New MIT paper finds that you can can delete an artist from an AI model's training data & nothing changes.
CSAIL engineers also observed that it's hard to trace AI-generated images to specific data. Their work raises new questions about AI copyright: bit.ly/3U6ZncH
MIT CSAIL on X: "New MIT paper finds that you can can delete an artist from an AI model's training data & nothing changes. CSAIL engineers also observed that it's hard to trace AI-generated images to specific data. Their work raises new questions about AI copyright: https://t.co/FvvWJpXMHO"
But isn't this the inverse of the Sorites paradox? You can delete one artist from the training data and nothing changes, but if you delete all artists and then you have nothing. This shows that what emerges depends on all the artists, not on just one, no?
Have they tried deleting all of the artists work? Or just one art piece? They should re-do the research and delete more than just one artist. Ideally all of them.
If nothing changes then one did not "delete the artist" they merely removed some bits but not enough.
