Paperclip Director's Commentary, Milestone celebration part 2!!
Added 2020-05-10 20:10:40 +0000 UTCShout-out to all the moms, am I right folks? Anyway please enjoy this commentary on an old favorite.
Comments
You have fantastic taste! (It's my favorite as well)
Cody Fox
2020-05-15 12:08:03 +0000 UTCAny plans for a legacy of the haunted house commentary? That is quite possibly my favorite YouTube video ever : )
and
2020-05-13 14:07:54 +0000 UTCScott Alexander names this idea Moloch in an essay framed around Allen Ginsburg's poem *Howl*: "In some competition optimizing for X, the opportunity arises to throw some other value under the bus for improved X. Those who take it prosper. Those who don’t take it die out. Eventually, everyone’s relative status is about the same as before, but everyone’s absolute status is worse than before. The process continues until all other values that can be traded off have been – in other words, until human ingenuity cannot possibly figure out a way to make things any worse." His examples include the two-income trap, the race to the bottom, arms races, the Malthusian trap, and, indeed, capitalism. He also mentions the paperclip thing briefly, in the context of misaligned AI: same idea, except now the tradeoff continues well beyond the normal limits of human ingenuity. It's a really good essay. It's long, but one of my favorites. If anyone is interested, here's the link: https://slatestarcodex.com/2014/07/30/meditations-on-moloch/
Chas Evans
2020-05-12 14:12:19 +0000 UTCBadly aligned goals is basically it. Universal Paperclips is about a single-minded pursuit of a goal, deemed important for no other reason than it's profitable, at the expense of catastrophic environmental and human consequences. That is not a story inherent to AI at all; it might as well be about Exxon, except manufacturing barrels of oil is way more on-the-nose than paperclips.
Jacob Geller
2020-05-11 00:32:18 +0000 UTCRe Universal paperclips and capitalism: as a computer scientist, I have a pretty different perspective on this. I've always heard / considered the paperclip collector idea, as used by Nick Bostrom, to mainly be about AI alignment / goal specification, which is a very real challenge in this field (see eg Robert Miles' recent video for real current examples: https://youtu.be/nKJlF-olKmg and most of his other videos are related topics too). The closest links to capitalism I can see are the incentives of capitalism potentially encouraging unsafe approaches to AI research, or that you could argue capitalism is also a system with badly aligned goals compared to actual human interests. But I'm curious, what exactly is your perspective on this? :)
Péter Mernyei
2020-05-10 22:43:33 +0000 UTC
