The quantum computer that knows all

This is my first post in more than a month that’s totally unrelated to the covid crisis. Or rather, it’s related only insofar as it’s about a Hulu miniseries, the sort of thing that many of us have more occasion to watch while holed up at home.

Three weeks ago, a journalist named Ben Lindbergh—who’d previously asked me to comment on the scientific accuracy of Avengers: Endgame—asked me the same question about the miniseries Devs, which I hadn’t previously heard of.

[Warning: Spoilers follow]

‘Devs,’ I learned, is a spooky sci-fi action thriller about a secretive Silicon Valley company that builds a quantum computer that can perfectly reconstruct the past, down to what Jesus looked like on the cross, and can also (at least up to a point) predict the future.

And I was supposed, not only to endure such a show, but to comment on the accuracy of its invocations of quantum computing? This didn’t sound promising.

But, y’know, I was at home quarantined. So I agreed to watch the first episode. Which quickly turned into the second, third, fourth, fifth, sixth, and seventh episodes (the eighth and final one isn’t out yet).

It turns out that ‘Devs’ isn’t too bad, except that it’s not particularly about quantum computers. The latter is simply a buzzword chosen by the writers for a plot concept that would’ve been entirely familiar to the ancient Greeks, who called it the Delphic Oracle. You know, the mysterious entity that prophesies your fate, so then you try to escape the prophecy, but your very evasive maneuvers make the prophecy come true? Picture that, except with qubits—and for some reason, in a gleaming golden laboratory that has components that float in midair.

Devs Trailer Reveals New Look at FX-Hulu's Upcoming Limited Series
If you’re never visited a real quantum computing lab: they’re messier and a lot less golden.

At this point, I’ll just link you to Ben Lindbergh’s article about the show: Making Sense of the Science and Philosophy of ‘Devs.’ His long and excellent piece quotes me extensively enough that I see no need also to analyze the show in this blog post. (It also quotes several academic philosophers.)

Instead, I’ll just share a few tidbits that Ben left out, but that might be amusing to quantum computing fans.

  • The first episode opens with a conversation between two characters about how even “elliptical curve” cryptography is insecure against attack by quantum computers. So I immediately knew both that the writers had one or more consultants who actually knew something about QC, and also that those consultants were not as heavily involved as they could’ve been.
  • Similarly: in a later scene, some employees at the secretive company hold what appears to be a reading group about Shor’s algorithm. They talk about waves that interfere and cancel each other out, which is great, but beyond that their discussion sounded to me like nonsense. In particular, their idea seemed to be that the waves would reinforce at the prime factors p and q themselves, rather than at inverse multiples of the period of a periodic function that only indirectly encodes the factoring problem. (What do you say: should we let this one slide?)
  • “How many qubits does this thing have?” “A number that there would be no point in describing as a number.” ROFL
  • In the show, a crucial break comes when the employees abandon a prediction algorithm based on the deBroglie-Bohm pilot wave interpretation, and substitute one based on Everett’s many-worlds interpretation. Which I could actually almost believe, except that the many-worlds interpretation seems to contradict the entire premise of the rest of the show?
  • A new employee, after he sees the code of the superpowerful quantum computer for the first time, is so disoriented and overwhelmed that he runs and vomits into a toilet. I, too, have had that reaction to the claims of certain quantum computing companies, although in some sense for the opposite reason.

Anyway, none of the above addresses the show’s central conceit: namely, that the Laplace demon can be made real, the past and future rendered fully knowable (with at most occasional breaks and exceptions) by a machine that’s feasible to build. This conceit is fascinating to explore, but also false.

In the past, if you’d asked me to justify its falsity, I would’ve talked about chaos, and quantum mechanics, and the unknowability of the fine details of the universe’s state; I might’ve even pointed you to my Ghost in the Quantum Turing Machine essay. I also would’ve mentioned the severe conceptual difficulties in forcing Nature to find a fixed-point of a universe where you get to see your own future and act on that information (these difficulties are just a variant of the famous Grandfather Paradox).

But it occurs to me that, just as the coronavirus has now made plain the nature of exponential growth, even to the world’s least abstract-minded person, so too it’s made plain the universe’s unpredictability. Let’s put it this way: do you find it plausible that the quantum computer from ‘Devs,’ had you booted it up six months ago, would’ve known the exact state of every nucleotide in every virus in every bat in Wuhan? No? Then it wouldn’t have known our future.

And I see now that I’ve violated my promise that this post would have nothing to do with covid.

84 Responses to “The quantum computer that knows all”

  1. Jacob Says:

    Feel free to leave this in moderation but I have to at least ask. Any comment on this:

  2. Patrick M. Dennis, MD Says:

    Scott, I’ve read your blog and enough of your other writing (including that great SMBC piece) long enough to know that something like “QC derives its vast superiority over conventional computing because in QC there are 2^64 (or whatever) parallel processors going at once (or alternatively, that they are doing their processing in alternate universes).” is not, well, accurate.

    My limited understanding is that it, QC, has more to do with wavefront interference, controlled in such a way that the output emerges from the resulting pattern. Beyond that, I get the feeling that a more detailed comprehension would necessitate a deeper understanding of quantum mechanics than I currently posses.

    My question, then: the only example of wavefront interference that I’m familiar with is in some explanations of the double-slit experiment. So, is that experiment in any sense a (perhaps 1 qbit) quantum computer? If so, what problem is it solving?

  3. Scott Says:

    Patrick #2: The single-photon double-slit experiment precedes the idea of a quantum computer by ~70 years. So, rather than calling the double-slit experiment a “1-qubit quantum computer” (although in a certain sense it is that), I think it would be better to call a quantum computer a “massively scaled-up double-slit experiment”! 🙂 In particular, a QC is like a double-slit experiment scaled up to the point where it’s hard for a classical computer even to calculate the interference pattern, but where (ideally) the constructive interference happens only or mostly at configurations that reveal the answer to a problem that you care about. For more, see e.g. Quantum Computing Since Democritus or my undergrad lecture notes or any of the resources linked from the sidebar of this blog.

  4. John Figueroa Says:

    > A crucial break comes when the employees abandon a prediction algorithm based on the deBroglie-Bohm pilot wave interpretation, and substitute one based on Everett’s many-worlds interpretation. Which I could actually almost believe, except that the many-worlds interpretation seems to contradict the entire premise of the rest of the show?

    Would a deBrogile-Bohm–based interpretation give different results from an Everettian one? I thought the math was the same in both cases

  5. Scott Says:

    John Figueroa #4: Good question! In reality, deterministic prediction of the far-future course of human events is just flat-out impossible, for multiple reasons that include quantum randomness (i.e., the Born rule) and your lack of complete knowledge of the universe’s current quantum state. So the best you can do is probabilistic forecasts, which in the show gets identified with “Many Worlds.” As you say, though, this is true regardless of whether you believe Many Worlds or deBroglie-Bohm, as long as you accept standard QM.

    Hypothetically, if you could see the actual values of the hidden variables (which, in standard QM, you can’t), then in principle deBroglie-Bohm would yield deterministic predictions. And I assumed that was what the show must have been gesturing at with the original prediction algorithm (the one that didn’t work well). But of course, the fact that it wouldn’t work because you wouldn’t know the hidden-variable values would’ve been obvious to the engineers from the beginning! 🙂

  6. Andrei Says:

    “that can perfectly reconstruct the past, down to what Jesus looked like on the cross”

    Are you alluding to this, is the show alluding to this or was it just a coincidence?

