Yes, the output of generative networks like ChatGPT and Dall-E is plagiaristic. It is, ultimately, reusing someone's work without so much as attribution, and without any original contribution. Maybe you can't make that case for all possible outputs, but certainly for a lot of typical output.
Reminder: These devices are statistical models that produce outputs that are probabilistically likely (according to the model) as a match to a given prompt. They frankly barely rise to the level of 'machine' let alone 'intelligence', and it's only through gross abuse of processing power and a healthy dose of apophenia that their output seems to mean anything at all.
If you ask ChatGPT a factual question, on those occasions where it does regurgitate an actual real answer, it's getting that information from somewhere. It's using the work of someone who did the actual legwork of putting that information online, and not attributing it. Yes, an individual instance of that is negligible, but it's tremendously poisonous to the information environment that this thing is just slurping up the work of reporting or archiving information and spitting it out in a way that doesn't even point towards the source.
Say you ask Dall-E to generate an image of a specific subject, like say His Holiness Pope Francis. Dall-E only knows what Pope Francis looks like based on images of Pope Francis that were produced by people. Actual photographers had to go out into the actual world with their actual cameras and take shots of the actual pope so that Dall-E can smear their work into nonsense garbage. This is, again, plagiaristic.
And when people say "oh so you think it should be forbidden to take inspiration from other artists?", I want to scream. It's not 'taking inspiration'. Your anthropomorphization of a spreadsheet isn't an argument. It can't be 'inspired' because it's not a person. The user also cannot be 'inspired' because they haven't even seen the source images that are being munged, they're just being handed a ready-made object to use.
If you break into my house and steal a kilo of bananas and I come to you and find you neck deep in a banana smoothie bender, neither of these assertions are valid arguments:
- "Bro point out where your bananas are to me. You can't find an individual banana in my smoothie."
- "Bro I was just inspired by your bananas."
There is no reasonable use for this kind of technology. Their output is just an informational and aesthetic pollutant. They shouldn't exist. And I think that in a political environment where powerful people (and lots and lots of powerless idiots) want to push this garbage as hard as it will go, it's actually politically necessary to adopt a posture of total rejection and not go out of our way to rhetorically soften how bad this stuff is or try to have a nuanced conversation about how it could be used ethically in different circumstances. I bet there's probably also some ethical and safe ways to use asbestos. I don't think I much care. I want this bathwater out of my house and I don't care how much baby goes with it.
This is why, by the way, I won't even repost "ha ha look how bad the LLM is" type material. I don't even want people to perceive this stuff as accidentally funny. It's not funny to me. It's just gross.
Like, this is not a controversial stance to take: some tech is just a bad idea and shouldn't be deployed. Asbestos home insulation. Fracking. Browser fingerprinting. Nuclear weapons. The Rollie egg cooker. You just disagree with me on whether this specific tech is bad and I am begging you to listen to me that it is very, very bad.
