Shortform #71 A Restful & Productive Saturday | Brief rough thought about AI Alignment & Politics
Today I slept in, studied for my cert exam, finalized my V3 character sheet, virtually hungout with friends, and had a pleasant time puttering around the house.
My office desk is currently two stacks of three boxes per stack with an old Apple Xserve stretched across them for stability. I used my two existing desks for my setups in the loft to great effect, now I need to focus on finalizing my office setup...so, purchasing three desks is my next step there.
I did not write anything down about Politics & AI Alignment, that seems (mostly) to be an activity for tomorrow. Very briefly though: I believe that according to whose values an AI is aligned towards (provided such alignment is doable; I freely admit that my knowledge of technical AI Alignment is woefully inadequate, something that I intend to remedy), well...that choice of values is necessarily a normative political choice. What world does one wish to live in, and how might one use an AI aligned towards creating such a world to bring such a world about? I don’t think the question of bringing about a friendly aligned AI stops at mere friendliness, I think the question at that point becomes, to whom is the AI friendly and what does that AI do to those in that individual’s “outgroup”? It feels like the human race is not just in an arms race to bring about friendly AI...it’s in an arms race to control the power (superintelligent AI) that could ultimately reshape the world according to their own values.
Shortform #71 A Restful & Productive Saturday | Brief rough thought about AI Alignment & Politics
Today I slept in, studied for my cert exam, finalized my V3 character sheet, virtually hungout with friends, and had a pleasant time puttering around the house.
My office desk is currently two stacks of three boxes per stack with an old Apple Xserve stretched across them for stability. I used my two existing desks for my setups in the loft to great effect, now I need to focus on finalizing my office setup...so, purchasing three desks is my next step there.
I did not write anything down about Politics & AI Alignment, that seems (mostly) to be an activity for tomorrow. Very briefly though: I believe that according to whose values an AI is aligned towards (provided such alignment is doable; I freely admit that my knowledge of technical AI Alignment is woefully inadequate, something that I intend to remedy), well...that choice of values is necessarily a normative political choice. What world does one wish to live in, and how might one use an AI aligned towards creating such a world to bring such a world about? I don’t think the question of bringing about a friendly aligned AI stops at mere friendliness, I think the question at that point becomes, to whom is the AI friendly and what does that AI do to those in that individual’s “outgroup”? It feels like the human race is not just in an arms race to bring about friendly AI...it’s in an arms race to control the power (superintelligent AI) that could ultimately reshape the world according to their own values.
Ooh, adding https://www.lesswrong.com/posts/CQAMdzA4MZEhNRtTp/human-values-and-biases-are-inaccessible-to-the-genome to my reading list for tomorrow / soon. Germane to my interests in politics, AI, and values.