Re:If you believe in "Artificial Intelligence", take five mi...
by cetu86 on Saturday February 15, 2025 @04:32AM
(#43057232)
I'm currently using AI code completion.
Since then I sometimes have subtle errors in my code that didn't happen before. Here is how that happens:
AI suggests something to me that looks right at a glance. I accept it and move on. Then later I hunt down a strange bug. When I find it I'm like "wait, that line's wrong! I didn't write that".
It still fails at being confidently wrong.
The brief article hits at people trusting the tool without questioning the output.
Meanwhile, we have people using Codeium or Copilot to write code and that sort of works since the code eventually needs to be compiled and tested (unit, integration, system, requirement sell off)
There is no test for the truth available to everyone else.