LessWrong 2.0 Reader

Thank you for this, I still read it periodically.

Does "highest status" here mean highest expertise in a domain generally agreed by people in that domain, and/or education level, and/or privileged schools, and/or from more economically powerful countries etc?

I mean, functionally all of those things.  (Well, minus the country dynamic.  Everyone at this event I talked to was US, UK, or Canadian, which is all sorta one team for purposes of status dynamics at that event)

Thanks that does help clarify the challenges for me.

[Some thoughts that are similar but different to my previous comment;]

I suspect you can often just prove the behavioral selection theorem and structural selection theorem in separate, almost independent steps.

1. Prove a behavioral theorem
2. add in a structural assumption
3. prove that behavioral result plus structural assumption implies structural result.

Behavior essentially serves as an "interface", and a given behavior can be implemented by any number of different structures. So it would make sense that you need to prove something about structure separately (and that you can prove it for multiple different types of structural assumption).

Further claims: for any given structural class,

• there will be a natural simplicity measure
• simpler instances will be exponentially rare.

A structural class is something like programs, or Markov chains, or structural causal models. The point of specifying structure is to in some way model how the system might actually be shaped in real life. So it seems to me that any of these will be specified with a finite string over a finite alphabet. This comes with the natural simplicity measure of the length of the specification string, and there are exponentially fewer short strings than long ones.^[1]

So let's say you want to prove that your thing X which has behavior B has specific structure S. Since structure S has a fixed description length, you almost automatically know that it's exponentially less likely for X to be one of the infinitely many structures with description length longer than S. (Something similar holds for being within delta of S) The remaining issue is whether there are any other secret structures that are shorter than S (or of similar length) that X could be instead.

1. ^{^}

   Technically, you could have a subset of strings that didn't grow exponentially. For example, you could, for some reason, decide to specify your Markov chains using only strings of zeros. That would grow linearly rather than exponentially. But this is clearly a less natural specification method.

I was just scrolling through Metaculus and its predictions for the US Elections. I noticed that pretty much every case was a conditional If Trump wins/If doesn't win. Had two thought about the estimates for these. All seem to suggest the outcomes are worse under Trump. But that assessment of the outcome being worse is certainly subject to my own biases, values and preferences. (For example, for US voters is it really a bad outcome if the probability of China attacking Taiwan increases under Trump? I think so but other may well see the costs necessary to reduce the likelihood as high for something that is not something that actually involves the USA.)

So my first though was how much bias should I infer as present in these probability estimates? I'm not sure. But that does relate a bit to my other thought.

In one sense you could naively apply the p, therefore not p is the outcome for the other candidate as only two actually exist. But I think it is also clear that the two probability distributions don't come from the same pool so conceivably you could change the name to Harris and get the exact same estimates.

So I was thinking, what if Metaculus did run the two cases side by side? Would seeing p(Haris) + p(Trump) significantly different than 1 suggest one should have lower confidence in the estimates? I am not sure about that.

What if we see something like p(H) approximately equale to p(T)? does that suggest the selected outcome is poorly chosen as it is largely independant of the elected candidate so the estimates are largely meaninless in terms of election outcomes? I have a stronger sense this is the case.

So my bottome line now is that I should likely not hold a high confidence that the estimates on these outcomes are really meaninful with regards to the election impacts. 

I like this as a thing to think about, but, Shifgrethor [LW · GW] is just a way less easy-to-say or evocative word that "Moloch", "Child of Omelas" or "Dark Forest", alas, so I don't think this particular one will really make it into my repertoir.

I can secondhand lend some affirmation to the newcomb case.

Oh yeah, that's a cool example.

Another solution is illegible-ization/orthogonalization of preferences to the hostile telepath so that you don't overlap in anything they might care about or overpower you with.

You mean something like, look boring to them? Like, I don't care how good Putin is at reading people, I just don't have anything he wants, so I'm safe as long as I keep (apparently) not having anything he wants?

Have to divide by number of airships, which probably makes them less safe than planes, if not cars. I think the difficulty is mostly with having a large surface-area exposed to the wind making the ships difficult to control. (Edit: looking at the list on Wikipedia, this is maybe not totally true. A lot of the crashes seem to be caused by equipment failures too.)

Are those things that you care about working towards?

No, and I don't work on airships and have no plans to do so. I mainly just think it's an interesting demonstration of how weak electrostatic forces can be.

Your 'accidents still happen' link shows:

One airship accident worldwide in the past 5 years, in Brazil.

The last airship accident in the US was in 2017.

The last airship accident fatality anywhere in the world was in 2011 in Germany.

The last airship accident fatality in the US was in 1986.

I think that this compares favorably with very nearly everything.