I was going back and forth, changing my post title from "Growth is the solution" to "The solution is growth" and back again. So I figured I'd look it up:
Good answer! It's nuanced, even. Way better than what I expected after it fumbled "public" versus "public sector" the other day.
2 comments:
I don't know whether this is true or not, but I think: these robots might be very good at structure-of-language kinds of questions, since that's really what they have experience with and know best. Anything else (e.g. facts about the world, or what the public sector is, or what the public is) they have to sort of infer from the text (and maybe images, whatever else) that they've seen. It seems like: as they get larger and smarter, they are better able to make those deductions and they actually do learn about the world. But Google probably has a small/dumb one doing those overviews, since it has to be cheap enough to run billions of times per day (or however many google searches people do).
That is all just sort of speculation, though - who knows.
Sounds right to me, Jerry. The AI responses (this time, anyway) strike me as well-written. This suggests to me that they are indeed "very good at structure-of-language kinds of questions".
The thing sounded so human, telling me it sounds a bit more like you are identifying a solution...
and it pointed out something I had missed: without emphasizing that it is the complete answer.
Granted, I did what I almost always do: I accepted the observation as accurate rather than evaluating it for a high bullshit content. But it still seems right, a couple days later. No objections have suddenly popped into my head...
It still bothers me, though that one of the "AI Overview" responses I got included 6 links, and 4 of the 6 were links to Investopedia articles. When I find search results from Investopedia I usually don't put much faith in it.
Something else I use when writing: Ngram alternatives
https://books.google.com/ngrams/graph?content=is+the+solution%2Cthe+solution+is&year_start=1800&year_end=2022&corpus=en&smoothing=3
Post a Comment