News for hackers. Stuff that matters.
Re:The Pile: a dataset for language modeling [pdf] by idiotsecant on Tuesday July 11, 2023 @05:16PM (#36687325)
Is it useful to protect weights with copyright? What if I download your weights and retrain them for 5 seconds, changing each weight .0000001%? How much change is a new product? What if I change a single weight?
Re:The Pile: a dataset for language modeling [pdf] by hedgehog on Tuesday July 11, 2023 @05:23PM (#36687403)
Like the parallel scenarios of taking a book and changing a few words, slapping a new logo on someone else's app, or stylizing a photo with a filter, those are questions that will be answered in court if people can't come to an agreement on their own.