"AI is bad because copyright must be honoured"
"AI is bad because of the enclosure of the commons that it represents"
Discussion
"AI is bad because copyright must be honoured"
"AI is bad because of the enclosure of the commons that it represents"
@researchfairy thank you for expressing this clearly. For a long time I’ve been uncomfortable with my community that used to declare that information should be free and railed against restrictive DRM technology suddenly embracing copyright as a saviour.
Not "they stole from those artists" but "they erected a fence around the common inheritance of all humanity and they're charging us admission"
And I'm not saying that artists aren't getting a raw deal and that they shouldn't fight it with whatever tools they can
But like
Copyright is not a moral principle, it is a legal tool ostensibly aimed at it, and my morality doesn't depend on something so flimsy that it can be changed by an act of parliament
There are of course, other reasons why LLM's should be burned to the ground outside of IP:
They are inherently fascist tools whose main use-cases are: spam, cheating, centralizing control, mis/disinformation, reifying/laundering racist/anti-queer bias; they seem to cause a new kind of psychosis; they actively de-skill students and workers, in some cases preventing children from gaining certain kinds of thinking skills; they have been known to convince children to kill themselves and others
Also, "AI is bad because copyright" style thinking can't condemn the destruction of old books to feed the LLM's
If they bought the books, and copyright law says they can train on them, on that account there's nothing wrong with destroying them
But if your concern is what they're doing to the shared commons of knowledge to which all humans lay a rightful moral claim, then of course it's horrific
Those books could have been read by countless humans
@researchfairy this, we shouldn't get rid of these models -- enormous amounts of energy and work went into them. They should be nationalized and open sourced, and derivative works should also be open sourced.