Woit diagrams?

SUM OVER HISTORIES

OPUS 4.8 

" . . . Want me to draw the ℂP¹ × ℂP¹ picture with the Euclidean antidiagonal and Lorentzian torus slices marked? It’s the kind of thing that’s much faster to see than to read. . . "

WOIT 

Counterpoint,
I’ll let this kind of comment through just once, but please don’t submit more of them. It’s a mishmash of true statements, , not quite true statements, semi-insightful comments, and utter nonsense. Reading it, one is tempted to carefully think about each statement, and possibly try and respond, clarifying the unclear and correcting the not quite right or completely wrong.

The problem is that the effort required to do this would be orders of magnitude larger than the effort needed to generate it. With the current state of these agents, this is the big problem with them: teasing out what is correct and useful from what is unclear, a misunderstanding, or utter nonsense is more effort than it is worth. If this were coming from a student, it could be worth spending the time needed to interact with them because they would be learning a lot (and you might learn something).

For now, seems to me best to ignore this kind of thing and wait and see if, as widely advertised, it gets much better soon.

I do confess to being tantalized by the last paragraph, curious to see if it can produce a useful visualization of that point. It seems plausible that producing visualizations like this is something these agents might be useful for (and drawing is something that I personally am horribly bad at).