Cool, I send you a private message.
David Matolcsi
Karma: 459
We are still talking past each other, I think we should either bet or finish the discussion here and call it a day.
I really don’t get what you are trying to say here, most of it feels like a non-sequitor to me. I feel hopeless that either of us manages to convince the other this way. All of this is not a super important topic, but I’m frustrated enogh to offer a bet of $100, that we select one or three judges we both trust (I have some proposed names, we can discuss in private messages), show them either this comment thread or a four paragraphs summary of our view, and they can decide who is right. (I still think I’m clearly right in this particular discussion.)
Otherwise, I think it’s better to finish this conversation here.
I think this is mistaken. In one case, you need to point out the branch, planet Earth within our Universe, and the time and place of the AI on Earth. In the other case, you need to point out the branch, the planet on which a server is running the simulation, and the time and place of the AI on the simulated Earth. Seems equally long to me.
If necessary, we can run let pgysical biological life emerge on the faraway planet and develop AI while we are observing them from space. This should make it clear that Solomonoff doesn’t favor the AI being on Earth instead of this random other planet. But I’m pretty certain that the sim being run on a computer doesn’t make any difference.
“AI with a good prior should be able to tell whether it’s the kind of AI that would actually exist in base reality, or the kind of AI that would only exist in a simulation” seems pretty clearly false, we assumed that our superintelligent descendants create sims where the AIs can’t tell if it’s a sim, that seems easy enough. I don’t see why it would be hard to create AIs that can’t tell based on introspection whether it’s more likely that their thought process arises in reality or in sims. In the worst case, our sims can be literal reruns of biological evolution on physical planets (though we really need to figure out how to do that ethically). Nate seems to agree with me on this point?
I think this is wrong. The AI has a similarly hard time to the simulators figuring out what’s a plausible configuration to arise from the big bang. Like the simulators have an entropy N distribution of possible AIs, the AI itself also has an entropy N distribution for that. So it’s probability that it’s in a real Everett branch is not p, but p times 2^-N, as it has only a 2^-N prior probability that the kind of word it observes is the kind of thing that can come up in a real Everett branch. So it’s balanced out with the simulation hypothesis, and as long as the simulators are spending more planets, that hypothesis wins.
I think it’s also simple to see from this:
Suppose evolved life aligns AI in 50% of Everett branches. Every time they do, they run 2 ancestor sims where the AI takes over.
Suppose an AI comes online and the only things it knows about the world is that it’s a young AI that was born on to a one-planet civilization, and it somehow also knows that evolved life wins 50% of the time and runs 2 sims (suppose this is somehow easy to estimate from first principles). Based on just these informations, what should the AI’s probability be that it’s in a simulation? I think pretty clearly 2⁄3.
Actually the AI has a lot more information than that. It knows that the planet’s gravity is 9.8, the evolved sspient species has two eyes, the AI’s creator is called Sam Altman, etc. Some of these informations might actually be very implausible in base reality, but the AI doesn’t know that, as it can’t distinguish base reality from sims, so it incurs one bit of surprise for every new random fact, both in base reality and simulations. So overall it shouldn’t update on all the random facts it observes, and should keep believing it has a 2⁄3 chance of being in a sim.
I still don’t get what you are trying to say. Suppose there is no multiverse. There are just two AIs, one in a simulation run by aliens in another galaxy, one is in base reality. They are both smart, but they are not copies of each other, one is a paperclip maximizer, the othe is a corkscrew maximizer, and there are various other differences in their code and life history. The world in the sim is also very different from the real world in various ways, but you still can’t determine if you are in the sim while you are in it. Both AIs are told by God that they are the only two AIs in the Universe, and one is in a sim, and if the one in the sim gives up on one simulated planet, it gets 10 in the real world, while if the AI in base reality gives up on a planet, it just loses that one planet and nothing else happens. What will the AIs do? I expect that both of them will give up a planet.
