|
|
|
|
|
by __d
510 days ago
|
|
I’m interested in the plagiarism perspective. I feel like LLMs are not too dissimilar to humans. We listen to a lifetime of music, read text, watch videos, etc. and when we come to create something all of that influences what we produce. Like if you’ve listened largely to western music, and you look for a note to complete a provided two-note sequence, your choice is informed by that listening history. A non-western trained person is likely to pick a different note. Similar analogies can be made for eg English phrases, or even topics for songs. There’s clearly a boundary between influenced by and copied. Is it the same for generative AI as it is for humans? |
|
You can do it, sure. But you'll probably also start to wonder why nobody really wants to listen to it, and you can count me out before I do.