  7. Scott Says:

    Andrei #6: It’s something that’s actually in the show. I have no idea whether it was a reference to the thing you mentioned, but I can imagine that the idea of reconstructing what Jesus looked like on the cross has occurred independently to many people over the past two millennia.

  8. Matt Reed Says:

    My favorite part of this show is the quantum computer prop, which was clearly inspired by a real superconducting QC. It has SMA cables and circulators and cold plates and heat exchangers (though no thermal shields, so apparently it doesn’t get cold).

    But yeah, based on the number of circulators, “no point in describing as a number” means “about 20”.

  9. Thaddeus Says:

    Scott –
    A little embarrassing to me for my first response on your excellent and oftentimes inspiring blog to be about Devs, but these are strange times. And I think you do a bit of disservice to some of the quantum mechanical philosophical questions that have, shockingly, ended up on an FX/Hulu miniseries.

    Obviously I agree with you that because of chaos theory, exponential complexity etc. that projecting even seconds every atomic position in a quantum simulation with any accuracy is practically impossible, but foundationally, is it possible, in principle? The technical issue is that Schroedinger’s equation is deterministic in every way, and the “Heisenberg cut” is still up for debate. The “many-worlds” view vs. something with true randomness (using the Born rule as you put it) is actually a meaningful distinction. The former speaks to those who believe that all of the universe is quantum mechanically coherent, including every observer in it; that decoherence (and hence the Born rule) is an illusion of complexity. This is separate from hidden variables, this is described as a theory too complex to predict the outcome of random experiments because all states of the observer and the observer’s environment must be included. In contrast there are those who somehow separate the observer from the coherent phenomena, possibly appealing to gravity or more mystically to consciousness, or what not. In the former case, the universe is fundamentally deterministic, just in a very, very, very complex space requiring a number of simulating qubits, well, so large in doesn’t really make sense to express it as a number. In this case the question is how an observer observes only one universe. If the universe is simulatable, the observer lives in that universe, coherenty. As one very insightful character puts it: the box and its observers are inside the box, and there’s another box inside that box . . . “uh oh.” I think casting this concept as categorically “false” goes too far.

    Anyway, although Greek Oracles through Asimov’s Foundation and many others have grappled with the “all-predicting” box before, the question of the Heisenberg cut is a deep one in much more modern quantum foundations, and I’d challenge you to find a singular piece of pop fiction that captures it anywhere close to as well as the writers of this show have. I, personally, am impressed.


  10. Henry Says:

    I think your critique that the many-worlds theory contradicts the premise of the show isn’t valid because the contradiction is the very point of having that theory in the show. It is the basis of the conflict that Forest undergoes.

  11. Pattern Says:

    On that last paragraph:
    with the benefit of hindsight, people often think “that’s obvious”. The issue* seems more like…
    1. There is a lot of information. Options for dealing with this: learn everything[1], or prioritize.
    2. If your choice of priorities didn’t include things that turned out to be very important later, you were surprised like everyone else. (This probably only works for the future; good luck with the past.)
    [1] How big would a quantum computer have to be to describe the world? If not completely, then good enough? How big would it have to be that it could run simulations fast, so it could run enough times to get good results?

    Once you establish multiple, seemingly reasonable, possible futures (via your model),
    seeing farther through time requires:
    looking down multiple paths (which split up your time)
    or picking one, and dealing with the consequences of being wrong.**

    So seeing the future (or the past) is like looking ahead in chess (or trying to figure out how the board ended up that way). People may think of computers as being good at that sort of thing, and future computers as being better, etc. (Alternatively it’s magic/”Hollywood” doesn’t try to filter sci-fi for accuracy – the future will do that just fine.) This could stem from not knowing how quantum computers work, or just be an extrapolation based on:

    1. We’re really good with computers/Computers really exploded
    2. (We know how to make) General computers scale/Maybe QCs are the next big thing.
    3. If we just figure out the secret to general quantum computers, that’ll scale./What are the limits? If you went back in time and told people what computers could do would they believe you? Who’s to say QCs won’t be able to do unbelievable things?

    On the one hand, maybe the way QM works means that’s like falling for a pyramid scheme, or a company that proposes to build a super lucrative upside down pyramid – if the top floors are the most expensive because they have the best views, then an inverted pyramid scales size with profitability! (The higher floors are more difficult to support to begin with, hence the limits on height to begin with, and the inverted pyramid structure antagonizes this problem.)

    But people do fall for pyramid schemes.
    Advances may invalidate the limitations we thought there were before.
    Do you make it big by playing it safe? (Whatever the answer, what do people think the answer is?)

    * of predicting the future
    ** Unless you have a wormhole to the future. (Unless a magic portal from A to B is the same kind of thinking as “quantum + computer = more calculating power, or being able to consider more possibilities. The Ultimate computer can figure out anything past or future! Okay maybe that’s too much, but – a quantum computer can consider more possibilities in parallel, and isn’t that the biggest limitation to figuring out the truth – considering more possibilities makes it easier to find the right one. Won’t the truth make the future (and past) seem obvious in advance?)

  12. Scott Says:

    Thaddeus #9 and Henry #10: No, my point was that regardless of the determinism of the Schrödinger equation, the location of the Heisenberg cut, or any other deep issues about QM—no matter what, there are going to be many possible futures and pasts compatible with all the data about the present available for feeding into Devs. And that’s all the show ever seems to mean by “many worlds” anyway: the negation of the belief in a single and completely knowable-by-Devs past and future! So in the sense that the show uses the term, “many worlds” is just straightforwardly correct; it’s not even a question of philosophy or interpretation. And to the show’s credit, its plot seems to acknowledge this!

    But the problem for the show is that, once you’ve acknowledged it, you’ve also acknowledged that nothing even remotely like the Devs machine that’s shown could exist. For the “branching of possibilities” wouldn’t happen only in rare moments, but constantly; it would destroy your reconstructions long, long, long before you’d gotten to Jesus on the cross. Likewise, there would be astronomically many possible reconstructions of Amaya (or girls who might or might not be Amaya), which would differ in far, far more than the exact positions of the hairs on her head or whatever. I stand by my view that this is a fundamental and even obvious tension that the show never resolves.

  13. Ashley Lopez Says:


    “What do you say: should we let this one slide?”

    Well, at least your original blog slogan seems to have worked.

  14. Craig Freeman Says:

    It is plausible. The grandfather parodox has been proven a myth. Haven’t you seen Doctor Strange? Haha, or was it Infinity War??

    In either scenario, it is possible for QC to create the entity demon without triggering possible fallout. The information that is presented about past, present, future is only a representational estimate of that variance and not the actual variance.

    So, in a sense, the information you’re perceiving is information that comes from the alternate timeline created after you’ve perceived the information.

  15. Scott Says:

    Craig Freeman #14: Have you actually seen the show? Because what you write doesn’t match what’s on it. The show has the characters “conforming to prophecy” like in a Greek myth, in a way that might be interesting to think about but that struck me as wildly implausible not only at a scientific level but also at an ordinary human one.

  16. fred Says:

    It’s simpler to just point out that any finite system can’t simulate itself from the inside, because the simulating machine (a part of the system) needs to take itself into account, which creates an infinite regress that needs infinite resources.

  17. James Cross Says:

    How about a somewhat related (in my mind at least) question?

    Could we have multiple possible pasts and no way of ever knowing, for sure, which one is the true one?

  18. Bunsen Burner Says:

    The problem I have with all such productions is that the technical/scientific geeks never act properly. They blindly accept things, never try to break shit, and generally never act how I’ve always known such people to act. You have a machine that can tell the fututre… great! Program some code to switch on a light in a second if and only if that light is not switched on in 1 sec in the future. See what happens!

  19. Miles Mutka Says:

    The show has been entertaining, often unintentionally, but seems like it could have been condensed. Stanley Kubrick (whom the makers evidently admire) would have made the same points in half the screen time. From a humanistic standpoint, the determinism is treated as that of a fictional character in a locked-down script.

    But if it were actual scientists who had access to a real devs machine, I hope they would use the machine in some more useful ways, like making backups of the library of Alexandria before its destruction for example. I would think the complete original works of Arkhimedes, Aristotle, etc. down to Zeno would be more interesting than some anonymous cave paintings.

  20. Scott Says:

    fred #16:

      It’s simpler to just point out that any finite system can’t simulate itself from the inside, because the simulating machine (a part of the system) needs to take itself into account, which creates an infinite regress that needs infinite resources.

    Yes, that’s another way of seeing the problem! The show even raises this exact issue in a later episode—with the “resolution” being that, well yes, that’s just how mind-bogglingly powerful the Devs system is. I suppose this suggests that the number of qubits that there’s “no point in describing as a number” was actually infinity? 😀

  21. Scott Says:

    James Cross #17:

      Could we have multiple possible pasts and no way of ever knowing, for sure, which one is the true one?

    I’d say that that’s not only conceivable, but is the actual situation that we’re in, at least epistemically (i.e., relative to all the knowledge that’s now available on earth).

  22. Scott Says:

    Bunsen Burner #18 and Miles Mutka #19: Complete agreement with all of your points!

  23. fred Says:

    By far the best show about computer science/engineering was “Halt and Catch Fire” (total run was 4 seasons, on Netflix).

    Not only they nailed the details of the 1980s personal computer revolution and the rise of the Internet, but the show was really about what it means to try and change the world, the dreams and disillusions.

  24. Anonymous Says:

    I know it’s a bit presumptuous to ask you a question in the comments section of your blog, but if you have time and find this question interesting:

    Given that fiction in these cases are normally based on things that could not possibly happen with our current tech / knowledge of physics (such as time travel, changing “destinies”, etc.), there pretty much has to be a jump in logic, or a falsehood somewhere in the science of this show.

    So if you were an advisor on that show, where in general would you place the logic jump, as to maintain some scientific integrity while also maintaining the integrity of the plot? For instance, would you have changed anything in the logic or fantasy tech on the Avengers: Endgame?

  25. Scott Says:

    Anonymous #24: I didn’t watch Avengers: Endgame all the way through (only the scene where they “explain” the time travel), so let me not comment on that one.

    For ‘Devs,’ I suppose the minimal change I’d want to make is simply to replace the phrase “quantum computer” by “hypertronium ubercomputer,” or something else that’s totally made up onto which the show’s writers could project all the ideas they wanted. I.e., I might kindly ask them to avoiding hijacking a real topic that people actually know something about, in order to seize an unearned veneer of plausibility for a basically unrelated fantasy! Just lean in to your absurd premise; don’t use the word “quantum” to try and plausible it up! 🙂

    (Incidentally, this is precisely why I find that logically tenuous sci-fi plots, involving time travel and the like, often work the best in comedy—Futurama, Austin Powers, Spaceballs, …—rather than in ‘serious’ dramas. Comedies have no need to conceal the sheer silliness!)

    If someone wanted to make a compelling fictional TV series that involved actual quantum mechanics … well, that’s an interesting challenge, but if it’s possible at all it would look very different from the ground up.

  26. Deepa Ramani Says:

    I read a lovely article in the nytimes recently, where they interviewed children of various ages on how they’re coping with their new circumstances holed up at home. The very last line of the article is an insightful comment by an 18 year old, who comes to the same conclusion you do in this post : nothing in life is promised.

  27. Len Yabloko Says:

    I like your blog and read it regularly. This time I just want to point a spelling error when you spelled “central concept” as “central conceit” when talking about Laplace Demon. Or may be I misunderstood?

  28. mjgeddes Says:

    It’s interesting that fortune-tellers mainly only ‘predict’ good things happening to people. And it seems that you can have fixed points here. If someone predicted that I’m going to win the lottery by taking 6 numbers {n1 n2, n3, n4, n5, n6 }, and I know the predictor is accurate, I would indeed rush out to buy the ticket with those numbers, thus fulfilling the prediction. If, on the other hand, the predictor told me a hitman was going to get me in Damascus , I’m not going to stay in Damacus, thus contradicting the prediction.

    Jacob #1

    The problem with Wolfram-style ‘theories of everything’ is that if you postulate that a near infinitude of possible outcomes can happen (infinitely complex possible outcomes from maximally simple rules), then it’s hard to see how you can predict any particular thing – the theory has zero information content unless additional assumptions are added , but then the real explanatory work is being done by these ‘extra’ assumptions, and not the simple starting rules.

    So Wolfram starts with some really simple rules, then he says after 10^400 time-steps or whatever there will be a branch that matches our universe, we have a near infinite number of possible branches to search through and ‘all’ we have to do is ‘get out and find it son’ 😀

    The more determinism you have, the simpler the theory right?

  29. Scott Says:

    Len Yabloko #27: My posts have typos more often than I care to admit, but in this case, I did mean conceit!

  30. Craig Freeman Says:

    Scott, I’ve seen a few episodes.

    Though, I’m more responding to this statement.

    You said: “Anyway, none of the above addresses the show’s central conceit: namely, that the Laplace demon can be made real, the past and future rendered fully knowable (with at most occasional breaks and exceptions) by a machine that’s feasible to build. This conceit is fascinating to explore, but also false.”

    I understand Laplace and his teachings, and I delve into studies of epistemology as well.

    Epistemology states a simple fact: Knowledge cannot be classified as a true belief without it being justified.

    If it can be justified, say with those ‘finer details’ you’re referring to, then it is possible to create as close to true past and true future as humanly capable.

    And, if you think about it, life is only %99.99 percent real. It’s the job of the QC to understand that .000001%. That tiny percentage is the true present reality simulated to the human perspective by Quantum Physic calculations, which we’ve only had little understanding of until very recently.

    I think what it signifies is that if you are to observe the fabric of reality, you are to first understand its nature, like atoms, quantum occurrences, chaos, and where these things descend from. We can now do that like we never could.

    If math is the study of life, then a quantum computer linked to the brain of an individual(s) is your new God.

    These people are prophetic in the sense that the reality they perceive is the reality brought to humans after the Laplace Demon is released. ????????

    In a sense, they are not human and never were.

  31. Filip Says:

    Jacob #1: IMHO, Wolfram’s theory has only one valuable piece: a graph that looks like the earlest cave paintings.

  32. mjgeddes Says:

    Filip, Jacob

    I did think there were some fascinating ideas in what Wolfram had to say, but these had nothing to do with simple automation rules. The automation rules seem irrelevant to actual useful explanation, as Wolfram himself vaguely seemed to acknowledge towards the end of his 20 000 word tome. He then seems to have made a rather incredible move, postulating a ‘rulial space’, where ‘all possible’ graph rewrite rules are realized – i.e., a multiverse where every possible computational rule is tried. Then he concludes that ‘you make your own reality’, in this mind-boggling excerpt:

    “There’s actually an almost infinite diversity of different ways to describe and experience our universe, or in effect an almost infinite diversity of different “planes of existence” for entities in the universe “

  33. David Says:

    Interestingly, Wolfram’s technical reports contain a claim relevant to our shared interests:

    > This [multiway formulation of QM] leads to an immediate, and potentially testable, prediction of our interpretation of quantum mechanics: namely that, following appropriate coarse-graining…the class of problems that can be solved efficiently by quantum computers should be identical to the class of problems that can be solved efficiently by classical computers. More precisely, we predict in this appropriately coarse-grained case that P = BQP…

    I’m of two minds about this. Their theory is precise enough to derive a clear, bold prediction about an unsettled question, which is a lot better than theories of everything usually manage (and much more than A New Kind of Science did!). That makes me think there might be something interesting to this, even with all the model-selection issues. On the other hand, it seems like they’re probably wrong on the concrete question, so IDK.

  34. Nick Nolan Says:

    After watching the last episode, I can speculate that the Devs-universe as a whole can be described as a holomorphic function.

    They can see into the past, into the future and replicate the whole universe starting from observing only a small local part. The function of the whole universe can be constructed from a part if you just measure it accurately.

  35. YD Says:

    Scott #20: If they built two of these computers, would they be able to convince a classical skeptic that their predictions (and postdictions) are correct? 😉

  36. Filip Says:

    David #33

    If Wolfram’s theory implies that P=BQP like that paper claims, then:

    (1) he’s right and the universe is a terrible boring place
    (2) his theory is wrong because it proves a statement without containing a technique strong enough to prove that statement

    I hate both cases.

    I’m of two minds about this. Their theory is precise enough to derive a clear, bold prediction about an unsettled question, which is a lot better than theories of everything usually manage (and much more than A New Kind of Science did!).

    Do you realize how big solving that “unsettled question” is? Have you looked at Ten Signs a Claimed Mathematical Breakthrough is Wrong? (I don’t know if P=BQP matters since he claims that he has a “beautiful theory” of all physics with one hypergraph technique ????) … Wondering if there are PSPACE and NP results too.

  37. Anonymous again Says:

    Scott #25: This makes sense, and according to your rules of engagement with science, I think another movie that did this pretty well (aside from the ones you’ve already mentioned) is Being John Malcovich. It holds hostage no real science concepts (the portal is just in a weird magical closet). And aside from that, its own rules are pretty consistent, unlike in Avengers where it doesn’t seem to hold its own assumptions to even some self-contained fantasy science. Avengers endgame was a good movie, though, although I find the references to actual science cringeworthy. Not sure if that’s by principle or pure aesthetic. In the Avengers, it was said that they can’t go back and meet their past selves, but Steve Rogers did, and nothing changed.

  38. Jair Says:

    These kind of stories always take some scientific premise and then approach it like it is magic or mysticism. I’ve always wanted someone to write a story that does the opposite – take a mystical premise and explore it scientifically. Like, some scientists discover a magic ring that turns people invisible. But instead of going on a quest to destroy it, they just start investigating it. How far down your finger, exactly, does it need to go before it turns you invisible? Does it work on animals? Why not? What about microbes on it? And why does it turn your clothes invisible too? How does it know what counts as your clothing? Maybe the magical essence would get tired of all your probing at some point and refuse to cooperate. Very embarrassing if you’ve already scheduled the press conference about your findings.

  39. David Says:

    Filip #36

    I don’t think they have anything for classical computers, for them P=BQP falls out of their treatment of QM. I think that makes sense as far as it goes; if you have a theory of everything, you _should_ be able to use it to resolve P=BQP one way or the other. It would be weird if you couldn’t! But one man’s modus ponens is another’s modus tollens: I think probably P!=BQP, so I’m (even) less inclined to believe that they’ve got something worthwhile.

    I just like it better than if they weren’t able to make that prediction in the first place, if that makes sense.

  40. Anonymous Says:

    Jair #38

    That sounds a lot like the plot of a Ted Chiang story called “Seventy-Two Letters”. It is set in a world where golems can be animated by inscribing them with particular magical symbols. From the second paragraph: “It wasn’t the sculpting that Robert enjoyed; it was mapping out the limits of the name. He liked to see how much variation he could impart to the body before the name could no longer animate it.”

    Much of the plot hinges on certain details of how this system works. But like most Ted Chiang stories, it has other interesting points to make. If you’re interested, you can find it here:

  41. Scott Says:

    Everyone: I’m really not looking forward to using my severely limited quarantine time to look at the Wolfram thing, but—exactly like what happened with A New Kind of Science 18 years ago—it might reach a point where I feel I have no choice.

    In the meantime, though, let’s not derail this thread from the important topic of a fictional TV miniseries about a holomorphic universe! (I.e., a universe wherein every part is perfectly predictable from every other part … at least until you hit a pole.) Which mistakenly bills itself as a miniseries about quantum computers.

  42. arch1 Says:

    Jair #38: This reminds me of a discussion in one of Feyman’s books about skepticism. His example was claims of mind control over a roulette ball. He asks what you do if you observe statistically significant behavior ostensibly due to mind control, and try everything under the sun to rule out other effects, errors, subterfuges, etc., and many people do this in many ways, and the effect doesn’t go away. His answer is that you ultimately believe that it’s real, and study the conditions under which it occurs – does it diminish with distance, can certain materials block it, and so on.

  43. Zalman Stern Says:

    Filip #36:

    Per ‘Do you realize how big solving that “unsettled question” is?’ I’m guessing they are inserting a set of constraints that means they aren’t claiming to solve that question. For example, if one insists on a discretization of the world such that the model of quantum mechanics is effectively simulating a quantum computer on a classical one, then it would be a given that the power of this constrained computer is no better than that of a classical one.

    Put another way, the question is whether they’re redefining what BQP means too much. I decided not to put the time in on any of this stuff so I am not saying one way or another, just framing the discussion.


  44. Chris Says:

    A show that has impressed me with the depth it takes sci-fi concepts to is Rick & Morty. The show regularly revolves around plots involving the implications of multiple worlds, AI, time travel, paradoxes, etc. There is a lot more to it than I expected. The episode where Jordan Peele and Keegan-Michael Key are the voices for time police characters is especially memorable.

  45. anon Says:

    Scott – I think the line about the number of qubits in the machine is supposed to be a bit tongue-in-cheek as well as to point out that the CEO isn’t a genius, or really very knowledgeable at all. My interpretation is that the show is a bit of an attack on tech culture, and in particular the fallacy of the genius tech CEO as savior. Alex Garland, the guy who wrote and directed the series also wrote and directed the movies Ex Machina and Annihilation, and while we are certainly dealing with science fiction here, I think that using the word quantum computer as opposed to say “hypertronium ubercomputer” is a good thing. These stories are thought experiments that happen in fictional worlds that resemble our own, which makes them more compelling and interesting. Staying away from scientificky sounding gobbledy goop is another thing that makes good science fiction compelling. Another one of the biggest merits of Garland’s work is that he makes great effort to be as scientifically true to our real world as he can be within the confines of the fictitious worlds he creates. So for example when the supposed QM expert in the show, Katie, talks about how there are no random events (lol) it is painful for the initiated, but it’s not really the point of the show. I don’t think it’s hijacking a concept to make a movie about a thinking robot (Ex Machina) or an alien like no other (Annihilation) simply because there are experts in AI or molecular biology who know things about those subjects as well. And if he is going beyond what other producers of science fiction do in creating these worlds, by consulting experts, and by giving it a veneer of plausibility, it’s certainly not “unearned”. I think that effort implies the opposite. Also, there will always be numbskulls who watch fictional works and assume they are now experts in a field simply because the fiction made it plausible to them, but one much more interesting and really great consequence is that an uninitiated person might become interested in something they would never have had any exposure to otherwise. I’d suggest watching Lex Fridman’s interview with Garland for some context. He’s not a dummy. PS I think Lex’s interviews with both you and Garland were awesome.

  46. Scott Says:

    anon #45: On reflection, you’re right. I was wrong. Viewers of ‘Devs’ will come away with an extremely accurate impression of what quantum computers will and won’t be able to do, and why I and others want to see them built. In fact, as an attempted popularizer of QC for the last 18 years, ‘Devs’ will save me a ton of effort, since people who’ve seen it will know so much about the subject already. And if this sounds like the opposite of what I was saying just before … well, I wasn’t responsible for what I said before, and I’m equally not responsible for contradicting it now. We live in a deterministic universe, wherein every atom completely constrains every other atom. The world runs along tram lines. Click. Clack. Cause. Effect. I, like you, am just another billiard ball posting whatever blog comments I was fated to post.

    —Parallel Scott from Devs world

  47. Anon Says:

    No, it sounds like the opposite of what I was trying to say couched in bullyish sarcasm. I have complete respect for your work and for the work of scientists everywhere – I am a scientist, as it turns out. I was trying to point out that narrative science fiction is better (compelling, interesting, having wider appeal) when it resembles our reality, even if it is completely contradictory or inaccurate scientifically, than when it’s full of Star Trek type technical jargon, and I say that as a total Trekkie. You admit that you didn’t watch the rest of Endgame but did watch all of Devs to date. Garland is doing more of a service than a disservice to the scientific community by popularizing these ideas. And that is not a discredit to those enduring bright minds such as yourself that have endeavored for years to move beyond rote journal articles to do something truly original in communicating science beyond it’s cloistered communities. Sure we can still hope for better work from the fiction writers, but I respect his attempts to mainstream the “hard science” sub genre of science fiction. I should have pointed out in my comment though that you yourself say this show isn’t much about quantum computers, as it is a recapitulation of an old narrative technique, which is an idea I was echoing in saying that the technical points can be bothersome. I get that it hits close to home, so I’d probably be somewhat bothered if someone made a show about monoamine transporter structure/function relationships and their consequences for monoaminergic neuronal signaling and got some key facts wrong. That would be a terrible show, though, whereas Devs isn’t all that terrible. Is Devs really that much more of a bastardization of it’s scientific subject matter than the average pop science article on QC? (Exclude articles for which you were interviewed from your consideration:p)

  48. YD Says:

    anon #45: I haven’t seen Ex Machina, but Terminator was a great movie because it brought the ideas of AI safety to the masses. /s

  49. Richard Gaylord Says:

    Jacob Comment #!:

    for anyone interested in Wolfram’s model of the universe, here is where to go to learn more:

    the general URL:

    papers on the QM and GR properties of the Wolfram model :

    “Some Relativistic and Gravitational Properties of the Wolfram Model” by Gorard at

    “Some Quantum Mechanical Properties of the Wolfram Model” at

    the forthcoming book by Wolfram (due to published on 5/29) :

    “A Project to Find the Fundamental Theory of Physics”

  50. fred Says:

    Scott #46

    “I wasn’t responsible for what I said before, and I’m equally not responsible for contradicting it now. We live in a deterministic universe, wherein every atom completely constrains every other atom.”

    The funny thing is that this view is actually the one that’s consistent with not just science but also with subjective experience (given a actual serious introspection of our own consciousness).

    What’s the alternative?

    What do we mean by “choices” when the atoms we’re made of don’t have any capacity for choice?
    Are we magical sources of causality that are disconnected from everything else, and for which we can take all the credit and/or all the blame?
    Where do we draw the line beyond which that freedom/responsibility stops? The whole universe, Earth, our body, our brain, individual neurons?
    Does our brain control its own neural connections? (a program that rewrites its own source code is still a program)
    What and where is this mythical thing we call “I”, the supposed unchanging truth of our essence?
    Do we choose to be born as ourselves? Why am I living my life instead of the life of Scott Aaronson?

  51. gentzen Says:

    The bottom of Fig. 2 in Teaching Quantum Computing to High School Students is interesting: “The bottom left plots shows the perceived importance of quantum computing before (blue) and after (orange) taking the module. The bottom right plot shows the distribution of interest in learning more quantum computing (5 indicates the maximum level of interest).” The plots make it pretty clear that both the perceived importance and the interest in learning more quantum computing went down after taking the course. The text (on page 4) manages to give this outcome a positive spin:

    The distributions of perceived societal importance of QC are shown in the bottom left plot of Fig. 2. Using our metric where 1 (5) indicated the minimal (maximal) importance of QC, the median of the before and after data is 4. The change in the distributions highlights that the students appreciated that QC technology is in its infancy, and much of the “hype” generated by popularmedia is hyperbole. Feedback from the students supports this understanding: “Sci-fi and science magazines make quantum computing seem more useful than it currently is. . . ” This is not to undermine the enormous potential of QC, and many of the students showed appreciation for the ongoing research efforts: “I had zero knowledge about quantum computing before this project, and I now have a broad understanding of why so many scientists are invested in quantum computing.”

    The general enthusiasm of students in pursuing further QC learning is shown in the bottom right plot of Fig. 2. Prior to completing course, the median was 4, and this did not change after completion. The change in distribution demonstrates that students successfully learned about QC and some wished to continue studying QC further at university while others learned enough to satisfy their curiosity. This also highlights the effectiveness of the module.

    But in the end, you have to admit that this is what happens if you inform people without hyperbole: their interest in the subject goes down.

  52. Rainer Says:

    Richard Gaylord #49

    I only skimmed over his web site . All the text is simply TL;DR for me.
    My first impression (which might be wrong) is that it is kind of a graph rewrite engine.
    This sounds to me like a Turing machine in disguise.

  53. Bennett Standeven Says:

    I hope I won’t derail the discussion too much if I mention one interesting prediction of Wolfram’s model: he defines an observation as a “locking” of the time flow around a gauge variable and argues that, since this locking will expand over time, the observation will eventually need to be released (to prevent the complexity of the computation from growing to unmanageable levels). So if we perform a measurement of some ‘quantum’ system, the measurement result should eventually vanish!
    In fact, we can also deduce this from the fact that the classical ‘worlds’ corresponding to different pathways in the gauge space differ only in the choice of foliation (because of causal invariance), and hence can’t have different matter distributions, which they would if any measurement of quantum phenomena was involved.

  54. Scott Says:

    gentzen #51:

      But in the end, you have to admit that this is what happens if you inform people [about QC] without hyperbole: their interest in the subject goes down.

    Please amend that to: it’s what happened when they tried to inform people about QC without hyperbole. 😀

  55. exmachina Says:

    Sorry Scott, i’m a physicist *and* a big fan of your blog, but I don’t understand this bashing Devs. I take you were having a slow day and stressed about CoViD 🙂

    Devs is a /tale/ , is not a science documentary. People do want to hear /tales/ , it’s in our nature of aboriginal cavemen gathered around the bonfire at night, listening to the stories and legends of ghosts and spirits. You don’t need to believe that the technology depicted in Devs corresponds to reality in order to enjoy the /tale/ .

    I push further to say that it is actually a /useful tale/ as much as Star Trek has been a useful tale for generations of scientists nowadays. Because if you’re a parent and you think “oh wow, this quantum Devs computer is really cool stuff, I wish I knew more about it” then it doesn’t matter that reality and fiction are distant: it is more likely that you will encourage your children to take a career in science.

    In this respect, I think calling the Devs machine something different from “quantum computer” would surely make it less annoying for a scientist, but would also make quantum computers less fascinating for the non-scientists. Given the ratio of scientists to non-scientists in the general population, I think the Devs /tale/ serves a Greater Good by abusing the language the way it does.

    Someone could object that I am advocating for using “lies” in order to lure more people into science. Yes, I am, and I’m not ashamed of it. Because I think that a general population that on average understands more the concepts of exponential growth, evolution, clathrate gun, and so on leads to a better world for everybody. I do believe that ends justify means here, also because the alternative is to leave the “lie business” to preachers, wizards, politicians, and supremacists. Moreover, unlike in these cases, a “sci-fi lie” is just a short-term solution that is actually a lesson: you need to stop believing in lies at some point if you really want to understand how things work.

    That said, if Devs is guilty of lying on the tech, it’s a pretty minor offender in the whole sci-fi genre.

    I also enjoyed a lot Ex Machina.

  56. Raoul Ohio Says:

    Anyone wanting to while away some time watching — plausible right now SF — that is highly entertaining for lots of reasons, check out the (2013-2017) Canadian series “Orphan Black”, available on Amazon Prime. The first couple episodes will be totally WTF? before you start to figure out what is going on. Despite main characters narrowly (usually) escaping death regularly, it is at times very funny, and the science is pretty good. Some of the scientists are likable, regular folks (plus a couple evil ones).

    SPOILER ALERT! Stop reading here if you want to try to figure out what is going on on your own.

    The overall story is that a number of women figure out that they clones, being observed in a giant experiment by an evil corporation, that, along with religious cults, and others, are murdering them — what’s not to like? All the clones are played by Tatiana Maslany, who nails the totally different personas and lifestyles.

  57. Scott Says:

    exmachina #55: But I didn’t bash Devs. I watched a show that doesn’t merely get a few details wrong, but that’s entirely about taking a steaming dump on everything that I’ve spent my entire life trying to get through people’s heads—e.g., that quantum computers are not magic oracles, that they’re interesting because the stock sci-fi plots that you already knew don’t map onto them, because they illustrate how the actual world is more imaginative than our tropes, and also, that the people who work on these topics are something like the characters on “The Big Bang Theory” but nothing whatsoever like the characters on spooky dramas—and I described the show on my blog as “not that bad” (because it wasn’t). Do you have any idea what an effort of will that took? 🙂

    Look, I’m going through a deep depression right now. Indeed, I’m finding it hard to understand anyone who isn’t depressed, given the terrifying state of the world, the morgues running out of room for more corpses, the collapse of people’s plans for their lives, the food deliveries that ominously no longer show up as the machinery of the world starts groaning to a halt, the clowns running wild in the control room, how easily this all could’ve been prevented but wasn’t, one’s own personal failure to foresee it. Why were we fated to be alive right now, to try to raise children now, right when the music of civilization finally stopped?

    So, as I’m staring out at the abyss, when people are like: “can you, for social reasons, pretend that the creepy Greek tragedy you just watched has a genuine connection with quantum computing?” Or: “can you please respond to the exciting news that, even as the walls close in, Wolfram is still around, continuing to say the same stuff he’s said for ~40 years?” … well, it just makes the abyss seem all the wider. I apologize if that colors my responses.

  58. Anonymous again Says:

    exmachina #55: Are you arguing that making fantasies out of science, and representing real concepts as mystical would actually increase societal scientific literacy and engagement? Because the opposite could very well be true as well – and probably is. Once people realize that it’s not magical like in the shows, they’re probably in for a rude awakening, but the real mysticism and beauty lies in GENUINE curiosity, not a fake carrot on a stick from movies. And it could backfire – “I heard QC’s could do such and such, so I’m against them” (as a possible example). I’m not a physicist – I studied physics but my ultimate goal was eventually to get into tech world & start a co., which I did – but I think the innate curiosity that led me to study it was there, and probably more incited by friendly people like Scott being willing to explain it how it *really* is. I remember reading those people growing up. The joy of discovery and understanding is a distinct animal from movie thrills, in my opinion.

  59. Aaron G Says:

    Scott #57, I’m sorry to hear that you are feeling a big depression given the COVID-19 pandemic.

    While the situation in the US is particularly serious, perhaps you can take comfort in knowing that other parts of the world (e.g. Taiwan, South Korea, New Zealand, and now increasingly Canada, the state of California, even hard-hit countries like Italy and Spain) have succeeded or are succeeding in slowing the growth in new cases (aka “flattening the curve”).

    Also that now there are so many groups around the world working diligently on developing COVID-19 vaccines, several of which are also starting early phase clinical trials. Or new antiviral treatments being developed as we speak.

  60. Thomas Shaddox Says:

    Minor quibble, but I’m pretty sure he vomits because he’s extremely nervous about committing espionage on behalf of a foreign government. Nervousness, which, it quickly turns out, was warranted.

  61. mjgeddes Says:


    Wolfram has moved on from cellular automata to hypergraphs, has to be an improvement right? With hypergraphs and graph rewrite rules, behold I give you… a universal computer!

    Of course, it’s doubtful any useful insight could be obtained from describing physics as an enormous string of simple computational rules, but who knows – the journey might lead to some interesting stuff. Wolfram’s other project (the quest for a ‘universal language’) seems a lot more promising though.

    If it’s a theory of everything one is after, you probably won’t find that in physics, but rather in trying to find a paradigm for unifying all the different branches of knowledge – i.e., a ‘universal language’. A TOE cannot be a specific theory, but it could be a general framework, i.e, a paradigm.

    Never fear Scott, I’m on the case!
    The ‘compass rose’ of knowledge will point us in the right direction

    N Computer Science (Theoretical)
    NE Information Technology
    E Social Science
    SE Engineering
    S Biology
    SW Physics
    W Cog-Sci
    NW Math (Pure)

    Center Complex Systems Theory

  62. Bunsen Burner Says:

    For me, it wasn’t even just the science that was a problem with Devs, though it did twist the QM a bit too far. The problem is the ridiculous quasi-religious psychobabble that went with it. I’ve never in my life heard scientists/philosophers/techies/business people talk like dimestore Yodas. Is it really that hard to create a TV series a bit more philosophically sophisticated? Why can’t dualism, essentialism and all the rest of that quack theology just die?

  63. Bunsen Burner Says:

    exmachina #55

    I’d like research as to the effect of bad SF on future career choices. And I say that as a scientist and a lover of SF! One thing that surprised me greatly when I started my academic career was how disinterested, and even hostile, most of the physicists I was working with were to science fiction. Finding it silly and childish. I also have to admit that over the years I’ve found the vast majority of SF lovers to be total science illiterates. Worse in many ways than the general public, because they think their interest in SF somehow gives them insight into complex scientific topics.

  64. Mitchell Porter Says:

    David #39 said

    ”if you have a theory of everything, you _should_ be able to use it to resolve P=BQP one way or the other. It would be weird if you couldn’t!”

    Despite the connotations of ‘everything’, in contemporary usage, theory of everything means primarily a theory of fundamental physics. And despite the fact that ‘quantum’ is part of the definition of BQP, BQP is fundamentally a mathematical (or computer-science) concept. One should not expect a theory of subquantum physics to resolve whether P=BQP, unless mathematics specifically relevant for computational complexity theory is inherently prominent in its derivation of quantum mechanics.

  65. Dmitri Urbanowicz Says:

    In particular, their idea seemed to be that the waves would reinforce at the prime factors p and q themselves, rather than at inverse multiples of the period of a periodic function that only indirectly encodes the factoring problem. (What do you say: should we let this one slide?)

    Well, if you include the classical part into the quantum curcuit as well, wouldn’t it be exactly as they say?

  66. fred Says:

    I just noticed that Devs is produced by Alex Garland, now I’m really interested to check it out.

    I loved Ex Machina and Annihilation.

  67. Anonymous again Says:

    I’d argue that the best work science fiction does is to explore human psychology (on a mass & individual level), by removing some limits of reality. It seems to me that its value is in actually *removing* the constraints of science from story-lines. It’s true that many conflate their love of science fiction with some sort of actual scientific literacy. Using actual terminology, rather than their own self-contained fantasies, would further that misunderstanding. I have seen so many Rick & Morty fans self-congratulate for being smart enough to understand it, but the show’s very own writers make fun of their ridiculous uses of “science” and sci-fi to explore the human experience in different ways, so I think a lot of the fans miss that point. “Quantum carburetor”? Jesus, Morty. You can’t just add a Sci-Fi word to a car word and hope it means something!”

  68. Aaron G Says:

    On the general topic of science fiction – I generally appreciate it more if the underlying isn’t too closely tied to existing technologies (for example, Star Trek or Star Wars), or if the supposed “science” is only hinted at or is secondary.

    From the description of “Devs” it sounds to me that the series tries to bring the story in a quasi-realistic setting, tying in actual areas of science (e.g. quantum computing) to move the story forward without understanding the science.

  69. I Says:

    Scott, would you like a summary of what Wolfram’s claiming now?

    Its basically just: graph-rewrite systems have natural interpretations that generate enough structure to replicate QFT+GR (the actual equations). Though that depends on the initial graph and rule, its surprisingly easy to re-cover swathes of physics, and it looks like chunks of mathematical physics (string theory).

    And this interpretation yields some insights (vacuum energy of QFT directed towards maintaining space, measurements must vanish,…). Wolfram is excited by how easy this all was (the concrete insights started coming half a year ago, and have kept coming).

  70. Rainer Says:

    I says #69
    I am very curious how the quite heavy tensorial equations of GR can be derived from simple graph rewrite rules. I think if that is possible then it will not be intelligible.

  71. Scott Says:

    I #69: Thanks. So, is this graph rewrite system classical or quantum? If it’s classical, can you explain in a self-contained and comprehensible way how it reproduces the Bell inequality violations? (ANKOS didn’t do this, despite Wolfram claiming that it did, and Wolfram never acknowledged the error in 18 years, so the skepticism level that needs to be overcome here is close to infinity.)

  72. Bennett Standeven Says:

    Scott #71:
    The graph rewrite system is classical. But the gauge space of foliations (ie rewriting orders) is “quantum mechanical” (a Rieman/Kaehler manifold, in other words; I don’t really understand how Wolfram is getting complex numbers out of it). So it should be possible to have entangled foliation-states. But being gauge variables, they are not observable; so there is no Bell inequality violation associated with them.

  73. I Says:

    Rainer #70:

    They ape the “Chapman-Enskog” hydrodynamic expansion used in deriving navier stokes for discrete molecular dynamics, which Wolfram used in “Cellular Automaton Fluids 1: Basic Theory” when deriving hydrodynamic equations for cellular automota. They assume the dimensionality of the graph (which they define by the growth of the number of nodes reachabl by traversing a r-edges e.g. r^n growth means dim. n. Call this N(r) ) is the same, but the curvature varies (corrections to the prior term).

    The key point here seems to be: generalise metric spaces w/ volume to measure spaces w/ probability measures. Using notions of distance and degree for hypergraphs, generate some measures that you can use to find the generalisation of Ricci scalar, the Ollivier-Ricci scalar curvature.

    Proceeding to use the natural notion of parrallel transport in graphs, treating geodesics as the shortest paths between nodes w/ all edges having distance 1, and edges are parrallel iff they share the same vertices, you can derive the Ricci tensor by looking at the volume of cones of causal graphs.

    From there it seems by taking averages over vertices of the contraction of the ricci tensor you can construct the discrete analog of Einstein-Hilbert action of the vaccuum, in the continuum limit.

    They define energy in terms of causal graphs: the flux of edges flowing through a timelike foliation of events. Momemntum is via a space-like foliation. Apparently this creates an alteration in N(r) and requires a correction to the action’s lagrangian.

    That should be it.

  74. I Says:

    #70, sorry got that wrong. They do not clam to have derived QFT, just QM. Yes, its all deterministic and classical. No, he and Gorard (a collaborator giving some detail and referring to the literature!) don’t acknowledge the issues with Wolfram’s prior model for entanglement and QM

    They say they’ve moved onto a new model which is non-local and proves CHSH in the same way as de-Broglie Bohm or other deterministic non-local theories. They don’t detail how that works.

    Since they don’t seem to have answered your objections explicitly, and their derivations aren’t fully rigorous, there’s decent odds somethings gone wrong Scott.

    Though Bennett Standeven might be better placed to comment on the QM stuffs rigour.

    By the way, here’s the papers with the technical details:

  75. Bennett Standeven Says:

    As far as I can tell, they are trying to define an observable as a linear superposition of rewrite rules. So from the “multiway invariance” condition (and the fact that causally unlinked rewrite rules commute automatically), it follows that every observable commutes. Which means most of their discussion of QM is nonsense.

  76. fred Says:

    Bunsen Burner #62
    “Why can’t dualism, essentialism and all the rest of that quack theology just die?”

    Erwin Schrodinger noted:

    What puts one off when examining what are called objective, historical accounts of ancient or modern philosophy, is that one keeps finding such statements as: A or B was a ‘representative’ of this or that view; so-and-so was an X-ian or a Y-ian, holding allegiance to this system or that, or partly to one and partly to another.
    Different views are almost always opposed to each other as though they really were different views of the same object. But this kind of account practically forces us to regard one or other of these thinkers, or both of them, as crazy, or at the very least as totally lacking in judgement. One is then very apt to start wondering how posterity, including oneself, can possibly think the ill-considered babblings of such blockheads worth any closer attention. But in fact one is dealing, at least in very many cases, with well-founded convictions of highly competent minds, and hence one can be sure that differences in their judgement correspond to differences in the object of it, at least in so far as very different aspects of that object were given prominence in their reflective consciousness. A critical account of their thought should, instead of stressing the contradictions between them, as is usually done, aim at combining these different aspects into one total picture—needless to say, without compromise, which can only lead to confused and hence a priori untrue statements.
    The real trouble is this: giving expression to thought by the observable medium of words is like the work of the silkworm. In being made into silk, the material achieves its value. But in the light of day it stiffens; it becomes something alien, no longer malleable. True, we can then more easily and freely recall the same thought, but perhaps we can never experience it again in its original freshness. Hence it is always our latest and deepest insights that are voce meliora.

    It was said by Epicurus, and he was probably right, that all philosophy takes its origin from philosophical wonder. The man who has never at anytime felt consciously struck by the extreme strangeness and oddity of the situation in which we are involved, we know not how, is a man with no affinity for philosophy — and has, by the way, little cause to worry. The unphilosophical and philosophical attitudes can be very sharply distinguished (with scarcely any intermediate forms) by the fact that the first accepts everything that happens as regards its general form, and finds occasion for surprise only in that special content by which something that happens here today differs from what happened there yesterday; whereas for the second, it is precisely the common features of all experience, such as characterise everything we encounter, which are the primary and most profound occasion for astonishment; indeed, one might almost say that it is the fact that anything is experienced and encountered at all. It seems to me that this second type of astonishment—and there is no doubt that it does occur—is itself something very astonishing. Surely astonishment and wonder are what we feel on encountering something that differs from what is normal, or at least from what is for some reason or other expected. But this whole world is something we encounter only once. We have nothing with which to compare it, and it is impossible to see how we can approach it with any particular expectation. And yet we are astonished; we are puzzled by what we find, yet are unable to say what we should have to have found in order not to be surprised, or how the world would have to have been constructed in order not to constitute a riddle!

  77. Jair Says:

    Anonymous #40: I’ve read that story and I enjoy Ted Chiang quite a bit!
    arch1 #42: Sounds somewhat similar yes! Except it would be interesting to consider an unambiguously “magical” mechanism. I guess the point is that even if such a thing exists, it would be amenable to the same scientific process as everything else, and would cease to be magic. Unless it completely defied logic itself.

  78. Anonymous Says:

    In Wolfram Model, the universe is an ever-growing finite graph, which obeys deterministic rules.

    He proposes that we may have a chance to know the exact configuration of this graph in an earlier moment and its update rule, so basically we have a chance to know everything about everything. I really hope, that this is the case.

    (Okay, I’m shy, I don’t like the idea that future civilizations may be able to reconstruct my life and my thoughts, but that is a sacrifice I am willing to make for the idea of Knowing All.)

    Besides that the whole idea is reassuring, likely to be provable, and it could solve every question about free will and epistemology forever.
    Basically like what Laplace did with his daemon, but this finite-but-evergrowing world behaves much better computationally.

  79. Hash Says:

    Lilly and Forest dead and they are in the Devs system now. How they get old and die in the “perfect” simulation?

  80. Bogdan Says:

    May I have an unrelated question about quantum computers? I am looking for a way to understand (and explain to students) the “Oracle Separation of BQP and PH” theorem in an easiest way.

    Ok, oracle A is just an arbitrary language, easy! BQP^A is easy as well: this is the set of languages recognised by quantum computer which can also ask extra questions in the form “does x belong to A”? P^A would be easy as well – same definition without the word “quantum”. Similarly, we can easily define P^(PH,A) – standard computer with 2 oracles. But what is PH^A? The “PH-powerful device” with oracle A? But what is the “PH-powerful device?” Low-depth alternating circuits? Too complicated!

    So, my questions are: Is P^PH = PH ? Is, for any oracle A, P^(PH,A) = PH^A? Is the main result in the “Oracle Separation of BQP and PH” paper is equivalent to the statement that there exists an oracle A such that BQP^A is not a subset of P^(PH,A)?

  81. Scott Says:

    Bogdan #80: Yes, PH=PPH. PHA just means that you get to specify a predicate with a constant number of alternating universal and existential quantifiers, at the end of which is a polynomial-time algorithm that gets to query an A oracle.

  82. Paul Chapman Says:


    Something which I couldn’t understand at first (and to which neither you nor Lindbergh refers) is the room with the strange metal arms hovering over a table, on which we see a dead mouse, a clock, and various other artifacts. Somehow the data collected from that room is aiding the computer.

    And I think those scenes are intended to solve one of the problems you point out: how is the quantum computer getting enough input?

    The computer (or the program it’s running) is *extrapolating* the (block) universe from that collection of objects. Ie, there is much, much less Information in the universe than we think (or calculate according to our understanding of the laws of physics). Physical extrapolation is possible in some ways. For example, if we have a perfect crystal of salt, we can extrapolate its orientation from the positions of a handful of atoms.

    I’m not arguing that the universe really *can* be extrapolated in this way, just that it’s a reasonable premise upon which to base the computer’s ability to have complete knowledge of past and future.

    And what the members of Devs are doing when they conjure up a view on their Big Screen is asking the computer (program) to do the calculations to extrapolate to that particular time and place. Ie, the computer’s memory never has a *complete* picture of the universe in its state (impossible anyway because it is embedded within it), but can ‘look’ at one time and place, and present what it finds (using aesthetically pleasing camera angles and movements).

    Similarly, in the final simulation, it doesn’t have to simulate Forest’s and Lily’s universe completely, but just enough for them to experience it as real. The moon really isn’t there if no one’s looking at it. And that falling tree makes no sound.

    Caveat: I may have fallen for the intentional fallacy here, of course.

    Cheers, Paul

  83. Bogdan Says:

    Scott 81 – Thank you! Below is my attempt to make formal what you said. I hope I did not misunderstood anything. I also hope that “$ latex $” (without spaces) will display the latex symbols correctly.

    Let $latex \Sigma=\{0,1\}$, $latex {\Sigma}^n$ be the set of words $latex \sigma_1\dots\sigma_n$ with each $latex \sigma_i$ being $latex 0$ or $latex 1$, and let $latex \Sigma^*=\bigcup\limits_{n=0}^\infty \Sigma^n$. Let $latex |x|$ denote the length of any word $latex x\in \Sigma^*$. Let $latex A:\Sigma^*\to \Sigma$ be an arbitrary function which we call oracle. Imagine a computer which can do everything a regular computer can do, but, in addition, can instantly compute $latex A(x)$ for any $latex x\in \Sigma^*$. Let $latex P_k^A$ be the set of functions $latex F(x_1, \dots, x_k)$ in k variables $latex x_i\in \Sigma^*$, for which there is a polynomial P and a program on such a computer which computes $latex F$ in at most $latex P\left(\sum\limits_{i=1}^k|x_i|\right)$ operations. Let $latex \text{PH}^A$ be the set of functions $latex f:\Sigma^*\to \Sigma$ for which there exists integer $latex k\geq 0$, polynomial P, and function $latex F\in P_{2k+1}^A$ such that $latex f(x)=1$ if and only if $latex \forall y_1 \in \Sigma^{P(|x|)}\,\exists z_1 \in \Sigma^{P(|x|)} \dots \forall y_k \in \Sigma^{P(|x|)}\,\exists z_k \in \Sigma^{P(|x|)}$ such that $latex F(y_1,z_1,\dots,f_k,z_k,x)=1$. Let $latex \text{BQP}^A$ be the set of all functions $latex g:\Sigma^*\to \Sigma$ for which there is a polynomial P and a program on a quantum computer (again with extra ability to instantly compute A) which computes $latex g(x)$ in at most $latex P(|x|)$ operations. The Theorem proves the existence of oracle $latex A:\Sigma^*\to \Sigma$ such that $latex \text{BQP}^A \not\subseteq \text{PH}^A$.

  84. Albert Veli Says:

    Hello Scott. I haven’t read your blog in a while so could you give a short update on where you think we stand regarding an Artificial General Intelligence? Do you still think it is at least 100 years into the future before we need to worry about singularity?