In the 50’s they thought we would have intellegent robot butlers by the 70’s. They had solved more structured problems that seemed hard, like chess, and figured language and simple physical tasks couldn’t be much different. They came up with some hacky chatbots and things in the 20th century, but it was all cheap tricks like strategically changing the subject - I talked to these things enough to tell. ChatGPT passes basically every test of short-term language reasoning we can throw at it. It’s solved the problem for really basic purposes. It can take your Wendy’s order without any fine-tuning.

Alright, I’m going to respond to the rest of this in quip-like fashion, since you’ve touched on a lot of separate-ish points here, but the tone intended is still neutral.

Image recognition was great beginning the late 2010s.

That was literally the same tech we’re talking about here, just earlier and with a slightly different structure.

For instance, machine translation and spam detection (document classification) were getting really good by the late 2000s.

You and me have different memories of older machine translation. It could replace words and a few phrases fine, but it broke or produced awkward phrasings very often. It didn’t engage with the underlying meanings at all. Spam detection worked well, but not similarly smart, and IIRC in some case was neural nets again.

violating copyright,

Disagree.

It can also be somewhat useful as a search engine so long as you want your answer to be authoritatively worded but don’t care if it’s true or not.

Or if the answer is easily verifiable, like it has been in my own cases.