leogao's Shortform

leogao

This is a special post for quick takes by leogao. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.

what's the strongest argument for why i shouldn't auto-ignore any acausal arguments that involve hypothetical entities extremely far away (or which only exist in other Everett branches or whatever) such that we will never interact with them causally at all? a razor i have is things which are entirely epiphenomenal should be ignored because they are unfalsifiable.

in particular, this seems consistent even if you accept one-boxing and paying in the counterfactual mugging. the key question is what kinds of evidence you accept as evidence of the existence of an acausal connection. in these hypotheticals, we simply declare by assumption that Omega is truthful and capable of predicting you. in reality, we would either arrive at such a belief by empirical observation (Omega has a strong track record), or pure theoretical deduction. all of the epiphenomenal acausal theories depend on pure deduction. that empirical observation has to be conveyed to us causally. it seems reasonable to draw a line and say we simply don't trust pure deduction to be able to convince us of an acausal link.

also, with both the counterfactual mugging and Newcomb, even though in each instance you can't prove the other branch could have happened, in the long run those who do the right thing (pay, one box) will win. whereas with purely epiphenomenal acausal theories you will literally never find out any difference whatsoever, because the branching point happened long ago.

13Linch6mo

The story I'm most sympathetic for acausal trade[1] to look something like this: 1. Several factions on Earth tried pretty hard to hash out their differences for what to do with space probes but end up deciding that they can't. 2. Rather that put compromise solutions on the probes, and rather than go to war, they decided to "agree to disagree" and split fractions of space according to the values of different factions.[2] The key agents on the different probes are tasked to satisfy their respective factions' values. 3. Once the probes arrive, they "unfold" and after an initial buildup, a key intermediate stage is turning a lot of stellar mass and energy into computrionium, so they can do more calculations than Sol's supercomputers were ever able to. The calculations are initially done for entirely instrumental reasons (to figure out the best approach to maximize/satisfy the values loaded into the relevant agents) 4. One aspect of interest to these probes are what the other agents are doing. They figure this out via simulation. 5. Meanwhile, perhaps these agents learn highly useful empirical information about the respective regions of the other probes, from light transmissions in galaxies within your past but not future (due to space expansion) lightcone. 6. 4+5 lead many agents to mutually conclude that their preferences will be better satisfied if they swapped places. 7. At this point, the agents acausally "trade" (in the sense of doing what another faction wants) iff their simulations suggest (with sufficiently high probability) that the other agents will do what they want conditional upon trading. In such an overly specific, somewhat convoluted setup, I think refusing to consider acausal trade will be obviously foolish. Maybe you think this setup unreasonably stacks the deck towards acausal trade. I agree! I think it's plausible enough that simulation limits means trading across multiverses isn't viable. But I wanted to present the example above as a kin

2Fabien Roger6mo

My understanding of your argument is "we will have only very weak evidence of beneficial acausal effects". I agree this is probably true now. (Understanding "evidence" is the Bayesian sense.) But I don't think this would be the case in situations like the one I describe above with the ASI simulating other civilizations? The ASI might be uncertain (though it's unclear, maybe with good enough intelligence you could be quite confident in this kind of thing), but probably not radically uncertain. It doesn't seem to me like a different kind of uncertainty to the one I have when I donate to a medium-to-high-risk high-impact opportunity / when I try to help the world by working on AI safety. (Uncertainty about morality (do you care about goodness produced by distant civilizations) and decision theory (are you EDT-ish) might not be resolved by intelligence and you thus might still have a kind of uncertainty different from empirical uncertainty about the effects of your action. In my comment above I conditioned that uncertainty away. I think it's unclear how to act when you have this kind of uncertainty, but I'd be surprised if the answer was "auto-ignore".) Maybe your argument is more like "you should only use causal evidence when making decisions" (as a possible update to decision theory, not as an empirical claim about what Bayesian evidence is weak vs strong)? I don't think there is a way of making this sentence more precise that you would result in a decision theory you would find reasonable.

if god comes down from the sky and tells you that there exists a beneficial acausal trade, sure, you should take it. ditto for a sufficiently competent ASI. i'm mostly making an empirical claim that you are very unlikely to gain the requisite level of confidence in practice as a human, because humans are simply not capable of reasoning about this in a sensible way, and so we should not really spend time thinking about it today. and it also isn't worth planning ahead for what to do if the ASI recommends making acausal trades in the future, because our plans... (read more)

leogao's Shortform

8

Ω 4

leogao's Shortform

8

Ω 4