Explain This Correlation

At SciFoo Camp last weekend, famed quantum gravitist Lee Smolin mentioned that he’d noticed a correlation between these beliefs:

  1. Many worlds for quantum mechanics,
  2. Anthropic arguments in physics, and
  3. Conscious computer-based AIs could be built.

This correlation seems intuitively right, but puzzling.  Any explanation for why it exists other than the obvious, that some people tend to be right about everything?

Added: Lee and most who came to the particular SciFoo session where he made this observation disagree with these beliefs, yet were creative sharp physicists, hackers, sociologists, etc.

GD Star Rating
Tagged as:
Trackback URL:
  • Allan Crossman

    Many Worlds has significant anthropic implications. For example, problems surrounding the improbability of abiogenesis disappear immediately, as far as I can see.

    So people might be thinking about the two together. As for AI, I have no idea.

    • The anthropic arguments about OOL apply equally well if there is only one world. The universe we see would *still* be expected to be life-friendly. The MWI would not relevant in this area – even if it was the only “multiverse” theory – which it isn’t.

      • Allan Crossman

        It seems to me that the existence of life is made nearly inevitable by MWI. If life was extremely unlikely otherwise *, then the fact that life actually exists would be evidence for MWI **.

        * I don’t actually assert this.

        ** As you say, other big cosmologies need to be considered as well.

  • anon

    Perhaps people who believe in feasible computer-based AI are likely to believe in simulation arguments, which also provide an explanation of anthropic arguments.

    Also, saying that we are running in a simulated world makes it likely that many other worlds are being simulated, which sounds superficially similar to many-worlds.

  • Each of these positions seems to violate the literal interpretation of Occam’s Razor: “entities should not be multiplied unnecessarily.” I think that people who adopt these beliefs have a different notion of what Occam’s Razor is really about (i.e., based on a formal definition of conceptual complexity such as program length instead of counting “entities”), and that accounts for the correlation.

    • Mike

      Yes… I think so. I’ve heard it said by many, and consider so myself, that the “multiverse” pictures satisfy a notion of “simplicity” because they avoid having to explain where many of the complex details of our universe come from, by supposing a large degree of randomness.

      For instance if Earth were the only planet, we would seemingly have a lot of explaining to do, to account for the convergence of so many details that happen to be important for life — which collectively make Earth a very “complex” system. But knowing Earth is one among countless planets, we can consider simpler underlying dynamics, which involve random aspects (like “initial conditions”), producing a large variety of planet types. The complexity of Earth is then just a symptom of myopia, combined with anthropic selection.

  • Scott Aaronson thinks many-worlds and related beliefs are tied by the bullet-avoiders vs bullet-swallowers axis.

  • Kevin Dick

    All of these beliefs resonate with common themes in science fiction?

  • kvn

    Well, the great majority of human beings have never heard of (let alone seriously contemplated) these things.

    People who have heard of all of them (they are in two disparate fields, physics and CS/AI) are usually “big thinkers” who are interested in broad explanations. These are also “impractical”, theoretical academic concepts. They don’t have lucrative practical short or medium term applications.

    The personality characteristic the believers in these concepts is “bullet-swallowing”. I would bet they are more likely to accept other “big explanations” like Marxism or libertarianism as well.

    • Constant

      Many worlds takes Schroedinger’s equation seriously. In contrast, Marxism does not take market economics seriously but rather is itself a bad, alternative, crackpot theory. The appropriate analog of Marxism is not many-worlds, but rather take-your-pick of the crackpots rejecting modem physics. Moreover many worlds is a simplification: eliminate the ad hoc hypothesis of a wavefunction collapse and realize that decoherence already accounts for observed phenomena. In contrast Marxism is a big convoluted mess that resists summarization. It is no simplification. It is no elimination of any ad hoc hypothesis.

      • mjgeddes

        The third belief in the cluster (conscious-AI) is pretty clear-cut rational, but the other two are highly controversial (as shown by my lower probability assessments). So some explanation other than rationality is needed.

        >Many worlds takes Schroedinger’s equation seriously.

        So do numerous alternative explanations. I think Yudkowsky successfully rebutted all explanations involving the notion of a ‘wave-function collapse’ in his series on this blog, but readers should be aware that numerous other alternative explanations still remain. The Bohm interpretation, for example, has no wave-function collapse (according to this interpretation reality is stratified into different levels – both wave-function and particle levels equally ontologically real – this is a similar idea to my own ‘3-level causality’ theory.)

        See for example, the good wikipedia article summarizing the ontology of some of the more popular alternatives (go to comparison table at end):

        Ontology of QM Interpretations

        >Marxism does not take market economics seriously but rather is itself a bad, alternative, crackpot theory.

