I feel similar (now, one year after). But for the first few weeks of holding the title my dominating feeling was that the stress wasn't worth it and that whole thing is crazy stupid (still not sure I was wrong)

I love your remix. But it make me wonder who would be a receiver of this visual. Researchers generally know that it is more complicated, and for popular audience version 1 seems to be the best one: show familiar storytelling with a basic model as a commentary.
Still, I would like it as a 💻 sticker

It should be some kind of holiday

the problem is, that the very next day one need to get back to office and at least count all the pottery/lithics/bones from wet sieve heaping from previous season. And hope for next bit of tar.

I guess whatever is at hand at the moment. But I also enjoy writing in general too much, to avoid changing produced output^^'

Yeah, sorry, I was imprecise. What I meant was these lower level AI branches, where one can tract and supervise the training process, are more trustworthy for me. LLMs with added complication are not predictable as a research tools. But they write e-mails faster than I do, which is help enough.

yeah, like I sad, it is a bit of overstatement to that. it is only LLMing out datasets is bad research practice at this point, because level of confidence is so hard to estimate. As I see it, LLMs are more research subjects than tools at his point.

Btw that part of the 'BAD tool' is a bit of an overstatement, depending where you cut AI from ML etc. It is just LLMs that can't do reliable research (which is fine, for not all are born scientists)

I think that what most of the people in comments to the linked post try to convey is 'AI is a BAD tool for research'.
It is not to say that google is ideal

Population continuity in Western Europe is the actual model, see:
www.nature.com/articles/s41...
www.nature.com/articles/s41...

"what a nice example of the recultivation you have here!"

Oh! Also my personal favourite!