - no US university would employ him. I assume you have tenure MH so can't be threatened with that fate.
The representation of AI as something that just "happens spontaneously" reminds me of how medieval Europeans believed that snakes and worms and insects generated spontaneously out of the mud. What is the "mud" from which AI emerges??
This is getting deeper all the time. I was surprised yesterday to get an email from Zentralblatt Math (for whom I review on occasion), that they are getting a lot of inappropriate AI-generated reviews. Why would a reviewer do that?
From the grave, Simons continues to influence mathematics the way he wanted, embracing private, closed, elitist things (many of these LLM experiments are prohibitevly expensive for everyone but staff of corporate labs) over public, open, non-profit, etc. Think of the way they severely affected computer algebra research in US by funding Magma licences for all US campuses about 10 years ago.
Whether or not Quanta's editors are communing with Simons with the help of mediums or Ouija boards, I think they can take responsibility for their own editorial decisions.
I'd be interested to hear more about the effect of Magma licenses on computer algebra research, since I wasn't personally affected.
see https://sagemath.blogspot.com/2015/09/the-simons-foundation-and-open-source.html?m=1
For the context, Magma is a closed source system with the biggest customer also being the biggest employer of mathematicians in US.