2 years into this mess and I still haven't figured out a reasonable way to talk about the intersection of the AI "ethics" and "utility" arguments without getting completely derailed. the degree to which it "works"[1] does actually matter to the ethics; if it does "cure cancer" all over the globe then sure a bit of copyright infringement is small potatoes. But also the ethics argument feeds back into the utility argument; if it's just plagiarism then it doesn't matter if it really "can code".
0 Boosts
0 Likes
We have to get past the “does it work” to get to the “is it worth the tradeoffs” part of the argument; that’s where we move the needle. Meat tastes good. Agreed! How important is that pleasure in your life? What if you made meat a special occasion thing? Cars are required to live the life I want. True! So how can we mitigate the harm? Could you bike to work when the weather’s nice? And so on. We do this all the time everywhere, why not with AI? (2/2)
@jacob the more I stare at this argument the harder I find it to parse, because it seems to take accelerationsim as a given? I’m probably very confused 😅
@glyph But, like, this is true of almost everything. Cars are bad for myriad of reasons yet we use them because of their utility in our lives. Eating meat harms the environment (much more than LLMs btw) but we do it for pleasure (and in some cases for health). Shopping at Walmart harms workers but sometimes it’s the only store in town. Utility absolutely matters for ethics - we’re constantly trading off the “right” thing for the useful/convienient/pleasurable/etc thing. (1/2)