4 ms·
Perhaps there needs to be laws that require AIs to state the exact source of data used for training. This goes for art AIs as well. I feel bad for artists who h
by senttoschool 4y ago
Perhaps there needs to be laws that require AIs to state the exact source of data used for training. This goes for art AIs as well. I feel bad for artists who had their work "stolen" and then the AIs could very well put them out of work or greatly reduce their income.
- superchroma 4y agoThat doesn't resolve the quagmire. If you train an image generator based on scraped youtube video frames (this actually happened), how do you account for a frame which has copyrighted artwork in it? The person behind the camera could be walking around an art gallery and just panning around and you'd capture works you have no right to.
- senttoschool 4y agoIt doesn't. But it will allow the person who copyrighted the work knowledge that a particular AI potentially used the copyrighted material. This person can sue or do nothing about it. Right now, we don't even know what sources these AI used to train these AIs.
- superchroma 4y agoIt will for primary sources (e.g. training an AI with images from Shutterstock; direct link from consumer to artist), but for secondary, unclean sources (e.g. youtube frames) there will still be uncertainty.