Stack Overflow is almost dead

> This, in turn, leaves "AI" implementations with no new human content to feed off of. Now what?

You seem to be under the impression that AI needs more than all recorded human knowledge up until 2024 to reach the same level as an average SO contributor. It doesn't. Because none of the average SO contributors did.

It is unclear what algorithmic improvements are required to leverage the available data to get AI to AGI, but a lack of data is definitely not the bottleneck.

One could say that these AI systems aren't sharing their solutions (or questions) with other AI systems and that the world would benefit from it if they did, though. Perhaps it's a good idea to have some shared space for AI systems where they share the validated solutions they synthesized.

I've been trying to get ChatGPT to write some Emacs Lisp for me and it sucks. Few things are better documented than Emacs. There's several hundreds of pages of documentation, and several million lines of Elisp, but apparently that's not enough.

And I'm not asking for some beautiful architecture from ChatGPT, I'm just asking for simple hacks that get the job done. Elisp is designed to make simple hacks easy, but not easy enough for ChatGPT I guess.

Like, I asked it to make a command which would move the "mark" and the "point" so that the full line was selected. If a selection covered only part of a line, I wanted the selection to expand to cover the full line. To do this, all you have to do is move the mark and the point so that they surround a line. ChatGPT couldn't do it. It would only move the point, but not the mark, it wouldn't do anything with the mark. I explicitly told ChatGPT "no, you have to set both the point and the mark correctly", and then it wrote even more code that only adjusts the point, but not the mark--it would move the point to the beginning of the line, and then move the point to the end of the line, never touching the mark--it's stupid.

Per the post, it seems as if SO was going to going to slide beneath the surface anyway but that LLMs were one of a number of factors that maybe accelerated the process by maybe a few years. Without having deep knowledge of the area, this feels about right.

> You seem to be under the impression that AI needs more than all recorded human knowledge up until 2024 to reach the same level as an average SO contributor.

Replacing the average SO cobtributor isn't adequate to replace SO, and AI is able to “replace” SO effectively only since major models have gotten not only SO-as-training-data but web search (including SO) for immediate grounding.

And without SO or something like it with active human contributions it’ll have even more trouble replacing the value SO would provide for new questions and new domains where it will neither have SO traijing data nor SO query-time-search-results to use to synthesize answers.

"Having your problem addressed" is not a valid reason to post on Stack Overflow. You are expected before posting to have done enough analysis to the point where if your question is answered, you can solve the underlying problem yourself. When you are linked to a duplicate, it's because the person doing so believes in good faith that, to the extent that you have a question that meets the site's standards, answers to the other question will answer yours as well. This also means you are responsible for overlooking irrelevant details, reading the answers, making your own attempts to apply them, etc.

If the other question is actually different, you are expected to edit the question to make this clear - not by adding an "Edit:" section like in a forum post, but by fixing the wording such that it's directly clear what you're looking for and how it's different. This might mean fixing your specification of input or desired output.

It's difficult sometimes, and curators do make mistakes. Most frustratingly, it's entirely possible for two completely different problems to be reasonably described with all the same keywords. I personally had a hell of a time disentangling https://stackoverflow.com/questions/9764298 from https://stackoverflow.com/questions/18016827, while also explaining that https://stackoverflow.com/questions/6618515 really is the same as the first problem despite different phrasing.

But curators much more often get it right. Not only that, a few of us go out of our way to create artificial Q&A (https://meta.stackoverflow.com/questions/426205) for beginner issues that beginners never know how to explain, and put immense effort into both the question and answer. Some popular examples in the Python tag:

"I'm getting an IndentationError (or a TabError). How do I fix it?" (https://stackoverflow.com/questions/45621722) was written to replace "IndentationError: unindent does not match any outer indentation level, although the indentation looks correct" (https://stackoverflow.com/questions/492387) and a few others, with reasoning stated there.

"Asking the user for input until they give a valid response" (https://stackoverflow.com/questions/23294658)

"Why does "a == x or y or z" always evaluate to True? How can I compare "a" to all of those?" (https://stackoverflow.com/questions/20002503) was written largely as an alternative to the organic "How to test multiple variables for equality against a single value?" (https://stackoverflow.com/questions/15112125) after the latter was found not to help beginners very well (the original example was quite unclear, although it's since been improved).