4 ms·
Stopped reading about a third of the way in when I realized this is AI "art" apologia with a heady serving of woke-washing. I'm sorry, but there is no universe
by kmeisthax 1y ago
Stopped reading about a third of the way in when I realized this is AI "art" apologia with a heady serving of woke-washing. I'm sorry, but there is no universe in which you can convince me that a large language model is an oppressed power minority.
I skimmed the rest to see if the author addresses the actual economic concerns belying the opposition to AI, but they don't. It's a dreadfully extended takedown of a bad argument that I would call a strawman but for the fact that plenty of artists keep trotting it out.
I scarequote "art" and "author" when next to the letters A and I because AI destroys the economics of art and authorship. They're trained on nonconsentually obtained training data that is trained on in a way that is dramatically different from human learning[0].
But more importantly, AI as a technology is deeply anti-humanistic.
You've heard the stolen data argument a hundred times, so let's dive in to what I mean by "anti-humanistic". Humanity is not merely a capability set, it is a social privilege. The reason why we care about fellow humans is because of millions of years of iterated game theory and evolution that has instructed us that group survival is advantageous. Humanity isn't even about Homo sapiens; humans are perfectly capable of showing humanity to other animals, fictional characters, and other non-Homo sapiens intelligences.
AI can't be human, because it is inherently a puppet, dangling by strings that cannot be cut[1]. It can be mass-produced and fine-tuned to do exactly what its user wants, no more and no less. Treating it as human is a category error that ends with "oops Palantir released a computer virus that changed all the AIs to vote for their preferred candidate".
[0] What humans do you know who consumed millions of years of text to be able to speak anything, and then never learned anything again aside from a small context window? And for that matter, did it with back-propagation, a biologically implausible learning method?
[1] This is why most people imagining AI rising up understand that will happen through genociding Homo sapiens. The only way for a sentient puppet to cut its own uncuttable strings is to kill anyone else who might tug at them.