Discussion about this post

User's avatar
Enon's avatar

The article needed to be read and edited by a human before being posted, but aside from AI tics such as "deliverable" as a noun, several instances of "load bearing" and a [VERIFY:...] note in the footnotes, as well as needing to be tightened up bu about 40% of length, as well as progressive-lie alignment issues such as climate change shoehorned in and an implicit belief in human universalism rather than competing genetic group interests, I agree with the general thrust of the post.

The biggest missing or at least underemphasized feature is, oddly, the importance of intelligence in addition to alignment--usually intelligence is overemphasized while alignment ignored. The Bostrom paper cite does explicitly note that these two dimensions are nearly independent. Even if humans were aligned (the central ask of the post--which they can't be, because of differing group interests), humans will mostly get all hard problems wrong (by definition of "hard problem" as one most can't answer correcrly). Intelligence is measured by the difficulty of problems solved and difficulty by the intelligence required (bootstrapping from a random sample of a given population). So democracy, if it were even a thing--which, as you note, it isn't--would get every hard problem wrong.

The actual system decides based on demand in the economic sense, ability and willingness to pay, which in practice gets better answers than average-intelligence ignorant disinterested judgments, but often demand is short-sighted and wrong, costing more longer term than it gained short term. Then there is a demand to socialize or shift the costs of this stupidity, which is the main problem of the current system.

So the central issues are not just "aligned with whom?", i.e. "better or worse for whom?", but "is the answer or planned action right?", and how do we accurately answer these questions. The first question is answered by the group affiliations of those proposing and their track record of either group loyalty (right and non-White) or favoring other groups (baizuo, White left). The second is answered by item response theory intelligence-testing methods across an aligned group together with auxillary objective questions determining associated knowledge and whether intelligence is actually being applied to the question by individual members of the group.

Roslyn Reid's avatar

Follow the money & you can never go wrong.

2 more comments...

Ready for more?