Friday, April 18, 2014 ... Français/Deutsch/Español/Česky/Japanese/Related posts from blogosphere

David Mermin on Quantum Bayesianism

Many physicists, when they get older (and, in some unfortunate cases, long before that), have the tendency to reduce their powerful brains back to the era of Newton or ancient Greece and "undo" their knowledge of quantum mechanics. People like Gerard 't Hooft – and even, to a much lesser extent, Steven Weinberg and Leonard Susskind – start to pay lip service to fundamentally deluded ways to squeeze the laws of quantum mechanics into the straitjacket of classical physics, using one (or many!) of the several popular, comparably misguided strategies: hidden variables of one kind or another (including the Bohmian pseudoscience), collapse mechanisms (including GRW), the many-worlds interpretation (with some parallel universes that "really exist" just like other planets), and others.

David Mermin is an optimistic counterexample. His views have been evolving. Mermin is the actual originator of the "shut up and calculate" dictum often attributed to Feynman. But I think that when he said it for the first time, he coined it in order to humiliate Feynman's – and generally orthodox – positivist attitude towards these questions. Over the years, he got fully converted to the Copenhagen school's positivist, intrinsically subjective understanding of the quantum phenomena. He learned how to love Bohr. He realized that Einstein was just wrong in his debates with Bohr, and so on.

Now, he published a piece in Nature

Physics: QBism puts the scientist back into science
which argues that the scientist (observer) cannot quite be eliminated from the scientific process, as classical protophysics and physics has been – for a very long time, successfully – assuming. Instead, since the switch of physics to quantum mechanics, physical theories have been predicting probabilities of some particular experiences by the observers which means something about the relationship between the observer and the external world.

When experiments with the same initial conditions are repeated many times, we may interpret probabilities in the frequentist way – as a fraction of the repetitions that lead to a desired outcome. But before a single repetition, we always care about the probability because it quantifies our subjective belief that the outcome will obey a condition. Before a single experiment, the probabilities are Bayesian ones. The "collapse" is just a change of our expectations – subjective Bayesian probabilities – about our future experiences. It happens in our heads, and so on.

The same thing has been said by your humble correspondent many times and Mermin says it in new ways.

The particular buzzword "Quantum Bayesianism" (or "Qbism") is meant to describe this 2001 preprint (and 2002 PRA paper) by Carlton M. Caves, Christopher A. Fuchs, Ruediger Schack. It describes probabilities in quantum mechanics in the way they are. Your humble correspondent would probably agree with all papers in literature presenting Qbism. I just have a trouble with the "credits", with the claim that "Quantum Bayesianism" is some really new 21st contribution to physics (and also with the prominent role that is given to Thomas Bayes). It is really just a new brand that describes the very same thing that the Copenhagen school understood well. They just didn't expect that the meaning of probabilities in quantum mechanics – which is really simple and obvious for anyone who is not prejudiced – would remain a source of controversy among professional physicists for at least 90 years so they didn't write long essays and they weren't inventing new words to describe the same thing.

The 2001 paper has about 200 citations but relatively to the huge traffic in the industry of deluded, classical misinterpretations of quantum mechanics, it is not enough. So in a 2013 poll about the foundations of quantum mechanics, "Copenhagen" got over 40% while "Qbism" only received 6% of the votes.

Mermin nevertheless believes that not only all the mysteries have been solved but because of his crusade that began when he became a professor emeritus, the booming industry of misinterpretations of quantum mechanics will evaporate. It would be nice if he were right and successful but I doubt it because the human stupidity knows no borders, limits, genders, races, ages, calendars, or nationalities. It is everywhere. Only two observable things are infinite, the universe and the human stupidity, and I am not sure about the former.

Mermin also recorded an interview. Click at this podcast link, press the "»" button ("next segment") three times, and you will jump directly to the conversation about the "subjective science".

He also claims that Qbism solves "the Now" problem in philosophy, a discussion between Einstein and Carnap on whether or not the special awareness of the present is accessible by physics. One may say that this vague problem is the same thing as the presentism vs eternalism battle. And even though relativity wants us to look at things "eternalistically" because the whole spacetime is the only invariant thing, quantum mechanics really overrides it and does give the present a special status – because it assigns a special role to the observers' experiences and those are attached to some particular moments (or periods of time). I think that not-quite-well-defined debates of this sort become meaningless philosophical babbling after 5 sentences or so which is why I try not to go over this limit. The Now problem may either be viewed as an analogy (one that even exists classically) of the "existence of subjective experience" or as a special example of the latter. In both cases, the problem is "reification" of something (fallacy of incorrect attribution of "real existence" to some auxiliary theoretical constructs such as the spacetime or the quantum state, in these two cases).

Add to Digg this Add to reddit

snail feedback (37) :

reader Mikael said...

Dear Lubos, I was also catching this Qbism debate on the internet and thought that it would be a suitable topic for the reference frame just as it happened with this article. Indeed for a close follower of the reference frame there is not much new and one finds the same position you patiently explained here many times. Besides you I always found David Mermin on of the most insightful sources on the topic.
I found this more philosophical speech (at the risk of babbling) from him at Perimeter which is perhaps not so interesting for those who are already clear about the topic but for those readers who are confused and look of an orientation.

reader Luboš Motl said...

Tx for the comment and link, Mikael!

reader Tom said...

Lubos, what a great line!: [Only two observable things are infinite, the universe and the human stupidity, and I am not sure about the former.] As someone who swam in the turbid waters of the military industrial complex, I briefed manager types many times on probabilistic subjects and very often used Bayesian techniques, and can testify that very, very few had any clue as to what I was talking about. So, forget QM, any probabilistic analysis beyond rolling dice is beyond all but the brightest, and throw in Bayesian stuff and you might as well be talking String Theory.

Unlike 99.9% of statistical work, where distributional assumptions are the norm, one actually derives the distribution in QM and, amazingly, it fits the empirical data. What more could someone want from a theory? You are surely correct about the meaningless philosophical babble in this area.

reader Luboš Motl said...

LOL, thanks, Tom, you have a good taste. The great line is - aside from some tiny modifications - due to Einstein. ;-)

reader Rehbock said...

Lubos, The opening paragraph reminded me of a cartoon and I found it was here.

reader Gene Day said...

The double-slit experiment alone is enough to prove that nothing is new here. One has to do one of three things in trying to explain it:
1) Give up in exasperation.
2) Descend into Bohmian or some other fantasyland.
3) Surrender to Bohr’s view, that science is only about observables.

reader physicsnut said...

interesting video.
GFR Ellis and Tony Rothman had a paper in 2009 quant-ph/0912.0808
"the arrow of time arises simply because the future does not yet exist".
Yes - Weyl "reified" spacetime.

reader Giotis said...

Sorry Lubos but for some reason I dislike any discussion about interpretation of QM. IMHO it is just a waste of time and completely boring. QM postulates are clear and known for decades now; at the end of the day this is all you need.

The only thing that troubled me recently (due to Papadodimas Raju paper), which is not related to interpretation of QM of course but it’s kind of a basic issue, is whether it is legitimate in QM to have state dependent operators.

reader Luboš Motl said...

Dear Giotis, it's really an uncontroversial thing if you think about it in a right way. Almost every operator we use in quantum theories - except for some symmetry generators i.e. conserved quantities - is state-dependent in the PR sense.

For example, take the system of 10^30 carbon atoms (this many carbon nuclei and this many electrons). Some of the states of this system approximately look like a cubic diamond. For these states, you may define the creation operator for a phonon propagating inside the crystal. Then you have a graphite. It also has some phonon operators but they act differently on the positions and velocities of nuclei and electrons.

This comment doesn't mean that observables aren't given by linear Hermitian operators. They always are and PR aren't changing anything about that. Instead, what they say is that the relevant operators that look local or admit simple enough measurement apparatuses etc. have matrix elements that is only "clear" for a certain neighborhood of a pure state - plus minus the action of several (other) simple operators.

So the claim isn't that any observable that can be measured fails to be a linear operator (which can act on any state, so in this sense, it is state-independent). Instead, the PR claim - as I would interpret it - is that the subset of the linear operators on the Hilbert space that may be viewed "simple" (as a creation operator for a phonon in a wave packet) is state-dependent. There is no universal (across the Hilbert space) way to define which operators are simple to measure and which operators are not.

As the diamond example above shows, it's not really a new observation in quantum gravity. In quantum gravity, we have string theory with a complicated configuration space (landscape). The spectrum of excitations heavily depends on where you are in this landscape, and so does the spectrum of "simple" operators.

In the present of black holes, they look like a mostly empty space (hole) which may be acted upon with local field (creation/annihilation) field operators but again, the right way to define these operators has to depend on the microstate (it is nearly constant in its vicinity) because the similarity of the physics of distinct BH microstates is a "coincidence". Fundamentally, the Hilbert spaces around particular microstates are "as different" as low-lying excitations around different vacua in the landscape, or as the phonon spectrum in diamond/graphite.

reader Giotis said...

Thanks for this answer Lubos, it helped me a lot.

reader Eclectikus said...

Quantum tunnelling and its applications to e.g. the tunnel diodes and scanning tunneling microscopes are also perfect samples to explain the role of statistics in quantum mechanics, and how we're able to use this (purely statistical) knowledge. And If I'm not mistaken, they are also good examples to explain the transition from probability density to Bayesian probabilities ... isn't it?

reader David Brown said...

I don't understand the reference to Steven Weinberg and Leonard Susskind. Have they in any way supported 't Hooft in his cellular automaton work?

reader tomandersen said...


Thought the comments were a little on the congratulatory side...

More Bohmian madness, you might say?

"Why bouncing droplets are a pretty good model of quantum mechanics"

I am fairly sure that no theorist predicted these effects before they were explored on the lab bench.

At any rate the videos are fun to watch.

reader superok4luv2u said...

Dear Lubos,As i understand the Qbism interpretation basically says that what we are calculating is the probability of the observer's personal experience of the 'outcome'. basically if we go all the way through the correlations with the system
|catalive>|BRAINhappy> or |catdead>|BRAINsad> Qbism says that we are ultimately calculating our perception of the sates of our brain ,But what does it even mean to say that our brain cells 'perceive' their states ? for me it doesn't make sense unless of course you say that 'perception and consciousness' is not a property emergent from the sates of our brain cells, so i am confused :(

reader Luke Lea said...

Abstractions are not real, only observations are real -- is that the takeaway?

reader Luke Lea said...

The material world is the place where we all meet and communicate. Somebody said that and I always liked it.

reader cobaan heula said...

Obat Diabetes wrong were a little on the congratulatory side

reader Luboš Motl said...

Hi, you are completely missing the point of all of this.

I *know* when I perceive something - everything else, whether considered real or not, is being converted my perceptions which are the only thing I really understand and can check. So the whole point is that the material world is just an intermediate result that gets transformed to what is the final result for a user of quantum mechanics or anyone else - his internal, subjective perceptions.

You are trying to treat the brain as just another external objectively real object which is exactly what is *wrong* for the interpretation of predictions in quantum mechanics.

reader Milkshake said...

But we don't need this subjective stuff in General Relativity. It would be interesting to compare what's different.

reader Luboš Motl said...

Dear Milkshake, the subjective character of science is needed in - and because of - quantum mechanics. So it's not needed in classical GR because it's classical, and it *is* needed in any quantum version of GR - like string theory - because it is quantum.

The difference between classical GR and quantum theories is that it is classical. Have I answered every micron of your question? I surely think so.

reader Luboš Motl said...

It's cute as a visualization of something but the mathematical equations aren't quite the same and the physical interpretation is completely different. Presenting this as a part of research to foundations of quantum mechanics is pure noise.

reader RAF III said...

Lubos - Once again you are tying yourself in knots by adopting the ancient, or medieval concepts and modes of thought of the philosophers. Just as in political arguments, if you accept the premises, presuppositions, language, and taxonomy of your opponents, then you will fail.

reader Luboš Motl said...

Where did I accept anything from my opponents?

reader superok4luv2u said...

Hello again lubos thank you for answering my question and thank you again for your clarification. So according to Quantum Mechanics perception or the mind exists independently of the material world including the brain but isn't that as stated in the article a sort of solipsism because nothing in quantum mechanics says that your perception has to agree with mine i could perceive the cat as being dead while you are perceiving it as being alive what i perceive exist only in my mind and my mind alone what i am ultimately sure of is the existence of my mind.
and i need to clarify i am in no way attacking that position infact i think it is what QM ultimately says i'm not comfortable to some of it's consequences but nature doesn't care about my comfort ,Does it?

reader RAF III said...

1)In the use the objective/subjective disinction and all the meanings and connotations (which can often change within the same sentence) they attach to it.
2)In the idea of probabilities in QM arising from incomplete knowledge.
3)In accepting that the character of science is subjective.

reader Luboš Motl said...

Sorry, these are not points of my opponents, they are my points. Fundamentally speaking, quantum mechanics probabilistically predicts *subjective* experience. I don't know how else you want to say this thing.

One may attach some extra wrong baggage to the subjective character of QM but it's his fault. One can't deny the subjective character of QM - or any other important insight of science - just because people like to add wrong things to these insights, too.

The uncertainty principle means that the knowledge is incomplete even when it's maximal (a pure state) and the rules of QM guarantee that it cannot be completed.

reader Luboš Motl said...

Dear Superok,

physics - quantum mechanics with particular Hamiltonians and models for anything - is needed to explain the relationships of the perceptions to anything else. One could argue that without these relationships, the perceptions would be meaningless or wouldn't exist.

But what QM says that you shouldn't try to reduce your feelings to something else. Mechanisms by which external events influence your feelings or perceptions may be studied and "reduced" to other pieces but the feelings themselves can't.

Quantum mechanics is subjective but it is not solipsist because everyone may equally use its rules. Some things are unavoidably subjective in the sense of dependent on the subject who is perceiving something or using QM - his perceptions are surely subjective - but the laws of QM may also be used to show the correlations that are common-sense and that were holding in the classical, objective world. But quantum mechanics stops short of confirming any objective world - objective reality is an emergent concept.

Nature doesn't care whether you like Her rules, indeed.


reader RAF III said...

They are your points (and mine too) because, like the Red Queen, the words you use mean exactly what you want them to - no more and no less.
However, when one uses words to which extra baggage has already been attached it is not just the fault of the listener.
Although I know what you mean, I find your last sentence above to be almost incoherent. The idea of 'incomplete knowledge' implies that there is some more knowledge to be had. If the uncertainty, which is not synonomous with incomplete knowledge, is intrinsic and unavoidable then it is objective.
And so on...
Anyhow, I don't want to argue about the meaning of words. I only wanted to point out that by using words which have a well established meaning and support ages old philosophical prejudices it is very difficult to argue against those prejudices. But this is just what is required. QM is incompatible with these notions, and in fact supersedes them.

reader Eugene S said...

Humpty Dumpty, not the Red Queen.

'When I use a word,’ Humpty Dumpty said in rather a
scornful tone, ‘it means just what I choose it to mean — neither more
nor less.’

reader Luboš Motl said...

There is nothing ill-defined or vague about these words of mine.

There is nothing incoherent about the fact that the wave function describes incomplete knowledge about the physical system. Everyone with the right (e.g. "quantum Bayesian") understanding of the state vector will tell you exactly the same thing.

For example, even the abstract of the paper that started the term "quantum Bayesianism"

says, among other things, "In the quantum world, maximal information is not complete and cannot be completed."

This has a completely well-defined meaning. The information in the state vector is maximal because it's not possible, even in principle, to find a physical system in a state that would allow one to know more than what we know in a pure state vector.

On the other hand, even this maximal information about the physical system is incomplete because it doesn't allow one to answer all physically relevant questions (about observables' values).

reader RAF III said...

Quite right! Thanks.

reader RAF III said...

If there is nothing ill-defined then why did you write both *subjective* and subjective? Do they have different meanings? What DO they mean?

The only thing I wrote about physics was that QM supersedes the conceptions of philosophy. Was I wrong?

You wrote - "It's exactly your completely wrong opinion that "incomplete
information" means that "it must be possible to complete it" that puts
you squarely in the camp of the anti-quantum zealots who are simply
incapable of giving up classical physics". But this is nonsense. I understand that 'incomplete' refers to observables (under the assumption that a purely probabilistic prediction arises from 'incompleteness'). I was rejecting the use use of the word 'incomplete' because it implies that there is some completion possible when there isn't. I was writing about the word and it's obvious meaning, not about QM.

Why say 'incomplete' at all? Why not write -"The information in the state vector is maximal because it's not
possible, even in principle, to find a physical system in a state that
would allow one to know more than what we know in a pure state vector."? As you did. Neither classical physics nor QBism can deal with complementary observables.

If you continue to use words and concepts whose meanings and connotations are at odds with the physics you're describing then you will only have yourself to blame when you are misunderstood.

QBism is just an attempt to 'interpret' QM within the radically subjective Bayesian philosophy of probability of De Finetti (and possibly Jeffreys). There is nothing about it which is particularly quantum mechanical or scientific.

reader Luboš Motl said...

The *asterisks* around a word means that the word is emphasized. Every other kid in the kindergarten knows that.

Again, you are wrong about a fundamental thing if you think that "every incomplete thing can be completed". It's just not true, in any context. For example, a theory may be UV-incomplete, and it may have a UV-completion but it doesn't have to have a UV-completion. The information about observables in the state vector is incomplete and it cannot be completed because it is already maximal.

Otherwise it's a waste of time to discuss with you. I have already clarified all your mistakes. Why don't you read what I wrote ten times before you touch your keyboard again?

reader RAF III said...

Alright, I have read all your comments ten times. My only point has been about using everyday and/or philosophical words and concepts in areas where they are not applicable or easily understood by those without knowledge of those areas. Could you at least acknowledge this?

reader Luboš Motl said...

No, I have used no philosophical words outside their range of validity. In particular, the word "subjective" may be said to be a philosophical one but it is the kind of elementary philosophy that is important for thinking humans in general and it is essential to formulate claims about the meaning of concepts in quantum mechanics.

The statement that "the state vector encodes subjective knowledge" simply means that the knowledge is accessible to a particular observer and it may be different for different observers - the opposite words is "objective" which is the same for everyone. Philosophers or people calling themselves this way clearly cannot have any monopoly over this word.

reader RAF III said...

Lubos - I think that you (and Mermin, in the Nature article) are advocating something that could be called QBism lite -
accepting aspects which are consistent with a sensible use of probabilities in science (and calling it the 'beef' of Qbism) while
ignoring the philosophical pronouncements which are essential to the views of Fuchs, Caves, et. al.. Perhaps your eyes
glazed over. I know mine do.

Fuchs would never accept a frequentist view of probabilities after repeated experiments. For him they are only and always
degrees of belief which do not refer to or depend on anything in the external world. They may change as a result of data and
the rules of probability theory (which they view as objective), but they are always (somehow!) only degrees of belief which
quantify one's willingness to bet on a future outcome. There is for them no relation between belief and reality. Here is a typical

In this paper we address the problem of certainty in the Bayesian approach to quantum
mechanics. We show that a consistent treatment of quantum probabilities requires that even
if a measurement outcome has probability 1, implying certainty about the outcome [21], that
probability has to be interpreted as a Bayesian degree of belief. This would be the case,
for instance, if the premeasurement state is an eigenstate of the measured observable. Even
in this case we maintain that the measurement has no preassigned outcome; there is no
element of reality that guarantees this particular measurement outcome [22, 23]. Certainty
is a function of the agent, not of the system.
in which they also make clear their disputes (which I find somewhat disingenuous) with the Copenhagen view.

One need not accept QBism when one rejects the idea that a state vector is a 'real , physical thing, so Qbism does not provide
the only means of preserving locality in physics as claimed here

I described QBism as radically subjective, and I believe that Fuchs would accept this characterization without taking
offense because it is accurate. Even Mermin notes in Nature that 'QBists are often charged with solipism', and this is not
without reason. His defence, however, is of QBism-lite, not of Qbism as advocated by Fuchs and Caves.

I can not view QBism as identical to the Copenhagen view, nor can I view statements such as 'there is no
element of reality that guarantees this particular measurement outcome' as scientific.

I did not use the word *controversial* so I don't understand why you put it in quotes. Likewise, I don't understand why you
failed to put quotes around the other things you just made up and attributed to me.

reader Luboš Motl said...

Apologies, RAF, I don't know what you're talking about.

The quote by Fuchs you wrote here is totally OK with me, and so was a talk by Fuchs at Perimeter that I have listened to.

None of these things imply that one can't interpret probabilities in a frequentist way after many repetitions of the same experiment. One can *always* interpret probabilities in this way. The Bayesian vs frequentist dispute is really about the Bayesian probabilities' being "stronger" and "existing" even before the frequentist probabilities are definable, and I agree with Fuchs et al. that before a single measurement, the Bayesian interpretation of the probabilities is inevitable.

Mermin says clearly why QBism isn't sollipsist - it's the same reason as mine: everyone can really use this framework, not just me, and the framework is still completely compatible with all the explanations of the qualitative similarity (and common origin) between me and other people.

I don't know what you want to achieve by the extra labels "radically" before "subjective". How does "radically subjective" differ from "subjective"? You are making up a dichotomy that doesn't exist. Given the fundamental nature of subjective observations in quantum mechanics, I would be eager to support the claim that "quantum mechanics is radically subjective", too. I don't even know what incorrect corollaries one could derive out of this claim.