|
|
|
|
|
by zahlman
408 days ago
|
|
There was a noticeable inflection in the question-rate-vs-time curve around the time that ChatGPT was released. Which is fine. If your question is not answered by `site:stackoverflow.com how to do the thing` but it is answered by an LLM taking `how do I do the thing?` as a prompt and synthezising existing Stack Overflow content, that is inherently not a suitable Stack Overflow question. Because anyone else could put `how do I do the thing?` into the same LLM. It's not any different from using a traditional search engine. (And when the LLM fails by producing a wrong synthesis, then blessing that result by putting it on Stack Overflow is actively harmful - which is why it's banned by policy.) |
|