        I agree, but Marxists are not likely to see it that way. And in fact, its opposite number, Libertarianism, could also be regarded as just as crackpot. Even moderate versions are certainly not regarded as ‘rational’ by main-stream economists.

        Few take anthropics seriously. It is usually used as a sort of ‘God of the gaps’ – once causal explanations of something are available, you can dispense with anthropics, for instance anthropic arguments are used to explain the cosmological constant only because there is currently no causal explanation, once one is available, I bet that anthropics will be dropped like a hot-potato.

        I want to suggest that the believers in this belief cluster are not nearly as ‘rational’ as they think. They just fancy themselves as ‘cool smarties’.

      • Constant

        I was comparing MW with Marxism, not talking about anthropism or libertarianism. The category of “big explanation” is a crude, vague, hand-waving thing, of limited explanatory value here. You might call Schroedinger’s equation itself a “big explanation” since it applies to so much. This is what I was critiquing, using as an example MW versus Marxism. What KVN’s argument in effect does is damn MW by association with Marxism, and a highly questionable, forced association at that, which is the height of lazy reasoning. I’m aware of Bohm. Bohm adds new elements since it’s a hidden-variable theory. It’s not necessarily wrong, but it’s not taking the Schroedinger’s equation as seriously as MW, since Schroedinger’s equation does not say anything about the hidden variables, any more than it says anything about the wave-function collapse.

  • Mike

    I think (1) and (2) are linked by an explanation given above (which I placed a comment below).

    As for (3), this just comes from a materialistic view of the world (unless I misunderstand). Most people who think about questions related to (1) and (2) subscribe to such a view, so they accept (3) too. They would also tend to accept (4) there is no afterlife… but we do not consider that an odd coincidence.

  • mjgeddes

    I second kvn – they are all ‘big picture’ grand ideas, appealing to big-picture thinkers looking for grand explanations. And the fact they are all not widely understand by the general populace may also have something to do it (appealing to antiauthoritarian, unconventional thinkers who see themselves as keepers of deep truths not generally known to the masses).

    Are they right though? My probablities:

    (1) Many-Worlds: 50% ; good theory, seems to be consistent and have a lot of elegance ; but too many loose ends and alternative explanations to be very sure.

    (2) Anthropic Reasoning in Physics – 20%; never been a fan, don’t think anthropic reasoning explains anything, so much complexity means ad hoc explanations can be used to ‘explain’ almost anything, semantic ambiguities in the ‘referenece class’ probably invalidate anthropic reasoning anyway. Still smarter people than me think there’s something to it so 20%.

    (3) Conscious computer-based AIs could be built – >95% – consciousness is a physical process and is wholly computable as far as we know. Unlikely that these assumptions are wrong but still some things we don’t know, so a small question mark remains.

    • Mike

      Anthropics must play a role in explaining at least some aspects of what we observe. The only question is whether it plays a large role in certain aspects or not. (At it’s core, anthropic reasoning is merely conditioning the results of experiments based on some understanding of the experimental conditions. When I ask what I should see when I look into the sky, I should consider the fact that it is me looking at the sky, and not a randomly selected photon, for instance.)

  • They are all based on non proved stuff and are pretty modern and futuristic

  • And not only that, I bet they’re more likely than average to be atheists, to be against protectionism and in favor of international trade, approve of gay marriage, and have higher IQs. If this is *not* the case, *then* we can go looking for “special” explanations.

  • Rafal Smigrodzki

    To add to Eliezer’s list – they accept evolution and evolutionary psychology, do not have taboo trade-offs, blocked exchanges, forbidden base rates, and heretical counterfactuals. I have been recently reading Tyler’s new book, and I think that the association of beliefs may be explained by the autistic cognitive style. The style is caused by a neurological dysfunction, an impairment of the formation of long-range intracortical connections which create evolutionarily conserved presets necessary for fast acquisition of social, language and perhaps physical modeling (folk physics) skills. As a result, such thinkers let their hierarchical temporal memory, or the general cognitive engine that is our cortex, run wild, without the systematic biases that normal people have built in – and they follow premises to their conclusions, without intrusion of antediluvian, hardware-encoded modes of thought. Of course, they also frequently make mistakes but of a more random, non-systematic nature. And unfortunately they have difficulties with running the 150 simultaneous iterated Prisoner’s Dilemma games that normal people easily do with their hardware-accelerated social cognition.

    • Interesting… I guess this is the book you’re referring to.

