I'm a coder with 30 years of experience and I'm also excited about the potential of ChatGPT to help with my work. However, it's very important to remember to double-check the code it generates for accuracy.
Especially for anything that is not boilerplate or tutorial level, I've seen a lot of mechanically incorrect and unhelpful code generated.
> I've seen a lot of mechanically incorrect and unhelpful code generated
This has been my experience. I've attempted to use ChatGPC 4 times to generate a functions or classes with various levels of complexity and every time it has produced incorrect code. It often fabricates functions, classes, or entire libraries and the few times it produced regular expression the regexp didn't actually do what ChatGPT thinks it will do.
I used it at work yesterday. I asked it to write a promise that returns true after 30 seconds. It did. Easy stuff I suppose...
I gave it some code which was an obscure looking Typescript type that was using several utility types. ChatGPT gave me a decent, if not wordy, explanation back in plain English.
Then I asked it which parameter to use in Google Mediapipe SDK to enable long range face detection. It didn't get the answer perfect, but it was still correct, and it sent me down the same path that a Google search would have.
I only played with it for 15 minutes after a meeting, but I can start using this tool at work today because that's how useful and easy it seems. We'll see if I actually reach for it today.
I very recently relied heavily on chatGPT to write a contract in solidity.
The tool gave me lots of example code I could use, helped debug some error messages & was much more efficient than google & stackoverflow.
I don't think I'll ever deploy the contract (It's a game that isn't very good), but I don't think it would have been this easy to do without a semi-literate ai dev to help me out.
I am mainly using ChatGPT to inform my google search options.
So instead of searching on google "what data stores to use for saving embeddings" . I ask ChatGPT "What are recommended datastores for saving embeddings". The second pre query gives me sense of better starting point, that to start with a generic query.
"Pretend you're very well articulated developer. Explain to project manager why we can't develop this feature faster than initial estimate. Argue with him till he capitulates to our demands"
Now just need a chat plugin to shove the text back and forth...
Especially for anything that is not boilerplate or tutorial level, I've seen a lot of mechanically incorrect and unhelpful code generated.