I was hoping people other than Jessica would share some specific curated insights they got [from the CTMU]. Syndiffeonesis is in fact a good insight.
The reply I’d drafted to this comment ended up ballooning into a whole LessWrong post. Here it is!
It used to seem crazy to me that the intentions and desires of conscious observers like us can influence quantum outcomes (/ which Everett branches we find ourselves in / “wave function collapses”), or that consciousness had anything to do with quantum mechanics in a way that wasn’t explained away by decoherence. The CTMU claims this happens, which seemed crazy to me at first, but I think I’ve figured out a reasonable possible interpretation in terms of anthropics. (Note: I am not an expert in either quantum mechanics or anthropics. I’m pretty sure the big-picture here is correct, but I won’t be surprised if I get some details wrong.)
[EDIT: Based on the feedback I have received thus far, it seems worthwhile to clarify up-front that I am not trying to make a positive case for consciousness affecting quantum outcomes. I am trying to make a case that the arguments against consciousness affecting quantum outcomes rest on shakier assumptions than I think people often give them credit for, and that there may be room for consciousness to affect quantum outcomes if we drop these assumptions. If I had to bet based on my current inside view, I would bet against consciousness affecting quantum outcomes. My actual position is much weaker than what I’ve since realized the title of this post might connote — apologies for the confusion.]
My previous picture of quantum mechanics was that we’re in a “probabilistic clockwork universe” – very roughly, it’s got a set of initial conditions (initial quantum state of the universe) and deterministic laws (guiding the evolution of the wavefunction) that are hooked up to a random number generator (specifying quantum outcomes), and the initial conditions + the deterministic laws + the outcomes of the random generator fully constrain what we observe. (Under many-worlds, the RNG tells us which Everett branches we find ourselves in.)
Importantly, the “quantum randomness” of the random number generator is truly, actually random. I think this picture of quantum mechanics does not leave much room for observers like us to influence quantum outcomes, and paints a picture of reality that’s beyond the reach of God.
Let’s say we identify which Everett branch we’re in using a length-n bitstring that encodes the outcomes of the quantum RNG. If the Everett branch we find ourselves in is truly random, then the K-complexity of this bitstring should be close to n. If we should a priori expect to find ourselves in some particular “truly random” Everett branch, the “probabilistic clockwork universe” picture would be correct.
However, there might be a special class of pseudorandom bitstrings, for which the prior probability that our quantum RNG bitstring is in this class far outweighs the prior probability that the bitstring is “truly random” (/ has K-complexity close to n). If this were the case, the Born probabilities would be more like reflections of logical uncertainty about pseudorandomness (like “the probability of some far-out digit of π being 5 is 10%”) than like parameterizations of a “truly random” RNG.
One way this could turn out to be true is if it’s a priori more likely that there are special, nonrandom portions of the quantum multiverse we’re being sampled from. For example, if we had a priori reasons for expecting that we’re in a simulation by some superintelligence trying to calculate the most likely distribution of superintelligences in foreign universes for acausal trade reasons, then we would have a priori reasons for expecting to find ourselves in Everett branches in which our civilization ends up producing some kind of superintelligence – i.e., that it’s in our logical past that our civilization ends up building some sort of superintelligence.
Under this picture of reality, the deterministic laws of physics remain the same, but the initial conditions and “RNG outcomes” are underdetermined, and merely have high-level constraints applied to them, which get increasingly refined over time by our simulators, while possibly never reaching full resolution, if their full resolution is never needed (like in lazy evaluation).
This is kind of like “authoring a story over the course of logical time”, by fleshing out the details of a plot idea in parallel, stitching them all together in a logically consistent way, and iterating until the story is as complete as it needs to be. The order in which the author fleshes out these details may be totally different from the timeline of the story, and many unimportant details may never get fleshed out, including possibly many details about the initial conditions.
This contrasts starkly with the “probabilistic clockwork universe” picture, where, roughly speaking, the initial states are fully specified, the next states are then fully specified with a “turn of the crank” (an application of the physical laws + a consultation of the RNG, if we’re tracking which Everett branch we move to), the next states after that are fully specified with another “turn of the crank”, and so on.
