Monday, February 5, 2018

Aaronson gets on the Many-Worlds bus

Quantum computer complexity theorist Scott Aaronson has come out of the closet as a Many-Worlds supporter:
I will say this, however, in favor of Many-Worlds: it’s clearly and unequivocally the best interpretation of QM, as long as we leave ourselves out of the picture! I.e., as long as we say that the goal of physics is to give the simplest, cleanest possible mathematical description of the world that somewhere contains something that seems to correspond to observation, and we’re willing to shunt as much metaphysical weirdness as needed to those who worry themselves about details like “wait, so are we postulating the physical existence of a continuum of slightly different variants of me, or just an astronomically large finite number?”
This surprises me. H is the one who always insists on describing quantum mechanics in terms of probabilities. But in MWI, there is no Born rule, and there are no probabilities. All possible outcomes occur in different universes, and there is no known way to say that some universes are more probable than others.

Tim Maudlin has criticisms in the comments. He prefers the de Broglie Bohm interpretation, aka pilot wave theory, aka nonlocal hidden variable theory, aka (as he prefers) something with Bell's nonlocal beables.

Scott tries to answer why so many physicists have signed onto something so ridiculous:
Pascal #67:
There was a time when MWI was considered completely outlandish, but now it seems to be taken much more seriously.
What do you think caused this change in perspective?
Interesting question! Here are the first seven answers that spring to mind for me:

1. The founding generation of QM, the generation that had been directly influenced by Bohr and Heisenberg, died off. A new generation of physics students, less under their influence, decided that MWI made more sense to them. (You may want to read Max Tegmark’s personal account of his “conversion” to MWI, as a grad student in Berkeley, in his book. I suspect hundreds of similar stories played out.)

2. The quantum cosmologists mostly signed on to MWI, because Copenhagen didn’t seem to them to provide a sensible framework for the questions they were now asking. (Did the quantum fluctuations in the early universe acquire definite properties only when we, billions of years later, decided to measure the imprint of those properties on the CMB?)

3. David Deutsch, the most famous contemporary MWI proponent, was inspired by MWI to invent quantum computing; he later famously asked, “to anyone who still denies MWI, how does Shor’s algorithm work? if not parallel universes, then where was the number factored?” Anyone who understands Shor’s algorithm can give sophisticated answers to that question (mine are here). But in any case, what’s true is that quantum computing forced everyone’s attention onto the exponentiality of the wavefunction—something that was of course known since the 1920s, but (to my mind) shockingly underemphasized compared to other aspects of QM.

4. The development of decoherence theory, which fleshed out a lot of what had been implicit in Everett’s original treatment, and forced people to think more about under what conditions a measurement could be reversed.

5. The computer revolution. No, I’m serious. If you imagine it’s a computer making the measurement rather than a human observer, it somehow seems more natural to think about the computer’s memory becoming entangled with the system, but that then leads you in MWI-like directions (“but what if WE’RE THE COMPUTERS?”). Indeed, Everett explicitly took that tack in his 1957 paper. Also, if you approach physics from the standpoint of “how would I most easily simulate the whole universe on a computer?,” MWI is going to seem much more sensible to you than Copenhagen. It’s probably no coincidence that, after leaving physics, Everett spent the rest of his life doing CS and operations research stuff for the US defense department (mostly simulating nuclear wars, actually).

6. The rise of New Atheism, Richard Dawkins, Daniel Dennett, eliminativism about consciousness, and a subculture of self-confident Internet rationalists. Again, I’m serious. Once you’ve trained yourself to wield Occam’s Razor as combatively as people did for those other disputes, I think you’re like 90% of the way to MWI. (Again it’s probably no coincidence that, from what I know, Everett himself would’ve been perfectly at home in the worldview of the modern New Atheists and Internet rationalists.)

7. The leadership, in particular, of Eliezer Yudkowsky in modern online rationalism. Yudkowsky, more so even than Deutsch (if that’s possible), thinks it’s outlandish and insane to believe anything other than MWI — that all the sophisticated arguments against MWI have no more merit than the sophisticated arguments of 400 years ago against heliocentrism. (E.g., “If we could be moving at enormous speed, despite feeling exactly like we’re standing still, then radical skepticism would be justified about absolutely anything!”) Eliezer, and others like him, created a new phenomenon, of people needing to defensively justify why they weren’t Many-Worlders.
Wow. I think that Scott is mostly correct about these reasons, but it is surprising to see an MWI-advocate admit to them.

It is bizarre that radical rationalist atheist skeptics have somehow bullied mainstream physicists into believing in parallel unobservable universes. How does that happen? Yes, I know Bohr is dead, but can't anyone else fill in for him?

400 years ago, heliocentrism had the same predictions as the alternatives, with the heliocentrists starting to find some technical advantages. MWI does not make any quantitative predictions, and does not have any technical computational advantages. Some argue that MWI has a philosophical advantage in that it eliminates the observer, but that hasn't really given us any new physics.

MWI is still completely outlandish. It is amazing how many otherwise intelligent men have been sucked in by it.

Update: There is some technical discussion on Aaronson's blog about whether Born's rule can co-exist with Bohm's mechanics and with MWI. A couple of comments do a good job of explaining why failing to predict probabilities really is a fatal blow to MWI.

This issue goes right to the heart of what science is all about. Copenhagen quantum mechanics does a great job of predicting experiments, and became very widely accepted. Then comes MWI, which doesn't predict anything in our universe, but predicts all sorts of wild fantasies in unobservable parallel universes. So many of the smart professors jump ship, and endorse MWI? Weird.

Update: Lubos Motl piles on Aaronson, as usual.

Motl credits Aaronson with being about 80% correct, expecially when he (Aaronson) thinks for himself. Aaronson correctly explains what's wrong with the pilot wave and transactional interpretations. The big remaining issue is Copenhagen versus MWI.

That big issue goes right to the heart of what science is all about, and Motl explains it well. He creates an analogy of MWI to creationism (as Copenhagen to biological evolution). These creationism analogies get tiresome, but he makes a good point. MWI adds a huge belief structure way beyond what there can ever be evidence.

Update: In a later posting, Aaronson says:
Which interpretation of QM you espouse (e.g., MWI, Copenhagen, or Bohm) has no effect—none, zero—on what you should predict about the scalability of quantum computation, because by explicit design, all interpretations make exactly the same predictions for any experiment you can do on any system external to yourself.
This is contrary to the opinion of others like David Deutsch, who say that the many-worlds interpretation is what justifies quantum computing.

1 comment:

  1. Scott is being a good sport here and telling the truth about quantum computers. They can't work without MWI! RIGHT! There is something really superficial about quantum mechanics and we have really made no progress toward understanding quanta. To me, that's why people are going MWI. I still think it's basically crazy and overly classical (hell, even space-time is a presumption of the math) but shut up and calculate is an incredibly stupid attitude toward the world. Neo-Copenhagen thinking along Bayesian and psi-epistemic lines of thought seem to be aiming for a realism but not traditional realism. I like this direction. However, science is advanced through experiment. There have been real experiments limiting the scope of certain theories, so we clearly can go farther than QM in debates about foundations. Motl is a one-hit-wonder nobody and Scott has more meaningful publications to his name. Motl just rants but offers no real argument. He's a crank without curiosity, except when it comes to totally bullshit theories you can't test.