My favorite tell is when a write-up starts with a verbose explanation of given knowledge on a subject. Yes, we all know what ‘World Wide Web’ and ‘Internal Combustion Engines’ are.
Get to the f’ing point.
No “we” all don’t. Ask anyone who works support how fucking stupid the general population is about shit they use daily. Let alone stuff they heard years/decades ago. Seriously. Just start asking people to point to “the computer” and see how many point at the monitor even when it’s clearly an 80” wall hung TV.
Biggest issue I see is that these LLMs tend to repeat themselves after a surprisingly short number of times (unless they’re sufficiently bloated like ChatGPT).
If you ask any of the users of Sillytavern or RisuAI they’ll tell you that these things have a long tail of not being very creative.
Seriously, fuck off with the AI shit. At best it’s intelligent regex. And “intelligence” here means a specific thing.
Do you even know what regex is? Your neural network needs more training.
Do you even know what regex is?
Do anyone? I know I just add more .+? until it works
Probably the same people who use git cli instead of ide plugin equivalents
Yes but those people also use vim, so we don’t talk to them.
It’s a language model… Not a regexp. :)
I agree it’s overhyped and I think we will have these models in everything from phones to cars, giving us voice commands and summary functions.
It is not going to become self aware anytime soon. :)
intelligent regex
That would be much, much worse than what we actually have. Complex regex are positively Lovecraftian. You’d be chanting “Ia! Ia! Cthulhu ftaghn!” before you knew it.
i prefer the upgraded autocomplete analogy but yeah
