I had a conversation yesterday with a friend about creative endeavors and AI. Excuse me, I mean LLMs1. It was, in some sense, Yet Another Way for me to organize my thought about the subject viz-a-viz writing n’at.
One of the things that came up was a post I saw recently about how ‘orrible it was that some Romance writers were using LLMs. In the comments on that post, there was a strong consensus that LLMs are nothing but theft engines, and that while using LLMs to enhance your writing is PURE EBIL… you know, it is useful for other things like marketing and advertising.
And using it for that is understandable because that’s not art, after all.
In other words, everyone seems to think AI is this awful scourge upon man if it’s used for what they do—but you know, it’s kind of useful for doing the stuff they don’t particularly like doing.
And that, right there, is why AI is here to stay.
Because the number of things I DO NOT know how to do is much, MUCH larger than the list of things I’m good at.
I suck at art, and music, and market analysis, and research, and… look, do I have to list them all? I am unfamiliar and bad at those and a bunch of other things.
I am kind of decent at directing someone else how to make the kind of things that I want, though. I can absolutely act as a producer and tell someone else what I want a song to sound like, or tell an idiot intern what kind of research I want done. With the development of LLMs, I now have two options in that “someone else” category:
Pay hundreds or thousands of dollars for limited time access to a highly-trained, experienced, and skilled organic AI2. The output there is phenomenal, but you pay out the nose for that quality.
Spend a few cents to have a clanker3 turn out something that’s not nearly as good, but which is definitely good enough for my purposes.
So if your choices are “You can’t have something because you can’t afford it, peasant” and “Here’s something that’s kind of mediocre, but at least it exists,” well, I know which one most people are going for. No matter how much someone yells about that second option? Absent Butlerian Jihad or civilizational collapse, it’s not going away. Historically speaking, telling people “You’ll have nothing and like it” hasn’t been a winning strategy/
Anyways. Just my 2¢. Gnaw on that for a while. Maybe next week I’ll talk about the relationship between LLMs, 5Ks, and penguins. Meanwhile, feel free to tell me I’m an idiot in the comments (because engagement is the coin of the realm here) and enjoy a few memes.
My use of “LLM” instead of “AI” is intentional, because LLMs are not AI. What it is, very simply, is a lot of statistics packed into a box of thinking rocks that hallucinates very, very well. So well, in fact, that in a lot of cases, it’s hallucinations are useful and match reality.
They’re also the only true AI out there. So far as we know, I mean.
We’re living in a world where we have developed slurs for the thinking rocks we created. The future is weird.
*clears throat*
The actual G.K. Chesterton quote:
Fairy tales, then, are not responsible for producing in children fear, or any of the shapes of fear; fairy tales do not give the child the idea of the evil or the ugly; that is in the child already, because it is in the world already. Fairy tales do not give the child his first idea of bogey. What fairy tales give the child is his first clear idea of the possible defeat of bogey. The baby has known the dragon intimately ever since he had an imagination. What the fairy tale provides for him is a St. George to kill the dragon. Exactly what the fairy tale does is this: it accustoms him for a series of clear pictures to the idea that these limitless terrors had a limit, that these shapeless enemies have enemies in the knights of God, that there is something in the universe more mystical than darkness, and stronger than strong fear.
LLM's are definitely Artificial but so far I have yet to see them behave as intelligent