I think the argument is that OpenAI should have done the work of cleaning up the proofs in the first place before dumping them onto the math community (or worked with mathematicians to guide the research process). It's as if an energy company dumped 20 tons of waste next to a village and the townspeople decided to clean it up. "How does this benefit the energy company" might not be the right question to be asking.

I'm struggling with the right analogy but the "waste" is more like...idk, hundreds of those giant crates they go through on antique shows.

The contents are actually extremely valuable, it's better to have them then not, but it still takes nontrivial expert work to sort and parse them.

A landfill with a hard drive full of Bitcoin is still a landfill.

this is a strange analogy. no one is forcing the math community to read this

  >I think the argument is that OpenAI should have done the work of cleaning up the proofs in the first place

So the complaint is, OpenAI didn't replace the role of human mathematicians.... enough?

If this gets 'fixed,' I highly doubt the math community would be any happier.

The complaint is that OpenAI is taking the job out of mathematicians. But obviously this doesn't sound that appealing of an argument so you get people piling on complaints like "this is slop and we have to clean it up". When in reality, the fact that they have anything to do at all is a privilege haha

Uh the ultimate training data... Turning their slop into human accessible anything even if proven wrong is training data they should be paying tens of millions of dollars for.

Maybe, but I don't see why it would be better than selecting random maths papers

Because it corrects the model output using model input

If that's the goal then they can do it the other way round.

1. Pick random high quality maths papers that are outside of the training window 2. Convert them into a form which preserves the quality of the results but throws away the quality of the writing. For example, formalise them or keep only a high level summary of the proofs and throw away the introduction/conclusion 3. Generate full maths papers out of these summaries 4. Try to make the output maths papers be as similar as possible to the input papers.

You can see this as similar to training an auto-encoder, and I don't see any real downside to doing this over training on these corrected papers. Yes, the individual results would be less significant but you can't exactly mass produce training data by publishing papers and hoping someone fixes it.