OK, I played with ChatGPT for a while.
-
OK, I played with ChatGPT for a while. At least for programming tasks, Claude is head and shoulders above it.
You give Claude a task, it generates valid, working code (tested in a virtual environment), which might or might not do exactly what you wanted, and then shuts up, awaiting further instructions. It also doesn't shy away from telling you that you're wrong when you say something that's not correct.
As opposed to that, ChatGPT generates 3 different kinds of bullshit that was never tested in anyway and almost certainly doesn't work. It's only "criticism" is to say "the bug in your code is..." when you paste the very same fucking code it gave you a moment ago. It keeps repeating "Ah, I got it now" and "Here's a final version that works" while misunderstanding stuff you've explained it 2 prompts ago and delivering code that's still just as broken as before. And at the end of each delivery, it always begs to generate another 2-3 kinds of bullshit that you never asked for.
-
R relay@relay.infosec.exchange shared this topic
-
OK, I played with ChatGPT for a while. At least for programming tasks, Claude is head and shoulders above it.
You give Claude a task, it generates valid, working code (tested in a virtual environment), which might or might not do exactly what you wanted, and then shuts up, awaiting further instructions. It also doesn't shy away from telling you that you're wrong when you say something that's not correct.
As opposed to that, ChatGPT generates 3 different kinds of bullshit that was never tested in anyway and almost certainly doesn't work. It's only "criticism" is to say "the bug in your code is..." when you paste the very same fucking code it gave you a moment ago. It keeps repeating "Ah, I got it now" and "Here's a final version that works" while misunderstanding stuff you've explained it 2 prompts ago and delivering code that's still just as broken as before. And at the end of each delivery, it always begs to generate another 2-3 kinds of bullshit that you never asked for.
@bontchev Given that this comes across like a session with a three-year-old who is mentally challenged, the question I have is are AI tools being mandated for companies so that the employees can train the models so that they can *ultimately* replace the workforce?
-
@bontchev Given that this comes across like a session with a three-year-old who is mentally challenged, the question I have is are AI tools being mandated for companies so that the employees can train the models so that they can *ultimately* replace the workforce?
@AAKL Dunno; I haven't worked in the industry for 2 decades already. I'm in the academia, about to retire.
The point I was trying to make is that, at least for programming tasks, the models are *very* different. ChatGPT wasted me a whole day before it reached the conclusion that what I wanted couldn't be done. (It kept insisting that it can be.) Claude is very good. Despite the occasional mistake or misunderstanding, talking to it feels like talking to a person. It really helped me produce useful stuff. ChatGPT feels just like chatty computer and produced mostly bullshit.