      • Rafal Smigrodzki

        Yes. I just finished it. I allowed myself to wax poetic on the extropian list, here is what I wrote:

        I have some of my own reflections on autistic cognitive style:

        Autism is a form of abnormal development of the brain, where the most
        basic neurological substrate of cognition, such as the cortical
        mantle, cortical columns, and the subcortical support structures,
        develop normally, yet there is an impairment in the long-range
        connections (white matter tracts) between cortical modules. This
        differs from garden-variety retardation where the substrate of
        cognition is itself dysfunctional. Furthermore, the impairment of
        long-range white matter tracts is not random. Random damage to white
        matter is seen in multiple sclerosis and microvascular disease which
        produce a totally different cognitive outcome (although infrequently
        one may observe some savant-like traits). The tracts most affected in
        autism tend to be the highly specialized networks important for social
        and linguistic tasks, while general-purpose cognitive circuitry is

        The special purpose circuitry is very important for fast achievement
        of social competence, which is obviously extremely important for
        survival in the prehistoric jungle that shaped our genes. Yet, such
        speed comes at a cost: The social circuitry imposes a pre-determined
        structure on cognition, directing attention to stimuli and phenomena
        most important for survival in the jungle. This structure carries with
        itself evolutionary assumptions about social tactics, finely
        calibrated to life in small tribes. It teaches about dominance and
        submission games, coalition-building and exclusion of strangers, the
        uses of violence and deception in both offensive and defensive
        applications. It teaches to devote cognitive resources to tracking the
        alphas, rather than count petals in various flowers. It weaves lies,
        violence and the ability to self-deceive into the very fabric of the
        mind, making them into unseen yet all pervading facts of life.

        The mildly autistic mind is to some extent deprived of such
        pre-ordained structure. It is free to use its raw intelligence without
        compulsive focus on tribal structure. It tends to excel and then focus
        on tasks that do not require extensive long-range cortical
        connections, such as mathematics, or physical modeling of the
        environment. It is forced to come up with its own ways of imposing
        structure on sensory input which is much slower than pre-configured
        ideas – but it is free not to be shackled by tricks calibrated for the
        hunter-gatherer lifestyle. The autistic person is in a way weak,
        innocent, inept at bullying, lying and social manipulation, yet his
        mind is like a primal force, unbound by outdated assumptions. He is
        mentally undivided, applying his general cognitive engine to a greater
        scope of tasks than usual among the dominant social animals. His style
        reflects his neural structure – a Hawkins’ hierarchical temporal
        memory, fluidly building ever more complex hierarchy of world models
        from simpler units, without the shortcuts afforded by the specialized
        neural hardware seen in standard humans – slower but less likely to
        make systematic mistakes in understanding. He is the persistent
        truth-finder, at times appalled by what he finds out. He is the
        innocent, fluid force of truth.

  • rosyatrandom

    @ Eliezer: While I am a proponent of all 3 and, further, the additional ones you suggest, I would hazard that simply because these are unusual beliefs that require significant appreciation and grasp of abstract concepts, they would mandate higher intelligence anyway. As kvn said, you have to be fairly smart just to have heard of and understand them in the first place. What would be more telling would be the reverse correlation: of people who have heard and understand, what are the proportions who find them compelling?

    I also agree with Mike: these positions are tied together with a ‘why not?’ attitude for accepting umbrella explanations. The more general a system is, the simpler, and the larger number of more complex ‘special cases’ it can contain.

  • I just added to the post, which I think fits poorly with the atheism, abstractness, big thinker, and IQ explanations.

    • Constant

      It would be helpful to know some of their reasons for rejecting mw etc. Or do you suspect the stated reasons are rationalizations and thus not the true reasons?

    • Was there some kind of survey of participant beliefs? How did that work? What were the results?

    • fenn

      “smarter people than me think there’s something to it”

      They have blogs?

    • mjgeddes

      True, yes, OK, our first attempts to answer were wrong, so give us another crack at it –

      Here goes, next attempt at explanation:

      ‘The people willing to believe these things are more willing to question prevailing social beliefs – thus bringing them more in contact with unusual ideas. Most people at the SciFoo session were not really familar with the said beliefs, because they were more socially conventional, focusing mainly on theories relevant to their careers’

      Of course, this willing to question prevailing social beliefs cuts both way, it will lead thinkers to great new big ideas when they are rational, but insane big ideas when they are irrational – and these extremes are exactly what are observed on the transhumanist lists – great big ideas, mixed in with crazy rubbish such as Marxism for example.

  • Oliver Beatson

    The beliefs are all based on assumptions that would not be there without the existence of intelligent humans?

  • Douglas Knight

    What is “belief” in “anthropic arguments in physics” supposed to mean? I don’t believe than there has been a single serious anthropic argument in physics. That’s a claim about humans, not about fundamental physics.