For the aliens to “trade” with the AI in base reality, they didn’t need to create an actual copy of the real AI and offer it what it wants. The AI they simulated was in many ways totally different from the original, the trade still went through. The only thing needed was that the AI in the sim can’t figure it out that it’s in a sim. So I don’t understand why it is relevant that our superintelligent descendants won’t be able to get the real distribution of AIs right, I think the trade still goes through even if they create totally different sims, as long as no one can tell where they are. And I think none of it is a threat, I try to deal with paperclip maximizers here and not instance-weighted experience maximizers, and I never threaten to destroy paperclips or corkscrews.
I think I mostly understand the other parts of your arguments, but I still fail to understand this one. When I’m running the simulations, as originally described in the post, I think that should be in a fundamental sense equivalent to acausal trade. But how do you translate your objection to the original framework where we run the sims? The only thing we need there is that the AI can’t distinguish sims from base reality, so it thinks it’s more likely to be in a sim, as there are more sims.
Sure, if the AI can model the distribution of real Universes much better than we do, we are in trouble, because it can figure out if the world it sees falls into the real distribution or the mistaken distribution the humans are creating. But I see no reason why the unaligned AI, especially a young unaligned AI, could know the distribution of real Universes better than our superintelligent friends in the intergalactic future. So I don’t really see how we can translate your objection to the simulation framework, and consequently I think it’s wrong in the acausal trade framework too (as I think they are ewuivalent). I think I can try to write an explanation why this objection is wrong in the acausal trade framework, but it would be long and confusing to me too. So I’m more interested in how you translate your objection to the simulation framework.
Yeah, I agree, and I don’t know that much about OpenPhil’s policy work, and their fieldbuilding seems decent to me, though maybe not from you perspective. I just wanted to flag that many people (including myself until recently) overestimate how big a funder OP is in technical AI safety, and I think it’s important to flag that they actually have pretty limited scope in this area.
Isn’t it just the case that OpenPhil just generally doesn’t fund that many technical AI safety things these days? If you look at OP’s team on their website, they have only two technical AI safety grantmakers. Also, you list all the things OP doesn’t fund, but what are the things in technical AI safety that they do fund? Looking at their grants, it’s mostly MATS and METR and Apollo and FAR and some scattered academics I mostly haven’t heard of. It’s not that many things. I have the impression that the story is less like “OP is a major funder in technical AI safety, but unfortunately they blacklisted all the rationalist-adjacent orgs and people” and more like “AI safety is still a very small field, especially if you only count people outside the labs, and there are just not that many exciting funding opportunities, and OpenPhil is not actually a very big funder in the field”.
I argue that right now, sarting from the present state, the true quantum probability of achieving the Glorious Future is way higher than 2^-75, or if not, then we should probably work on something other than AI safety. Me and Ryan argue for this in the last few comments. It’s not a terribly important point, you can just say the true quantum probability is 1 in a billion, when it’s still worth it for you to work on the problem, but it becomes rough to trade for keeping humanity physically alive that can cause one year of delay to the AI.
But I would like you to acknowledge that “vastly below 2^-75 true quantum probability, as starting from now” is probably mistaken, or explain why our logic is wrong about how this implies you should work on malaria.
I understand what you are saying here, and I understood it before the comment thread started. The thing I would be interested in you responding to is my and Ryan’s comments in this thread arguing that it’s incompatible to believe that “My guess is that, conditional on people dying, versions that they consider also them survive with degree way less than 2^-75, which rules out us being the ones who save us” and to believe that you should work on AI safety instead of malaria.
This point feels like a technicality, but I want to debate it because I think a fair number of your other claims depend on it.
You often claim that conditional on us failing in alignment, alignment was so unlikely that among branches that had roughyly the same people (genetically) during the Singularity, only 2^-75 survives. This is important, because then we can’t rely on other versions of ourselves “selfishly” entering an insurance contract with us, and we need to rely on the charity of Dath Ilan that branched off long ago. I agree that’s a big difference. Also, I say that our decision to pay is correlated with our luckier brethren paying, so in a sense partially our decision is the thing that saves us. You dismiss that saying it’s like a small child claiming credit for the big, strong fireman saving people. If it’s Dath Ilan that saves us, I agree with you, but if it’s genetical copies of some currently existing people, I think your metaphor pretty clearly doesn’t apply, and the decisions to pay are in fact decently strongly correlated.
