Jeff Atwood"The survey found that 77 percent of respondents felt favorably about using AI in their workflow and that 70 percent are already using or plan to use AI coding tools this year." https://www.theverge.com/2023/6/13/23759101/stack-overflow-developers-survey-ai-coding-tools-moderators-strike
Augray@codinghorror the way I see it is this: even in an optimally designed tech ecosystem, you will still have times where you need to translate a question that is most easily expressed in natural language (“How do I do X in library Y?”) to pointers to the right code/docs. Reading all the docs is high latency. Searching the docs has lower latency. In some cases asking on/searching SO is even lower. In many cases, it turns out using an AI tool can have *much* lower latency.
@augray provided it's not wrong. ChatGPT told me ZZ Top had no song named "I Need You Tonight". That's.. fucking wrong?
@codinghorror true, I personally don’t use it to “learn new facts” in isolation. I verify it in some way (type the functions it references into my ide and look at their docs there). When it is accurate, I get a “cache hit” and find what I’m looking for faster than searching the docs myself. When it is wrong, it’s a cache miss and I fall back to traditional search mechanisms.
@codinghorror basically there are cache layers. The lowest latency “information caches” are also the least accurate. You just have to learn how to identify cache misses.