    • Mike

      Weinberg’s 1987 paper is serious. And it was preceded by serious, though more vague, arguments by Linde, and has been followed by serious arguments by many authors.

      Right now anthropic selection is the only plausible explanation for the size of the cosmological constant. The details are still uncertain — not really because of ambiguities in anthropic reasoning but because of dependence on the spacetime measure. But very few physicists would say this work is not serious… and I’d say it’s because they are prejudiced or haven’t thought long about it.

      • Douglas Knight

        You seem to be saying that, conditioning on physicists, there is no correlation between 1 and 2. Given that, is there any point to Smolin’s claim?

  • John Maxwell IV

    Maybe people who believe that stuff tend to be younger?

    • Mike

      No, I don’t think so. Weinberg and Susskind and Linde and Vilenkin are names of senior scientists that immediately come to mind.

  • Ah. Well, of course Lee would then think there was a puzzle. But I of course have no reason to induce a puzzle from the information given.

    Perhaps a way to phrase the question more interestingly would be as follows: Which of the beliefs in the many-worlds-believer cluster – which could identify rationalists or Big Theory believers or bullet-swallowers or something else – do *not* belong there, if any, according to you? That is, if we look at things that believers in many-worlds who attend SciFoo Camp believe more frequently than non-MWI SciFoo attendants, are there any things which you think aren’t there for rational reasons, and are instead there for some other reasons? And then how did they end up on that list?

  • Adam

    I think what they have in common is the idea that consciousness is not specially privileged in the universe. Many-worlds is an appealing alternative to theories like Copenhagen which require “observers” that must be defined on a physical level (though I’ve not seen any attempts, let alone successful ones, to actually provide such a definition).

    Next, the possibility of materialistic, computer-driven AI is a easy to infer if human consciousness is generated solely by our physical, biological brains, rather than by any special physical process or something inscrutable by science. Finally, anthropic reasoning fits well with this mindset because it avoids all the “why are we here?” speculation that’s inevitable if consciousness *is* special and we are therefore “supposed to” be here.

  • jimmy

    @ Eliezer: While I agree that you’d expect this correlation if some people just ‘get it’ better than others, there is more that can be interesting here.

    It would be rather boring if the correlation went away after conditioning on IQ, since the explanation would just be “smarter people are more likely to get any specific question right, and since IQ predicts performance on all questions, you’d expect correlation”.

    However, the correlation could be stronger, and persist even after conditioning for IQ (or whatever ‘intelligence’ proxy you use). It could be the case that there is some concept that some people grok and others don’t get at all (which correlates imperfectly with IQ) that directly predicts belief in 1,2,3. If this is the case, it would be interesting and useful to find out the cause.

    It could be, like Wei Dai suggested, that everyone that understands Occam’s razor believes 1,2,3. Another possibility is that the people that get it understand that their gut feelings come from the absurdity heuristic, and know to ignore them.

    From Robin’s added section, it appears that it is indeed not fully predicted by ‘intelligence’. Does this sound right?

  • Rob Spear

    Believers in these three things obviously prefer large-scale theistic type hand waving, rather than practical dealing with small problems. They are the type of person who has a lot of emotional investment in how smart they are, and is therefore driven to impressive sounding theories of everything, which tend not to help with actual problems in the real world. See also: the Singularity, most public Economists, etc etc.

    • Mike

      as opposed to those people who make themselves feel smart by making up grand generalizations that makes others look less smart.

    • Adam

      Is CI less frivolously “impressive-sounding” and more practical against real-world problems than many-worlds? Is the idea that consciousness/intelligence are purely physical processes, that could be simulated like any other, somehow more “large-scale theistic-type hand waving” than its opposite? Regardless of the merits and criticisms of anthropic reasoning, do you seriously judge those who favor it over your preferred cosmology as emotionally-insecure head-in-the-clouds types trying to sound smart?

  • Patrick (orthonormal)

    It seems to me that there is a common thread: a patternist theory of consciousness.

    In my case, I take MWI seriously for reasons of mathematical elegance. Once I accept that my consciousness is continually branching with the rest of the universe, it appears disingenuous to identify it with anything apart from the patterns of causality in my brain; this patternism then gives (2) and (3) as relatively easy corollaries.

  • Hal Finney

    For the anthropic case, you probably need to be looking at some kind of multiverse. Few object to using anthropic reasoning to explain the fact that we live on a planet where life can survive. It’s more troublesome if we want to explain why we live in a universe where life can survive, unless you have some reason to believe there are other universes. So the MWI-anthropic connection makes sense.

