Liquid Communism posted:

The real stall is that there isn't an infinite corpus to feed the models, and they've already pulled the low-hanging fruit of an internet scrape and started getting legal action in response. I've been quite happy to see more ways start being proposed to salt images to make them actively harmful to generative AI if included in datasets as well.

I am all for grifters wasting huge amounts of time and money only to learn that they should have just hired someone who went to art school for $40 an hour to do work for hire.

Edit: Same as I will never stop laughing when obvious bad code gets implemented because 'ChatGPT says it works' and they have to hire an actual software engineer at consulting rates to unfuck themselves.