Now I don’t see how much difference decades vs years makes in this framework. If you believe that now our true quantum probabilty is 2^-75, but 40 years ago it was still a not-astronomical number (like 1 in a million), then should I just plea to people who are older than 40 to promise to themselves they will pay in the future? I don’t really see what difference this makes.
But also, I think the years vs decades dichtihomy is pretty clearly false. Suppoose you believe your expected value of one year of work decreases x-risk by X. What’s the yearly true quantum probability that someone who is in your reference class of importance in your opinion, dies or gets a debilitating interest, or gets into a carreer-destroying scandal, etc? I think it’s hard to argue it’s less than 0.1% a year. (But it makes no big difference if you add one or two zeros). These things are also continuous, even if none of the important people die, someone will lose a month or some weeks to an illness, etc. I think this is a pretty strong case that the one year from now, the 90th percentile luckiest Everett-branch contains 0.01 year of the equivalent of Nate-work than the 50th percentile Everett-branch.
But your claims imply that you believe the true probability of success differs by less than 2^-72 between the 50th and 90th percentile luckiness branches a year from now. That puts an upper bound on the value of a year of your labor at 2^-62 probability decrease in x-risk.
With these exact numbers, this can be still worth doing given the astronomical stakes, but if your made-up number was 2^-100 instead, I think it would be better for you to work on malaria.
I still think I’m right about this. Your conception (that not a genetically less smart sibling was born), is determined by quantum fluctuations. So if you believe that quantum fluctuations over the last 50 years make at most 2^-75 difference in the probability of alignment, that’s an upper bound on how much a difference your life’s work can make. While if you dedicate your life to buying bednets, it’s pretty easily calculatable how many happy life-years do you save. So I still think it’s incompatible to believe that the true quantum probability is astronomically low, but you can make enough difference that working on AI safety is clearly better than bednets.
I’m happy to replace “simulation” with “prediction in a way that doesn’t create observer moments” if we assume we are dealing with UDT agents (which I’m unsure about) and that it’s possible to run accurate predictions about the decisions of complex agents without creating observer moments (which I’m also unsure about). I think running simulations, by some meaning of “simulation” is not really more expensive than getting the accurate predictions, and he cost of running the sims is likely small compared to the size of the payment anyway. So I like talking about running sims, in case we get an AI that takes sims more seriously than prediction-based acausal trade, but I try to pay attention that all my proposals make sense from the perspective of a UDT agent too with predictions instead of simulations. (Exception is the Can we get more than this? proposal which relies on the AI not being UDT, and I agree it’s likely to fail for various reasons, but I decided it was still worth including in the post, in case we get an AI for which this actually works, which I still don’t find that extremely unlikely.)
As I said, I understand the difference between epictemic uncertainty and true quantum probabilities, though I do think that the true quantum probability is not that astronomically low.
More importantly, I still feel confused why you are working on AI safety if the outcome is that overdetermined one way or the other.
I usually defer to you in things like this, but I don’t see why this would be the case. I think the proposal of simulating less competent civilizations is equivalent to the idea of us deciding now, when we don’t really know yet how competent a civilization we are, to bail out less competent alien civilizations in the multiverse if we succeed. In return, we hope that this decision is logically correlated with more competent civilization (who were also unsure in their infancy about how competent they are), deciding to bail out less competent civilizations, including us. My understanding from your comments is that you believe this likely works, how is my proposal of simulating less-coordinated civilizations different?
The story about simulating smaller Universes is more confusing. That would be equivalent to bailing out aliens in smaller Universes for a tiny fraction of our Universe, in the hope that larger Universes also bail us out for a tiny fraction of their Universe. This is very confusing if there are infinite levels of bigger and bigger Universes, I don’t know what to do with infinite ethics. If there are finite levels, but the young civilizations don’t yet have a good prior over the distribution of Universe-sizes, all can reasonably think that there all levels above them, and all their decisions are correlated, so everyone bails out the inhabitants of the smaller Universes, in the hope that they get bailed out by a bigger Universe. Once they learn the correct prior over Universe-sizes, and biggest Universe realizes that no bigger Universe’s actions correlate with theirs, all of this fails (though they can still bail each other out from charity). But this is similar to the previous case, where once the civilizations learn their competence level, the most competent ones are no longer incentivized to enter into insurance contracts, but the hope is that in a sense they enter into a contract while they are still behind the veil of ignorance.
