I wanted to clarify one or two points about my post on Terry’s blog.
Let’s begin with the following:
Evaluate all work as if it is AI-assisted.
I gave some broader context for this remark in the original post, but it never hurts to clarify important points that can be interpreted in different ways. The point of this suggestion was not to say “assume everyone is cheating and using AI without saying so.” Nor was it to say “AI disclosures are pointless.” Rather, I am making a broader point about what type of work we should consider valuable. I would encourage people who take this idea seriously to think about their own work. Are you trying to prove conjectures for fame and glory, or are you trying to further our collective understanding of mathematics? Those aims can sometimes coincide, but they are not the same.
I do not think this approach in any way preclude mathematicians from working without AI. It simply means asking how and why any new result helps us understand mathematics; whether that be an interesting question, example, a new proof of an old result, or a new proof of a new result. What further mathematics does it make possible? Importantly, the work we produce is more than simply the sum of the written words in our preprints; it includes the discussions we have with others about our work in seminars and elsewhere.
As for AI disclosures themselves, I anticipate that they may eventually become unnecessary. For now, I wish only that people were more honest about them. My main issue is with people trying to have their cake and eat it too. People are using AI to prove new theorems while acting as if being first should still earn them the same credit. By deliberately obscuring what the AI did, in otherwords, they hope to get credit that traditionally came with such proving such results first while simultaneously denying the new reality.
Twitter buffoons
Some people seem to have interpreted this as a reference to Levent Alpöge. That was not my intention. I will say this about Levent: he bet his career on AI at a time when very few mathematicians took it seriously. I do not think the style of his tweets have been entirely helpful to the mathematics community, but I did not intend to single him out.
Instead, If you really want to know who counts as a Twitter buffoon, I have prepared the helpful list of questions below. Only if you answer yes to 100% of these questions should you consider yourself a buffoon:
- Are you on Twitter?
Clarification: I don’t have a Twitter account, so perhaps I am missing a hotbed of serious, polite, and measured discourse. But what reaches me seems very far from the collegiality we claim to value in mathematics.