what have you got now?
Totally agreed. Systemically consequential implementation of AI algorithms which were trained on overfit datasets, EG disproportionately representative of those effected by these algorithms, would be and likely already are a problem. Less because the engineers are malicious in their bias, but more that these things are likely often designed without consulting those with more humanities-based expertise.we need to pay more attention to how AI is already manifesting in network effects and access to knowledge at a gigantic scale, and how those effects are already acute in some spaces and also quite absent in others
greaber has an excellent passage about the often contradictory nature of "rationalist" policy. let me see if i can find itarguably the craziest modern Internet cult of them all to me is ultra-rationalists
sometimes you get effective altruism
sometimes you get eugenics and Nick Land type bullshit
sometimes from the same people
the idea that you can rationalize every decision if you just try to hard enough is fucking crazy