Crucially, in the “authorship” picture, the specifics of how future details get filled in might depend on the specifics of how past details get filled in (where “future” and “past” refer to logical time here, not physical time). Some of these past details might be details about us, like the actions we take and the thoughts we think. Just as the desires of a fictional character might affect how the rest of the story gets composed, our desires might affect how the details of the rest of physical reality get filled in, via influencing quantum measurement outcomes and retrodicting initial conditions of the universe (the latter of which Scott Aaronson has speculated about).
Zooming way out, I want to be clear that I am highly uncertain about this “authorship” picture. In particular, the a priori likelihood of being in a not-truly-random Everett branch is still highly uncertain to me, and the bit about superintelligence simulations was meant more as an intuition pump to illustrate one possible way things could be than as a likely account of how things actually are. The current epistemic state of my inside view around whether our consciousness can influence quantum outcomes is “here be dragons”, which I think is importantly different from both “yes, obviously” (as Chris Langan thinks), and “no, that’s obviously crazy, are you one of those supernaturalist nutjobs who doesn’t understand physics?” (as I used to think).
That being said, Chris has (in his head, at least apparently) detailed and specific arguments behind his “yes, obviously”, having largely to do with how to think clearly about reality as a self-simulation, and the implications of that for anthropics. In other words, I think he puts forth a much more detailed case for how to think about the a priori likelihood of being in a “nonrandom” Everett branch. I don’t understand Chris’s views here yet, but my prediction is that I will end up updating toward Chris’s position after understanding his views better, because that is usually what happens to me when I understand Chris’s views better.
I’ll close by noting that, as far as I understand, the CTMU implies that the “authorship” picture of reality is correct; “telic recursion” refers to the iterative process of filling in these details throughout logical time; “intelligent design” (as Chris uses the term) is just another term for telic recursion; “UBT” essentially refers to “the story of reality” prior to any details getting filled in; “extended superposition” refers to the status of any “underdetermined portion of the story” whose details are “still in superposition”, but spread out across swaths of time.
Thanks to Marcello Herreshoff for coming up with the analogy of an author writing a story; thanks to Marcello Herreshoff and Nick Tarleton for reviewing earlier drafts of this post; and thanks to Scott Garrabrant for sanity-checking earlier versions of this general idea in private conversations.
CTMU insight: maybe consciousness *can* affect quantum outcomes?
From one of justinpombrio’s comments on Jessica Taylor’s review of the CTMU:
The reply I’d drafted to this comment ended up ballooning into a whole LessWrong post. Here it is!
It used to seem crazy to me that the intentions and desires of conscious observers like us can influence quantum outcomes (/ which Everett branches we find ourselves in / “wave function collapses”), or that consciousness had anything to do with quantum mechanics in a way that wasn’t explained away by decoherence. The CTMU claims this happens, which seemed crazy to me at first, but I think I’ve figured out a reasonable possible interpretation in terms of anthropics. (Note: I am not an expert in either quantum mechanics or anthropics. I’m pretty sure the big-picture here is correct, but I won’t be surprised if I get some details wrong.)
[EDIT: Based on the feedback I have received thus far, it seems worthwhile to clarify up-front that I am not trying to make a positive case for consciousness affecting quantum outcomes. I am trying to make a case that the arguments against consciousness affecting quantum outcomes rest on shakier assumptions than I think people often give them credit for, and that there may be room for consciousness to affect quantum outcomes if we drop these assumptions. If I had to bet based on my current inside view, I would bet against consciousness affecting quantum outcomes. My actual position is much weaker than what I’ve since realized the title of this post might connote — apologies for the confusion.]
My previous picture of quantum mechanics was that we’re in a “probabilistic clockwork universe” – very roughly, it’s got a set of initial conditions (initial quantum state of the universe) and deterministic laws (guiding the evolution of the wavefunction) that are hooked up to a random number generator (specifying quantum outcomes), and the initial conditions + the deterministic laws + the outcomes of the random generator fully constrain what we observe. (Under many-worlds, the RNG tells us which Everett branches we find ourselves in.)
Importantly, the “quantum randomness” of the random number generator is truly, actually random. I think this picture of quantum mechanics does not leave much room for observers like us to influence quantum outcomes, and paints a picture of reality that’s beyond the reach of God.
