Comment by cowboylowrez
3 months ago
it was more a simply reply to
>How can infinite AI content be strictly awful if it forces us to fix issues with our trust and reward systems?
I should have quoted what I just did in my original reply, I feel like I wasted your time by not including it. Still you did post interesting things so not all is lost.
>how can crime be bad if it forces us to police crime?
crime can be bad whether we police it or not. we actually police crime because its bad, at least in societies that are so inclined to have a police force. a desire to reduce somethings occurrence is not speaking positively of such occurrences.
> How can infinite AI content be strictly awful if it forces us to fix issues with our trust and reward systems?
this is neither a disqualifier for being "strictly awful", nor the newly arrived unique event finally necessitating fixes to trust and reward systems. I would hope that we dont evaluate the goodness of AI based on whether we have functioning trust and reward systems.
Fair points, and thanks for clarifying.
Your last point helps me tease out what I think rubs me the wrong way. Another analogy, "these newly introduced, extremely fast cars make it entirely unsafe to drive drunk."
Of course to be fair, we'd have to point out that the purchase, operation and production (and more) of said vehicles has a terrible impact.
I'd just love to hear that we are going to crack down on drunk driving which was even a problem when we were going slower. Obviously, the metaphor falls apart - trust and reward are much more interesting nuts to crack.
It's a really hard point to make because expressing an interest in wanting to see one part of the problem solved seems to indicate to others that I don't care about all the other aspects.
We're not cracking down on LLMs in any meaningful way. They're built on copyright infringement on an unprecedented way. It's the kind of thing where the law is looking at the other way while people's lives are destroyed, and some, if lucky, will be compensated 30 years from now, probably with pitiful amounts of money.
Corruption generally works by inflicting a diluted, distributed harm. Everyone else ends up a bit worse off except for the agent of corruption, which ends up very well off.