Somewhat stating the obvious, but training sets, and lack of data audit trails, are the next big IP brouhaha. Unlike home taping not seriously impacting the music or film industry, everything is going to be laundered and remixed with even less attention, attribution, and credit. I'm not sure the extent to which individuals can fight this one. The productivity difference makes AI assistance inevitable but the idea of any public facing IP being rewritten without attribution (maybe as derivatives from other AI processed data sets) strikes me as unfair and subject to error propagation if the original source material is unduly trusted. I wonder about inaccuracies (and/or) errors in old academic text books and whether they'll be found quicker or slower in the text books we are using now, if they and derivative works are used in training sets. Or will we get more sceptical, which seems like a potential positive response.

Feeling lemony as a result of tinkering.