    I’m surprised about the connection with AI though, surprised enough that I would want to see evidence that it’s true. From what I gather from way afar, most physicists neither accept the MWI nor any multiverse. Do most physicists then believe that conscious AI is impossible? That would be a surprise to me. Conscious AI is a common part of popular futurist culture, and I’m not aware of grumpiness from the physics community that such AIs are scientifically inaccurate (unlike things like FTL travel, where there are frequent low-key complaints).

    • Mike

      Perhaps this is why people writing above say what they do about “anthropics.” It should be made clear that the only flavor of “anthropic reasoning” that a significant number of physicists entertain is the idea that our universe is one among many with a variety of different properties, and we observe what properties we do in part because they, unlike others, are suitable to the evolution of observers like us.

      Thus, aside from a few who I would consider cranks, physicists believe in anthropic arguments only insofar as they believe in a multiverse.

      Regarding AI, I suspect a physicists belief in this is independent his/her position on (1) and (2)… so “correlation” might not be a good word. Maybe I am missing something, but to me the question is merely is it in principle possible to emulate a human mind using a device that is not merely a copy of the human brain. Physicists, regardless of positions (1) and (2), tend not to believe in a “soul” — and I think for most the question would boil down to whether thoughts ultimately reflect only the processing of information, in which case one accepts the possibility to build a computer to do the same (though the necessary resources could be enormous, and civilization may end before we figure it all out). Perhaps physicists readily accept the proposition because they are unaware of what is supposed to be the subtlety (that’s sort of how I feel).

      • Thus, aside from a few who I would consider cranks, physicists believe in anthropic arguments only insofar as they believe in a multiverse.

        That is exactly correct, but you miss the critical point of this willful ignorance, (no, you even join ranks with them), that this is also the reason that physicists are the cranks, since the weak interpretation is not what is observed… duh.


  • Aaron

    I’m sorry, I’m a perennial newbie at QM. But is there any particular reason why loop quantum gravity would necessarily be in opposition to with MWI? I thought string theory’s perceived flaw was appealing to higher dimensions in this universe.

    • Mike

      I don’t know about your first question, but I would have guessed the answer is “no.”

      I don’t think the “perceived flaw” of string theory is higher dimensions. In fact, that’s an enormous asset, if for instance you think anthropics is the only way to explain the cosmological constant.

      I would guess the loop folks would say the flaw in string theory is that it has to take for granted a “background” Minkowski metric, whereas one hopes a fundamental theory incorporating GR would not have to make any prior assumptions about the background geometry. But I’m not very initiated into either school.

  • mjgeddes

    I just want put in a plug for the Bohm intepretation of QM as a strong alternative to MWI. I didn’t know anything technical about it, about when I researched it I was amazed to discover that it lined up so perfectly to my 3-level ontological model of reality. This has really caused me to start doubting MWI.

    In Bohm, reality has a three level structure – the wave function is split in two – there’s the basic wave function at the top level, a new quantum potential at the mid-level, and the particle on the bottom level. All three levels are equally real. Although in Bohm there is no wave function collapse and all branches are equally real, there is only one actual concrete particle history (one world).

    Again, I must emphasize, this is a perfect match to my own recursive 3-level ontological model, which has really startled me.

    • Constant

      I learned about Bohm at about the same time as MWI, about 20 years ago. I find MWI more compelling for two reasons. First, MWI has fewer bits – Occam’s razor rules in its favor (as I view it). Second, if Bohm is correct about the objective reality in itself, then I think Bohm’s pilot wave would be capable of sustaining subjective experience without the help of the particle. That is, even if there is a pilot wave and a particle as Bohm says, we could just as well occupy some other part of the pilot wave.

      • mjgeddes

        Bohm’s ontology is far cleaner than the MWI ontology. The trouble with MWI is that is cannot explain why we observe the reality we actually do (how does our subjective experience emerge from the wavefunction?).

        It comes down to a matter of levels of abstraction. If you think reality only operates on a single level (reductionism) then you’ll go for MWI. If on the other hand, you think reality is best divided into different levels of abstraction, you’ll be sympathetic toward Bohm.

        An analogy might be the relation between deduction and induction; if deduction is just a special case of induction, does that mean we can dispense with the notion of deduction? By analogy, if the particle is somehow just a part of the pilot wave, can we dispense with the notion of the particle? I lean towards a ‘no’ answer in both cases.

  • mitchell porter

    This is a bit like asking why the elite intelligentsia of 1900 believed in Fabian socialism, eugenics, and the objective existence of atoms (which was an issue at the time). The common theme would be that these ideas look like the best available answer to some important question. But the details are a mass of historical contingency.