• donuts@kbin.social
    link
    fedilink
    arrow-up
    2
    arrow-down
    1
    ·
    edit-2
    11 months ago

    A) Not true. Many have been training models using various online data that that doesn’t belong to them, has not been licensed, and has been used without informed consent of the rights holders.

    B) Terrible comparison. Music sampling is a grey area that is much more complex and dubious than you’re suggesting. There are instances in which sampling has been considered fair use, but outside of that there are strict laws around sampling. Finally, human music creation and sampling have very little in common with generative AI.

    AI is here to stay. But the free ride of scraping every piece of information in human history without even a basic regard towards intellectual property or personality rights is unsustainable, unethical, and nowhere near the threshold for what can be considered fair use.

    Once people start needing to own or license their training data sets the technology will be just fine, but costs will rise dramatically and the VC investment bubble is going to pop bigtime.