Let’s say we identify which Everett branch we’re in using a length-n bitstring that encodes the outcomes of the quantum RNG. If the Everett branch we find ourselves in is truly random, then the K-complexity of this bitstring should be close to n. If we should a priori expect to find ourselves in some particular “truly random” Everett branch, the “probabilistic clockwork universe” picture would be correct.
However, there might be a special class of pseudorandom bitstrings, for which the prior probability that our quantum RNG bitstring is in this class far outweighs the prior probability that the bitstring is “truly random” (/ has K-complexity close to n). If this were the case, the Born probabilities would be more like reflections of logical uncertainty about pseudorandomness (like “the probability of some far-out digit of π being 5 is 10%”) than like parameterizations of a “truly random” RNG.
One way this could turn out to be true is if it’s a priori more likely that there are special, nonrandom portions of the quantum multiverse we’re being sampled from. For example, if we had a priori reasons for expecting that we’re in a simulation by some superintelligence trying to calculate the most likely distribution of superintelligences in foreign universes for acausal trade reasons, then we would have a priori reasons for expecting to find ourselves in Everett branches in which our civilization ends up producing some kind of superintelligence – i.e., that it’s in our logical past that our civilization ends up building some sort of superintelligence.
Under this picture of reality, the deterministic laws of physics remain the same, but the initial conditions and “RNG outcomes” are underdetermined, and merely have high-level constraints applied to them, which get increasingly refined over time by our simulators, while possibly never reaching full resolution, if their full resolution is never needed (like in lazy evaluation).
This is kind of like “authoring a story over the course of logical time”, by fleshing out the details of a plot idea in parallel, stitching them all together in a logically consistent way, and iterating until the story is as complete as it needs to be. The order in which the author fleshes out these details may be totally different from the timeline of the story, and many unimportant details may never get fleshed out, including possibly many details about the initial conditions.
This contrasts starkly with the “probabilistic clockwork universe” picture, where, roughly speaking, the initial states are fully specified, the next states are then fully specified with a “turn of the crank” (an application of the physical laws + a consultation of the RNG, if we’re tracking which Everett branch we move to), the next states after that are fully specified with another “turn of the crank”, and so on.
Crucially, in the “authorship” picture, the specifics of how future details get filled in might depend on the specifics of how past details get filled in (where “future” and “past” refer to logical time here, not physical time). Some of these past details might be details about us, like the actions we take and the thoughts we think. Just as the desires of a fictional character might affect how the rest of the story gets composed, our desires might affect how the details of the rest of physical reality get filled in, via influencing quantum measurement outcomes and retrodicting initial conditions of the universe (the latter of which Scott Aaronson has speculated about).
Zooming way out, I want to be clear that I am highly uncertain about this “authorship” picture. In particular, the a priori likelihood of being in a not-truly-random Everett branch is still highly uncertain to me, and the bit about superintelligence simulations was meant more as an intuition pump to illustrate one possible way things could be than as a likely account of how things actually are. The current epistemic state of my inside view around whether our consciousness can influence quantum outcomes is “here be dragons”, which I think is importantly different from both “yes, obviously” (as Chris Langan thinks), and “no, that’s obviously crazy, are you one of those supernaturalist nutjobs who doesn’t understand physics?” (as I used to think).
That being said, Chris has (in his head, at least apparently) detailed and specific arguments behind his “yes, obviously”, having largely to do with how to think clearly about reality as a self-simulation, and the implications of that for anthropics. In other words, I think he puts forth a much more detailed case for how to think about the a priori likelihood of being in a “nonrandom” Everett branch. I don’t understand Chris’s views here yet, but my prediction is that I will end up updating toward Chris’s position after understanding his views better, because that is usually what happens to me when I understand Chris’s views better.
I’ll close by noting that, as far as I understand, the CTMU implies that the “authorship” picture of reality is correct; “telic recursion” refers to the iterative process of filling in these details throughout logical time; “intelligent design” (as Chris uses the term) is just another term for telic recursion; “UBT” essentially refers to “the story of reality” prior to any details getting filled in; “extended superposition” refers to the status of any “underdetermined portion of the story” whose details are “still in superposition”, but spread out across swaths of time.
Thanks to Marcello Herreshoff for coming up with the analogy of an author writing a story; thanks to Marcello Herreshoff and Nick Tarleton for reviewing earlier drafts of this post; and thanks to Scott Garrabrant for sanity-checking earlier versions of this general idea in private conversations.