undefined | Better HN

0 pointsern1y ago0 comments

I recently needed to help a downstream team with a problem with an Android app. I never did mobile app dev before, but I was able to spin up a POC (having not coded in Java for 22 years) and solve the problem with the help of ChatGPT 4.0.

Sure I probably would have been able to do it without ChatGPT, but it was so much easier to have something to bounce ideas off-of. A safety net, if you will.

The hallucination risk was irrelevant: it did hallucinate a little early on. I told it it was a hallucinating, and we moved onto a different way of solving the problem. It was easy enough to verify it was working as expected.

0 comments

1 comments · 1 top-level

gexla1y ago

Seems to me this is the equivalent of fast retrieval and piecing together from a huge amount of examples in the data. This might take far more time if you were to do this yourself. That's a plus for the tools. In other words, a massively expensive (for the service provider) auto-complete.

But try to do something much more simple but has much fewer examples (a typical case is something which has bad documentation) in the data, and it falls apart. I even tried to use Perplexity to create a dead simple CLI command, and it hallucinated an answer (looking at the docs, it misused the parameter, and may have picked up on someone who gave an incorrect answer in the data.)

j / k navigate · click thread line to collapse