Cool stuff. Worth a deep read. Haven't had the time yet, but looking forward to it

any thoughts? @musiccognition.bsky.social @tecumsehfitch.bsky.social

This has been out as a preprint for a while. I expect if someone wants publish a critique then we won't have to wait very long

My art is not "AI slop"
I am very proud of this artwork, based on my own watercolors and drawings. It IS scientifically accurate, based directly on the work of the great comparative neuroanatomist Glenn Northcutt (and inspired by Georg Striedter's wonderful book). ChatGPT did the styling only...

I've just been shown the original watercolor paintings. Legit stuff.

There's an interesting question about the use of AI for assisting in art creation. But this is not a question of just using AI to generate slop.

Hi everyone! This was partly painted by a real person, and then AI was used to put on the finishing touches.

Perhaps Tecumseh Fitch will finally get on bluesky to clarify!

Probably this point will be obscured by the (well-intentioned) anti-AI-slop echo chamber...

Btw, this is a legally-binding poll, right?

Go go left

I should say also that I do use LLMS for coding a lot. And they are an incredible labour-saving device for me personally. So perhaps I take some of your criticism personally.

I like your work, and I agree with a lot of your posts. Yes, there are ethical considerations. But then you leave comments like this:
"It means you don't know these skills at any depth and have decided to stunt your growth. Truly saddening!"
It makes me pause, and think that you've gone too far.

Wow. I'm supportive of banning LLMs in the classroom. But wow. This discourse is getting odd and toxic. It is possible to both know skills and to benefit from LLMs. One could argue that instead of searching stackoverflow one should read documentation, otherwise you might stunt your growth.

I'm open to that. I find that the semantic distinction between the two is often not useful. Theories (and any logical frameworks), after all, are just boolean models. If you get this, then theories (as I think others think of them) are basically models that are not very specific/detailed.

Aside from the Box quote, I'd like to remind people that "All empirical observations are wrong, but some are useful" (Me)

Outside of physics the distinction seems more clear, since you rarely get a cohesive integrated theoretical framework that is mathematical and predictive. E.g. Darwin's theory of evolution can be boiled down to a few logical statements. Population genetics is a system of models of evolution.

I've been pondering the difference between models and theories for a while. I haven't heard much consistency in discussions. I certainly get different answers depending on the discipline. The one thing in common is that the word 'theory' seems to be better regarded. - is the standard model a theory?

Can you send it to me also? I'm now curious

Agent-based models by themselves are useful for extended thought experiments. If you want to know, given X, and Y, and some social dynamics, what is Z? But you can't compute it in your head.
Some great work has been done on evolutionary/game theory of cooperation. Sometimes good data is hard to get