That stat about AI art training data really got me thinking

I was reading some article from last month that said over 80% of the training images for popular AI art models come from personal blogs and small artist portfolios, not big stock sites. That blew my mind because I always assumed they scraped from places like DeviantArt or Pinterest mostly. It makes me wonder if those tiny blog artists even know their work is being used. Has anyone else looked into where the training data actually comes from?

3 comments

3 Comments

martin.riley1mo ago

Is it really that big of a deal if nobody even noticed?

phoenix_bailey1mo ago

I mean yeah but that's kind of the point isn't it? Who gets to decide what counts as "noticed"? Just because nobody called you out doesn't mean nobody saw it or that it didn't bother them. Small artists who post their work online often feel like they don't have a voice anyway, so why would they speak up when they're up against a giant tech company?

jamie8041mo ago

Question whether it really matters though lol. Most of those tiny blog artists probably wouldn't care or even notice their stuff got scraped for training a model.