2 ms·
But if it and we don't write tests for our software - and think through writing tests when we write the code - the losses and liability may exceed such measures
by westurner 3y ago
But if it and we don't write tests for our software - and think through writing tests when we write the code - the losses and liability may exceed such measures of return.
Current gen LLM (Large Language Model) AI are trained to regurgitate non-formally-verified code; and even if they were exclusively trained on labeled-as Formally Verified code, they would produce code that does not pass formal verification.
The LLM AI doesn't care about your liability or others'; we need to think through our code to write sufficient tests (given incomplete code specifications).
That $1.5T figure must be gross not net.
- westurner 3y ago... Though, I have enjoyed looking things up from the docs using LLMs in order to avoid getting distracted from focusedly coding.