Friday, March 13, 2020

Trying to make sense of superdeterminism

Sabine Hossenfelder writes:
It does not help that most physicists today have been falsely taught the measurement problem has been solved, or erroneously think that hidden variables have been ruled out. If anything is mind-boggling about quantum mechanics, it’s that physicists have almost entirely ignored the most obvious way to solve its problems.
Her "obvious way" is to replace quantum mechanics with a superdeterministic theory of hidden variables. It says that all mysteries are resolved by saying that they were pre-ordained at the beginning of the big bang.

Peter Shor asks in reply:
I don't see how superdeterminism is compatible with quantum computation.

Suppose we eventually build ion trap quantum computers big enough to factor a large number. Now, suppose I choose a random large number by taking a telescope in Australia and finding random bits by looking at the light from 2000 different stars. I feed these bits into a quantum computer, and factor the number. There's a reasonable likelihood that this number has some very large factors.

Just where and how was the number factored?
I agree with him that it is impossible to believe in both superdeterminism and quantum computation.

Quantum computation cleverly uses the uncertainties about quantum states to do a computation, such as factoring a large integer. If superdeterminism is true, then there aren't really any uncertainties in nature. What seems random is just our lack of knowledge.

If I could make a machine with 1000 qubits, and each can be in 2 states at once, then it seems plausible that a qubit calculation could be doing multiple computations at once, and exceeding what a Turing machine can do. (Yes, I know that this is an over-simplification of which Scott Aaronson, aka Dr. Quantum Supremacy, disapproves.)

But is the uncertainty is all an illusion, then I don't see how it could be used to do better than a Turing machine.

I don't personally believe in either superdeterminism or quantum computation. I could be proved wrong, if someone makes a quantum computer to factor large integers. I don't see how I could be proved wrong about superdeterminism. It is a fringe idea with only a handful of followers, and it doesn't solve any problems.

Update: Hossenfelder and Shor argue in the comments above, but the discussion gets nowhere. In my opinion, the problem is that superdeterminism is incoherent, and so contrary to the scientific enterprise that it is hard to see why anyone would see any value in it. Shor raises some objections, but discussing the issue is difficult because superdeterminists can explain away anything.


  1. To me, it's not that superdeterminism doesn't solve any problems, it's more that superdeterminism doesn't give us any productive, useful mathematics and physics.
    I suppose superdeterminism gives a more-or-less plausible solution for some problems that some people care about but that not many pragmatic physicists much care about.

  2. But [if] the uncertainty is all an illusion, then I don't see how it could be used to do better than a Turing machine.

    It can't, assuming it had to solve the full factoring problem using exactly the same assumptions on exactly the same data as the quantum computer. But like you said, the uncertainty is an illusion, which means the Turing machine may be solving a different problem and/or starting with more information and so not have to solve the full factoring problem.

    As the most basic example, consider how fast a Turing machine could solve a factoring problem if it already started with part of the solution. Or consider that some numbers might be easier to factor than others and so the number that gets chosen to factor will always be one of these.

    There are indeed a number of ways to address the types of problems that Shor raised, but the point is we won't know which ones are more or less plausible until people actually start analyzing and discussing superdeterministic theories.