Hacker News Reader

By and large, the original artists did not help train the systems.

Engineers figured out how to scrape sites, decode and filter tags, classify images, and feed them into the training dataset.

All the original artist did was publish an image online to be seen by other people.

It'd be like Google compensating people for putting things online that the search engine links to.

View on HN

Google provides value to the sites it links to by sending them traffic, how does an AI company provide value to the artist by using their work to train their systems? I don't see how your analogy works.

if you want an isolated example, there are several digital artists i had never heard of before stable diffusion started appearing in headlines and articles like this. out of curiosity, i installed sd and ran some prompts to see how well it actually emulated their styles. i was impressed enough that i ended up buying a print from one of the artists

Reply on news.ycombinator.com