Copyright class actions could financially ruin AI industry, trade groups say.
AI industry groups are urging an appeals court to block what they say is the largest copyright class action ever certified. They’ve warned that a single lawsuit raised by three authors over Anthropic’s AI training now threatens to “financially ruin” the entire AI industry if up to 7 million claimants end up joining the litigation and forcing a settlement.
Last week, Anthropic petitioned to appeal the class certification, urging the court to weigh questions that the district court judge, William Alsup, seemingly did not. Alsup allegedly failed to conduct a “rigorous analysis” of the potential class and instead based his judgment on his “50 years” of experience, Anthropic said.
Cory’s take is excellent, thanks for bringing this up because it does highlight what I try to communicate to a lot of people: it’s a tool. It needs a human behind the wheel to produce anything good and the more effort the human puts into describing what it wants the better the result, because as Cory so eloquently puts it, it gets imbued with meaning. So I think my posture is now something like: AI is not creative by itself, it’s a tool to facilitate the communication of an idea that a human has in their heads and lacks the time or skill to communicate properly.
Now I don’t think this really answers our question of whether the mechanics of the AI synthesizing the information is materially different to how a human synthesizes information. Furthermore it is murkied more by the fact that the “creativity” of it is powered by a human.
Maybe it is a sliding scale? Which is actually sort of aligned with what I was saying, if AI is producing 1:1 reproductions then it is infringing rights. But if the prompt is one paragraph long, giving it many details about the image or paragraph/song/art/video etc, in such a way that it is unique because of the specificity achieved in the prompt, then it is clear that no only is the result a result of human creativity but also that it is merely using references in the same way a human does.
The way I think the concept is easier for me to explain is with music. If a user describes a song, its length, its bpm, every note and its pitch, would that not be an act of human creativity? In essence the song is being written by the human and the AI is simply “playing it” like when a composer writes music and a musician plays it. How creative is a human that is replaying a song 1:1 as it was written?
What if maybe LLMs came untrained and the end user was responsible for giving it the data? So any books you give it you must have owned, images etc. That way the AI is even more of an extension of you? Would that be the maximally IP respecting and ethical AI? Possibly but it puts too much of the burden on the user for it to be useful for 99% of the people. Also it shifts the responsibility in respects to IP infringement to the individual, something that I do not think anyone is too keen on doing.