I’m reading that description and I don’t understand how this is different from literally every autocomplete/spellcheck function that learns to predict based on your past typing habits.
I thought the definition of LLM and why it’s different from other predictive text functions that have been on all sorts of devices for decades, is that it’s large and massively multi-victim and therefore not possible to localize to your device.
(I mean I do turn this shit off whenever I can, and I have for years, because it’s annoying. But I’m morbidly, dreadingly curious to know what’s new about this version of it)
actually, yknow what? I’m not done with this thought.
all throughout the 2010’s, when I was trying to get used to drawing and photo editing in various software, a lot of the programs had a lot of features that got described as “AI.”
they were all sorts of things! they ranged from stuff like “smoothing your brushstrokes” to “running a filter that made the picture look more like a painting.”
I don’t know how all of them worked. But they were not the same thing as the LLMs that get talked about today. They existed so far back in time that if any of us saw the date, we’d immediately say “this wasn’t AI, it’s from before AI”
but the programs, frequently, would call these features AI.
and probably a lot of them still do! even if they haven’t changed much, even when they exist alongside the other features that are powered by modern LLMs.
because there is no fucking standard for what can and can’t be called “AI.” people have been calling various software “AI” since the fucking 1960s. and not nearly enough people in the present debate even talk about this.
and very few of the people in the debate even know the differences between all the different things that get called “AI.”
I sure as hell don’t!
Like, this autocomplete thing described by OP? I still don’t know what makes it different from other similar things.
I know just enough to be aware that predictive text, and personalized spellcheck, and autocomplete functions, have been around for ages… and that the LLMs we hate are, in some ways, a more complicated version of that… and also, that some of those old functions that used to work pretty okay are now being sneakily replaced by LLM stuff that doesn’t work at all, because it’s trained on bullshit datasets like “everything that’s ever been typed into google docs.”
so every time I see someone say “if you use any AI in your art/writing at all, you are a bad person and not a real artist!” my immediate interpretation is that the person is, um… generating an automatic response to a term that they don’t understand and haven’t thought about the meaning of.
…if there even is a meaning, in any useful sense of the word “meaning.”
and like I said. I turn this stuff off too. I turn off predictive text and spellcheck when I can (and then turn it off again, and again, and again, because computers are assholes that don’t respect my decisions). I gave up on digital drawing a long time ago. My art these days is colored pencil on paper, and wire sculpture, and fiber crafts, because those are the only kinds of media I can trust to not change how they fucking work every few months.
but… to share my art online, I still have to digitize it, take a photo, make a scan. And the moment I do that, it’s gonna be influenced by whatever software I’m using.
and which of that software includes “AI”? and by which definitions? and is it possible to avoid? who knows!
there are lots of debates that terrify me because the people involved are expressing very strong opinions, to the point they seem about ready to commit physical violence over it– despite the fact they clearly aren’t even agreeing on definitions, or being conscious of any of the small nuances of difference between different cases.
But this one? is among the scariest, not gonna lie.
ALT