I've been watching the AI Helper on Perchance as it creates my gens, and it's actually pretty entertaining! It uses words like "ugh" when it finds something doesn't work, or look right, and phrases like "wait a minute!" to consider another angle. But it is also process and writing code, so why put the "human" language in there? I get that it's designed to let the user see the process (if anyone is interested, which I am), but it seems like the developers are trying to "humanize" the ai?
3 Comments
breadsmasher@lemmy.world · 4 pts · 8d
you answered your own question. its to humanise it
ccufcc@lemmy.world · 2 pts · 8d
why train it to not think like human if you cannot understand it? then you need an AI translator? no I know nothing about LLM but I think it make more sense if they mimic the way we think
daveed@lemmy.world · 2 pts · 8d
I'm not criticizing. As I said, I think it's entertaining, sometimes laugh out loud funny! I guess I just want to start a conversation about it. The Perchance AI Helper has a more subtle thinking section, in grey, and one in bold white that is actually meant for the user to see. The subtle thinking is the AI talking to itself, while the bold is it talking to the user. The subtle thinking doesn't expand, you have to expand it to see the whole thing. That's kind of what I was getting at: If it's going to output text for the user, why show the "internal" thinking at all? If I was a developer (of ai), of course I'd want to see all of that--but the ai has to translate it's actual "thinking" process into language the developer can understand, so the extra expressions of human frustration, rethinking something...just seems like unnecessary effort? As for "mimic the way we think," LLM (or any ai, for that matter), can't really even do that. It is translating processes we can't even comprehend into language we can, but someone had to tell it to include unnecessary expressions. It's like discussing philosophy: Makes you think, but...