> Is it everyone else that's wrong or.... This is disingenuous. OP is right. ChatGPT is mostly inaccurate and contextless by nature. It always produces somet

> Is it everyone else that's wrong or.... This is disingenuous. OP is right. ChatGPT is mostly inaccurate and contextless by nature. It always produces something confidently so it's very easy to think it's the right answer. Now, it can be extremely useful for any task that can be easily verifiable and you don't know the syntax, how to approach something, etc. Because any decent software developer can use it as part of the prototyping or brainstorming and get wherever they want to go in coordination with ChatGPT. What you can't do is assume it's "the expert". You are the expert and you're the intelligent one and the chat generates potential useful things. That's on verifiable stuff. On other things it can be so laughably bad it's impressive how much "everyone" (as in articles, hype, HN users who downvote criticism as being from luddites) pushes it as something that is not: an AI that can think and be relied upon and any inacuracy will just "get better with time" aa opposed to "the model and also the politics around it" doesn't have a path to get to that idealisation that is being sold. It's a tool, but it's usually sold as better than it is, like in this case, presumably with the intent of relying upon it to save cost in some key integration point. The problem is that it mostly won't work and comparing it to bad humans or worse integrations doesn't show the fundamental low ceiling that it has. I think people with authoritarian mindsets (and I don't mind Left or right but inherent trust in authority) easily want this to be a source of truth they can magically use, but there's no path for that to be true. Just to appear true.