Well, to be exact, your formulation of this problem has pretty much left this counterfactual entirely undefined. Naive approximation, that the world is just like ours, and Omega just lies in counterfactual, would not contain such weird calculators which give you wrong answers. If you want to complicate problem by saying that some specific class of agents have a special class of calculators that one would usually think to work in certain way, but actually they work in a different way, well, so be it. That’s however just a free-floating parameter you have left unspecified and that, unless stated otherwise, should be assumed not to be the case.
Hmm, no, I assumed that Omega would be using logical counterfactuals, which are pretty much the topic of the post. In logical counterfactuals, all calculators behave differently ;-) But judging from the number of people asking questions similar to yours, maybe it wasn’t a very transparent assumption...
I asked about these differences in my second post in this post tree, where I explained how I understood these counterfactuals to work. I explained as clearly as I could that, for example, calculators should work as they do in real world. I did this explaining in hopes of someone voicing disagreement if I had misunderstood how these logical counterfactuals work.
However, modifying any calculator would mean that there can not be, in principle, any “smart” enough ai or agent that could detect it was in counterfactual. Our mental hardware that checks if logical coin should’ve been heads or tails is a calculator the same as any computer, and again, there does not seem to be any reason to assume Omega leaves some calculators unchanged while changes results of others.
Unless, this thing is just assumed to happen, with some silently assumed cutaway point where calculators become so internal they are left unmodified.
Calculators are not modified, they are just interpreted differently, so that when trying to answer the question of what happens in a certain situation (containing certain calculators etc.) we get different answers depending on what the assumptions are. The situation is the same, but the (simplifying) assumptions about it are different, and so simplified inferences about it are different as well. In some cases simplification is unavoidable, so that dependence of conclusions on assumptions becomes an essential feature.
My current understanding of logical counterfactuals is something like this: if the inconsistent formal theory PA+”the trillionth digit of pi is odd” has a short proof that the agent will take some action, which is much shorter than the proof in PA that the trillionth digit of pi is in fact even, then I say that the agent takes that action in that logical counterfactual.
Note that this definition leads to only one possible counterfactual action, because two different counterfactual actions with short proofs would lead to a short proof by contradiction that the digit of pi is odd, which by assumption doesn’t exist. Also note that the logical counterfactual affects all calculator-like things automatically, whether they are inside or outside the agent.
That’s an approximate definition that falls apart in edge cases, the post tries to make it slightly more exact.
(Btw, I think it should be mentioned that a central piece of motivation for this “logical counterfactuals” thing is that it’s probably the same construction that’s needed to evaluate possible actions in normal cases, without any contrived coins, for an agent that knows its own program. So for example although a counterfactual scenario can’t easily “lead” to two different actions, two different actions in that scenario can still be considered as possibly even more (easily shown to be) contradictory “logical counterfactuals” that include additional assumptions about what the action is.)
Try as I might, I cannot find any reference to what’s canonical way of building such counterfactual scenarios. Closest I could get was in http://lesswrong.com/lw/179/counterfactual_mugging_and_logical_uncertainty/ , where Vladimir Nesov seems to simply reduce logical uncertainty to ordinary uncertainty, but this does not seem to have anything to do with building formal theories and proving actions or any such thing.
To me, it seems largely arbitrary how agent should do when faced with such a dilemma, all dependent on actually specifying what it means to test a logical counterfactual. If you don’t specify what it means, whatever could happen as a result.
I am not sure there is a clean story yet on logical counterfactuals. Speaking for myself only, I am not yet convinced logical counterfactuals are “the right approach.”
I am not yet convinced logical counterfactuals are “the right approach.”
Me neither. Have you seen my post about common mistakes? To me it seems more productive and more fun to explore the implications of an idea without worrying if it’s the right approach.
Note that the agent is not necessarily able to detect that it’s in a counterfactual, see Nesov’s comment.
Yes, those agents you termed “stupid” in your post, right?
