[bamboozled]

I stopped using ChatGPT when it wrote some code for me that was using libraries that don't actually exist. Even if it wrote said libraries for me, I'd have way too much work on my hands to validate whether or not it was correct.

I also asked it to write my sample code for other things like Cloudformation where it just made up directives and other configuration options that don't actually exist. Got a bit too abstract for me.

Personally, I wouldn't be replacing everything with ChatGPT?

[13years]

I've had very similar experience. After seeing all the hype of people claiming they have written all kinds of interesting things in code with ChatGPT I gave it a try.

Was never able to get it do what I wanted. It often seemed to make calls to non existent browser functions. I would tell it that function didn't exist, it would then rewrite it again, but still wouldn't be exactly correct.

Sometimes it was useful for doing something I've never explored, as I could get hints for how I might do it, but the accuracy was terrible.

[kranke155]

Ir doesn’t seem to have any real understanding of reality and it will just hallucinate answers. I literally don’t understand what the hype is about. I got lukewarm on it when it couldn’t give the right responses about Christmas traditions in Portugal and Belarus.

[bamboozled]

I'm fairly surprised Microsoft is going to try inject it into their search product for all the reasons you've cited.

[amf12]

I think it's all a hype right now. I feel they wouldn't enable it for most queries after a while and just have users use the chat interface. Hallucination is a major problem that hasn't been solved yet. Even with mostly accurate response, there are a lot of facts that are hallucinated, and it's difficult to trust because I have to use search to verify a lot of things.

For code, I'd prefer using something like Copilot, which is specifically trained for the task than using LLMs.