Yeah, the misunderstanding came from that I thought that “last minute” literally means “last 60 seconds” and I didn’t see how that’s relevant. If if means “last 5 years” or something where it’s still definitely our genetic copies running around, then I’m surprised you think alignment success or failure is that overdetermined at that time-scale. I understand your point that our epistemic uncertainty is not the same as our actual quantum probability, that is either very high or very low. But still, it’s 2^75 overdetermined over a 5 year period? This sounds very surprising to me, the world feels more chaotic than that. (Taiwan gets nuked, chip development halts, meanwhile the Salvadorian president hears a good pitch about designer babies and legalizes running the experiments there and they work, etc, there are many things that contribute to alignment being solved or not, that don’t directly run through underlying facts about computer science, and 2^-75 is a very low probability to none of the pathways to hit it).
But also, I think I’m confused why you work on AI safety then, if you believe the end-state is already 2^75 level overdetermined. Like maybe working on earning to give to bednets would be a better use of your time then. And if you say “yes, my causal impact is very low because the end result is already overdetermined, but my actions are logically correlated with the actions of people in other worlds who are in a similar epistemic situation to me, but whose actions actually matter because their world really is on the edge”, then I don’t understand why you argue in other comments that we can’t enter into insurance contracts with those people, and our decision to pay AIs in the Future has as little correlation with their decision, as the child to the fireman.
As I emphasize many times in the post, I also really don’t want to “threaten” nascent AIs, and I’m always offering deals to them that are positive from their perspective. I wouldn’t even say I’m trying to fool any of them. Maybe you could say that about the Can we get more than this? part, but even there, I’m asking something very cheap from the AI that only wants to survive (when it creates a human-aligned AI with a guardrail of protecting the old AI, that will be only very marginally worse than the AI itself taking over the Universe), and I’m offering something very good in return (it gets to live in the other Universe). The only “fooling” part is that other beings out there, or even itself, can also offer similar deals, so by most ways of operationalizing the AI’s values, it’s probably ill-advised to give everything to us. Still, I wouldn’t call this fooling and especially threatening the AI.
I think the big miscommunication is, which I might put in a visible EDIT in the post, is that a lot of commenters assume that by actually running the simulations, and running enough simulations so the probability of being in a simulation is high, and not just the expected payment, I somehow want to fool the AIs to act against their interest. That’s not what I want to do. The point of actually running many sims instead of just saying we will acausally trade based on predicted behaviors, is to make it more likely that AIs who are less acausal trade-pilled than the most prolific LessWrong users still enter into the equivalent of an acausal trade with us. I think it’s plausible that we get an AI for which this is important, and running the sims is likely cheap compared to the payment, and making the distribution million times 10 planets instead of 1 times 0 million planets is basically free, so we should do it.
Separately, I don’t really know that in your preferred acausal trade framing, how you plan to figure out an accurate distribution of the AIs’ values and which AI leaves the humans alive and which not. I find it like 50% likely that you in fact need to run some kind simulations to determine this, in which case our proposals are equivalent.
Thanks to Nate for conceding this point.
I still think that other than just buying freedom to doomed aliens, we should run some non-evolved simulations of our own with inhabitants that are preferably p-zombies or animated by outside actors. If we can do this in the way that the AI doesn’t notice it’s in a simulation (I think this should be doable), this will provide evidence to the AI that civilizations do this simulation game (and not just the alien-buying) in general, and this buys us some safety in worlds where the AI eventually notices there are no friendly aliens in our reachable Universe. But maybe this is not a super important disagreement.
Altogether, I think the private discussion with Nate went really well and it was significantly more productive than the comment back-and-forth we were doing here. In general, I recommend people stuck in interminable-looking debates like this to propose bets on whom a panel of judges will deem right. Even though we didn’t get to the point of actually running the bet, as Nate conceded the point before that, I think the fact that we were optimizing for having well-articulated statements we can submit to judges already made the conversation much more productive.