The smart ones too, I think. If you have a powerful calculator and you’re in a counterfactual, the calculator will give you the wrong answer.
Well, to be exact, your formulation of this problem has pretty much left this counterfactual entirely undefined. Naive approximation, that the world is just like ours, and Omega just lies in counterfactual, would not contain such weird calculators which give you wrong answers. If you want to complicate problem by saying that some specific class of agents have a special class of calculators that one would usually think to work in certain way, but actually they work in a different way, well, so be it. That’s however just a free-floating parameter you have left unspecified and that, unless stated otherwise, should be assumed not to be the case.
Hmm, no, I assumed that Omega would be using logical counterfactuals, which are pretty much the topic of the post. In logical counterfactuals, all calculators behave differently ;-) But judging from the number of people asking questions similar to yours, maybe it wasn’t a very transparent assumption...
I asked about these differences in my second post in this post tree, where I explained how I understood these counterfactuals to work. I explained as clearly as I could that, for example, calculators should work as they do in real world. I did this explaining in hopes of someone voicing disagreement if I had misunderstood how these logical counterfactuals work.
However, modifying any calculator would mean that there can not be, in principle, any “smart” enough ai or agent that could detect it was in counterfactual. Our mental hardware that checks if logical coin should’ve been heads or tails is a calculator the same as any computer, and again, there does not seem to be any reason to assume Omega leaves some calculators unchanged while changes results of others.
Unless, this thing is just assumed to happen, with some silently assumed cutaway point where calculators become so internal they are left unmodified.
Calculators are not modified, they are just interpreted differently, so that when trying to answer the question of what happens in a certain situation (containing certain calculators etc.) we get different answers depending on what the assumptions are. The situation is the same, but the (simplifying) assumptions about it are different, and so simplified inferences about it are different as well. In some cases simplification is unavoidable, so that dependence of conclusions on assumptions becomes an essential feature.
My current understanding of logical counterfactuals is something like this: if the inconsistent formal theory PA+”the trillionth digit of pi is odd” has a short proof that the agent will take some action, which is much shorter than the proof in PA that the trillionth digit of pi is in fact even, then I say that the agent takes that action in that logical counterfactual.
Note that this definition leads to only one possible counterfactual action, because two different counterfactual actions with short proofs would lead to a short proof by contradiction that the digit of pi is odd, which by assumption doesn’t exist. Also note that the logical counterfactual affects all calculator-like things automatically, whether they are inside or outside the agent.
That’s an approximate definition that falls apart in edge cases, the post tries to make it slightly more exact.
(Btw, I think it should be mentioned that a central piece of motivation for this “logical counterfactuals” thing is that it’s probably the same construction that’s needed to evaluate possible actions in normal cases, without any contrived coins, for an agent that knows its own program. So for example although a counterfactual scenario can’t easily “lead” to two different actions, two different actions in that scenario can still be considered as possibly even more (easily shown to be) contradictory “logical counterfactuals” that include additional assumptions about what the action is.)
Try as I might, I cannot find any reference to what’s canonical way of building such counterfactual scenarios. Closest I could get was in http://lesswrong.com/lw/179/counterfactual_mugging_and_logical_uncertainty/ , where Vladimir Nesov seems to simply reduce logical uncertainty to ordinary uncertainty, but this does not seem to have anything to do with building formal theories and proving actions or any such thing.
To me, it seems largely arbitrary how agent should do when faced with such a dilemma, all dependent on actually specifying what it means to test a logical counterfactual. If you don’t specify what it means, whatever could happen as a result.
I am not sure there is a clean story yet on logical counterfactuals. Speaking for myself only, I am not yet convinced logical counterfactuals are “the right approach.”
Hi Ilya,
Me neither. Have you seen my post about common mistakes? To me it seems more productive and more fun to explore the implications of an idea without worrying if it’s the right approach.
I like “breadth first search” or more precisely “iterative deepening” better than “depth first search.”
(DFS is not guaranteed to find the optimal solution, after all!)