How does the current AI hype compare to the hype prior to the 80-90 AI winter?

No comparison. The previous theory of AI was that we would accumulate and structure expert knowledge so that it could be repeated deterministically -- to train computers on things like medical diagnosis.

The current theory of AI is best described by The Bitter Lesson[0] essay and it's the exact opposite -- don't store any knowledge but make predictions from weighted models.

This new version is much more exciting to the public and to investors because it isn't constrained by expertise, knowledge, or anything else! Without "training" anything on Shakespeare or science, we can ask for a science paper in The Bard's voice. This seemingly miraculous parlor trick is what gets all the attention and makes people think that, with a enough of a Big Black Box, we could get anything and everything to come out the other side.

More succinctly, the ability of LLMs to posit on any topic and gleefully make up convincing and fake answers has lit a fire under the hype machine where before AI was only thought of or needed for "expert" use cases.

[0] http://www.incompleteideas.net/IncIdeas/BitterLesson.html

[1] Source: I worked on an expert system at IBM and now make RAG apps.

Kind of reminds me of the self-driving cars of the early 80`s. Yes, it worked! I a van, crammed with racks and racks of then-state-of-the-art electronics. But, as someone back then mentioned; While the car crawls at 30 km/h on an empty road with clear painted lines, it was more like driving 70 km/h with closed eyes for 5 seconds, before opening them briefly and then closing them again.

The I would say, we're doing the same now, with loads of electronics, even more power consumption and hallucinating models. Working, but don't bet your life on it.

Hard to compare, different times, no internet, no social media, but there was this idea that AGI was just around the corner, if only we could train a computer with the world's knowledge. There were even projects where people where inputting a lot of logic like statement about the world to use as a data source. The AI that was available was very rudimentary. Expert systems were the big breakthrough.