When You Ask AI, Everything Sounds Right: Until It Doesn't
When you actually try it, there are many cases where it doesn't work. In the end, verification has to be done by a human.
5 answers
I also pasted the code AI wrote exactly as it was, and it all blew up on edge cases, so I wasted a day. Since then, whenever I get an output, I first run it locally and check it while adding logs. It takes more time, but it hurts less.
Well, isn’t that more of a prompt issue? If you ask without giving context or constraints, of course you’ll get nonsense. It’s true that humans have to verify things, but it doesn’t seem right to blame only the AI.
I make and use a verification checklist. 1) Ask for sources 2) Double-check numbers and dates against the original 3) If it’s recent info, run a search and cross-check. Just doing this filters out half.
Agreed. Especially that plausible-sounding tone—that's the scariest lol
What's the source?