3 ms·
Considering the LLM doesn’t pull the code it generates from thin air, I’d argue that it has at least the same bug density with human written code. However, als
by bayindirh 2mo ago
Considering the LLM doesn’t pull the code it generates from thin air, I’d argue that it has at least the same bug density with human written code.
However, also considering that LLMs can’t judge the code they have emitted, humans have the advantage here.
Lastly, why we downplay the bugs in the code emitted by AI? Why the lower standards?
- AI generated patch allows access (https://news.ycombinator.com/item?id=49331423 https://news.ycombinator.com/item?id=49331423)
- Bug in AI scribe creates big problems (https://news.ycombinator.com/item?id=49294441 https://news.ycombinator.com/item?id=49294441)
As an addenda, human written code is much more maintainable with respect to AI emitted blobs. At least a band of humans can forge that code to a much better shape.