Hacker News new | past | comments | ask | show | jobs | submit login
OpenAI's DALL-E will train on Shutterstock's library for six more years (theverge.com)
2 points by benkarst 3 months ago | hide | past | favorite | 7 comments



How is it even remotely legal for Shutterstock to cash in on something that photographers never gave consent for? OpenAI is mindblowingly unethical.

Shutterstock was started in 2003.


Your notions of ethics are obsolete. You can’t stop ai. And yes it’s bad news. But also it’s inevitable.


Our legal system's idea of intellectual property wasn't designed for and had no way of knowing what it means to use a copyrighted work as 'training data'. Sadly it still has no protections and even there were, it's no clear how it would be enforced.


Precisely. Our legal system has two issues. First, the definitions are imprecise, non-objective and tethered to historical meaning that is no longer relevant. And second, law you can't enforce is nothing at all, even if defined perfectly.

AI is working on the scale of seconds, and nanoseconds. The legal system is working on the scale of years. It's too slow and dysfunctional even for human matters. But for AI it's hopeless. It's like a tree trying to catch up to a running cheetah.


What do you mean, never gave consent for? I helped a friend put their photos on there, and they most definitely did consent to practically anything.


I mean they never gave consent for their photos to be used as training data.


They did. Read the ToS.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: