Yeah, I know, good job posting in this thread about half a day after the discussion died down. I was busy, shuddup.
Either way, the short answer is: Yes, bots can create art. No, they cannot create art right now.
The issue with artificial intelligence in its current form is that it doesn't know what it's doing. We're using brute force methodology to make them guess with as much accuracy as they can what the thing they're meant to be guessing is. Note that a lot of this guessing is reactive- You give them an image of a frog and they say "I'm like 98% sure that's a frog right there". This is the most common and refined type of AI we have right now. It doesn't create anything- it's just able to recognize and react.
That doesn't mean there aren't AI that can't spontaneously create right now. Take Deep Blue, the chess computer- through playing itself more times than 100 humans combined could play in a lifetime, it's able to create this unique style of chess nobody else can with the sole purpose of winning. That said, Deep Blue is just operating on a bunch of self-gathered data, and is just recreating the most effective strategies it remembers. Still, it's a step in the right direction of "creation" and "creativity" since it did, on its own, learn how to play chess really good.
A step beyond even that would be those "image generator" bots you see. I hope you're familiar with the concept, but if you're not,
this anime doesn't exist. Here you can see that the bot certainly is doing something, and creating pretty good results considering it's a robot... but there's one critical flaw: It also cannot filter out its own imperfections. Just like how the image recognition bot can say "This is 98% a frog" the image creation bot can only ever make 98% of a complete image. It'll never create a 100% truly 'correct' looking art piece because it does not inherently understand what it's actually doing, it can only guess "this is what you humans want, right?" and even that comes with the caveat of only having gotten to this result after being told thousands of times "yes" or "no". It didn't spontaneously learn anything, thus it couldn't actually create any art.
This, currently, would be the same issue you'd have for literature. While generating an image is easy (
for a robot. I know how hard it is to draw.) since it's only one, large static combination of colour, creating the
nuance of language itself? That's a different story entirely- quite literally, given the topic of the thread. There are story bots out there, but they're more than a little jank, because there's just one issue: Language does not make any kind of sense.
Language is inherently subjective as it is an ever-changing, continually evolving process of people adapting their vocal chords to make the other guy be gooder at ur spoken tongue. In order to write in a language, it's not enough to have objective knowledge of the words in that language: you need to be intimately aware in the exact manner and nuance of how people speak and talk to each-other... and you need to be aware of that in a
subjective manner. That is what makes a text truly unique and yours; the fact you speak in the way
you have adapted to based on your experience conversing with other people. If a robot analyses each of these individual voices and finds and average between them, even if we made that coherent, the 'average' would lose all possible meaning because it would sound so boring and, well, everyday. The AI wouldn't have a voice of its own because it has everyone's voice.
This is why I said "unless the AI can be meaningfully subjective and arbitrary, it cannot create art" in the status bar yesterday. AI, in its current state, does not and cannot know why it's doing things right or wrong. It's arbitrary, yes, but not with a purpose- it's arbitrary because we told it to be. Unless an AI can start actually thinking for itself, having freedom of thought, and knowing
why it's doing what it's doing, it cannot ever make something we will truly enjoy without going "okay but this part's super wonky though".
Even if the premature AI we have could, by some miracle fluke, generate a piece of art of any kind that didn't have blatant flaws... well, it wouldn't matter, because it'd have generated that once. We ignore the millions upon billions of failed attempts prior, and we'll continue to ignore the attempts after, too. Because the AI is brute forcing until it finds something that works, it doesn't have a goal in mind beyond what the
human behind the screen tells it to. Without approval from the human, the AI would've tossed it out and kept trying- so it's also the human element that's the deciding factor in the end.
PPQ made the point of "but we ignore the thousands of wannabe manuscripts in real life too" and, yes, but how is that relevant? Even if we didn't like those manuscripts, each and every single one of them had more purpose and vision behind them than the robot stringing random words together. Even the worst manuscript among them would be better than 100's of attempts by the bot, on the sole basis that it has more of a cohesive flow and at least makes some kind of sense. Even if they didn't, at least the author could explain why they did things the way they did, and if their answer is "idk, i wanted to be random" that's still a
decision they made. The AI's answer will always be "because the data told me to" which means it will always generate predictable, meaningless results.
Freedom of thought is a critical component in creating consistent art. Even if an AI wrote the best piece of fiction on the planet, it couldn't do that twice.
A human
could do that twice, and that's what it means to make art.
