Hacker News Reader

Having half of the title is too triggering. That's not fair to the CS prof. Misleading.

But to comment on the article, although I've encountered several falsities in chatgpt output, I'm still very surprised that it could devise algorithms when I would give him the criteria.

And it would even output some code samples.

It shouldn't be underestimated as a programming aid.

View on HN

ChatGPT sometimes reminds me of a junior programmer.

I asked for a code example of clipping a polygonal shape with another polygonal shape, and it wrote code that just deleted vertices from the subject polygon that were outside the clipping polygon.

Interestingly, the correctness of the code it generated depended on which language I asked for it to generate the code in. I assume that says something about the corpus of code it analyzed and the relative quality between the languages.

In which languages were the results correct? I would assume Python as I've been told it was more prevalent in the training corpus.

I’m now curious. Anyone up for doing a breakdown of ChatGPT correctness vs generated language?

Reply on news.ycombinator.com