Thomas Dickerson @elfprince13@mumak.app

@dan where are the sidekiq queues though

@AJSWritesthings if we were actually waiting on them we’d have overlapping 2 year long Christmases 👀

@AJSWritesthings for today anyway

Amash is chilling on the house floor rn 👀

@Klutzleo @shriramk is even on Mastodon too

@rechelon huh, the only thing I ever heard about his professional conduct was, iirc, an open letter from a female grad student saying something to the effect of “my advisor is a lovely human please stop telling him to kill himself” but I can’t actually find it with Google now to confirm

@rechelon is this beef the whole Scott Aaronson/Amanda Marcotte/Arthur Chu clusterfuck, or is there some other reason people are mad at him that I've missed?

@rechelon third because it doesn't follow that all computationally unbounded consciousnesses are going to arrive at the same set of values (because the values guide the optimization, not the other way around)

@rechelon Second, because computational complexity matters! Questions about what to do with an unbounded (or even just exponentially bounded) computational resources are never going to be a good way to inform the behavior of polynomially bounded consciousness.

@rechelon As much as I love the categorical imperative, this doesn't feel like a useful framing for a few reasons. First off, with either a many-worlds or pilot-wave interpretation all of the possible states the future could take *meaningfully already exist*, and whatever rhetorical games you want to play with "choice" are irrelevant - the outcomes are already determined (and with other interpretations there doesn't seem to be any reason to stake out a position against libertarian free will).

@rechelon (3) If you're serious about viewing these problems through a computational lens, Scott Aaronson's "Ghost in the Quantum Turing Machine" should be required reading (as well as, to a lesser extent, his "Why Philosophers Should Care About Computational Complexity")

@rechelon
(2) there's not remotely enough technical detail in that description for me to evaluate your viewpoint but data compression and optimization algorithms that beat local gradient descent are both pretty well understood computationally and I'm not sure how they factor into "choice". Do Simulated Annealing optimizers have "choice"? Do Polynomial-Time Approximation Schemes? How do you even analyze combinatorial optimization in this framework, when typically no gradients exist?

@rechelon resurrecting this conversation after being mostly offline for a bit, with a few thoughts.
(1) one of the reasons I enjoy your writing (even when I disagree with it) is the frequency with which you raise interesting moral questions. In a purely mechanistic non-libertarian universe, making moral value-judgements seems an odd exercise (although obviously I can't fault you for this inconsistency if I accept the hypothesis that your particles made you do it)

@annaleen Vermont has extremely lax laws about public nudity, which inevitably get rediscovered every few summers by bored teens hoping to stir up drama in whatever random town, and there's always a big fuss in which town governments start debating an emergency ordinance and then October rolls around and the problem takes care of itself

@andrewliptak oh cool you presented at Session 5 next to my dad!

@KevinCarson1 bit of a slow-burn setup, but overall a great watch. definitely recommend watching Knives Out first if you haven't already.

@regehr this looks so cursed, I can't wait to watch it.