I went to St John’s College for undergrad, a small dissident college which was the only liberal arts & sciences college I could find that seemed mostly unironic in how it promoted itself & approached the curriculum. They didn’t offer merit scholarships but my grandfather was happy to pay for it.
I worked for Fannie Mae doing “credit risk analytics” after college, was promoted a couple times, ultimately to manager, while getting my MS in Math & Statistics from Georgetown, a top 25 university. I was key to standardizing mortgage data to a level that enabled high quality hedonic modeling of home prices; I think I’m on one of the related patents. I quit because it didn’t seem like I was doing anything to help anyone.
Then I went to work for GiveWell/OPP, and kept having my probationary period extended instead of being hired or fired because I wasn’t doing a bad job in any articulable way but somehow wasn’t giving them what they wanted.
I took a few years off to think about the situation, and now I’m on salary at a small research nonprofit, due to goodwill from richer friends from various kinds of work I couldn’t be paid for.
It’s not that I can’t get rewarded, but that I can’t get rewarded directly for contributing something good with integrity, which is demoralizing and confusing when mostly people are talking in a jargon that collectively invalidates that criticism despite obviously many of them believing it much of the time.
This way of organizing society seems to me to be obviously related to AI misalignment. A team would have difficulty coherently trying to develop aligned AI if it’s constituted the way this process constitutes elites.
As for alternate elites, I’d be interested in helping constitute one with an explicit theory of this class of problem, with a more overt acknowledgement of adversariality than the usual “Goodhart’s Law” deflection. Anything less and we just get OpenAI level slippage again.
I went to St John’s College for undergrad, a small dissident college which was the only liberal arts & sciences college I could find that seemed mostly unironic in how it promoted itself & approached the curriculum. They didn’t offer merit scholarships but my grandfather was happy to pay for it.
I worked for Fannie Mae doing “credit risk analytics” after college, was promoted a couple times, ultimately to manager, while getting my MS in Math & Statistics from Georgetown, a top 25 university. I was key to standardizing mortgage data to a level that enabled high quality hedonic modeling of home prices; I think I’m on one of the related patents. I quit because it didn’t seem like I was doing anything to help anyone.
Then I went to work for GiveWell/OPP, and kept having my probationary period extended instead of being hired or fired because I wasn’t doing a bad job in any articulable way but somehow wasn’t giving them what they wanted. I took a few years off to think about the situation, and now I’m on salary at a small research nonprofit, due to goodwill from richer friends from various kinds of work I couldn’t be paid for.
It’s not that I can’t get rewarded, but that I can’t get rewarded directly for contributing something good with integrity, which is demoralizing and confusing when mostly people are talking in a jargon that collectively invalidates that criticism despite obviously many of them believing it much of the time.
This way of organizing society seems to me to be obviously related to AI misalignment. A team would have difficulty coherently trying to develop aligned AI if it’s constituted the way this process constitutes elites.
As for alternate elites, I’d be interested in helping constitute one with an explicit theory of this class of problem, with a more overt acknowledgement of adversariality than the usual “Goodhart’s Law” deflection. Anything less and we just get OpenAI level slippage again.