Many-worlds implies computers are not conscious?

[jcap]

Consider the following thought-experiment in the many-worlds interpretation.

Suppose that I have a reversible conscious observer AI and a particle with +1/2 spin in the z direction.

Next the observer measures the spin in the x-direction and therefore spits into a version that measures +1/2 and another that measures -1/2.

Now imagine reversing both of the observers, together with the split particle states, so that we end up back with a single observer in the initial state and the particle with +1/2 spin in the z direction.

I think this would be the standard many-worlds interpretation.

But is there a contradiction here?

If both versions of the observer consciously perceive a whole particle with a definite x-spin then when everything is reversed surely we should end up with two particles with x-spin +/- 1/2 rather than one particle with z-spin +1/2?

I think that if many-worlds is true then no reversible machine (i.e. quantum or classical computer) can act as a conscious observer.

In contrast the operation of the brain seems to be fundamentally irreversible; when we consciously measure the spin of a particle our brain and the particle split never to be "put back together again" even in principle. Perhaps conscious awareness of the measurement results are a kind of "guarantee" that the evolution of the branches of the wavefunction will never be reversed in the future.

 

[Nugatory]

Now imagine reversing both of the observers, together with the split particle states, so that we end up back with a single observer in the initial state and the particle with +1/2 spin in the z direction.

This reversibility is only possible if the interaction that entangles the particle with the observer doesn't decohere the system enough to be thermodynamically irreversible (and in that case we wouldn't call the interaction an "observation"). It's irrelevant whether the interaction is with a human brain or a computer; both are complex enough systems that changing their macroscopic state will be irreversible.

MWI doesn't say anything different than any other interpretation here, it just manipulates the wave function differently to come up with the same result.

 

[PeterDonis]

Now imagine reversing both of the observers

If you can do this, then no measurement has actually occurred, and therefore according to the MWI no branching of the wave function has actually occurred. So neither observer actually observed anything.

 

[jcap]

If you can do this, then no measurement has actually occurred, and therefore according to the MWI no branching of the wave function has actually occurred. So neither observer actually observed anything.

Therefore it seems that you agree with me that in the many-worlds interpretation a reversible computer cannot function as an observer.

Am i right?

 

[PeterDonis]

Therefore it seems that you agree with me that in the many-worlds interpretation a reversible computer cannot function as an observer.

This has nothing to do with the MWI or any other particular interpretation; it's basic QM. If an operation is reversible, it isn't a measurement, and in basic QM, anything that can "function as an observer" must be making a measurement.

 

[jcap]

David Deutsch proposed this thought-experiment in order to test whether Copenhagen or many-worlds is true.

An artificial intelligence makes a measurement of a quantum system and then reverses its measurement in order to restore the system to its initial quantum state.

He thought that if the quantum state has collapsed then the AI has made a Copenhagen-type observation whereas if the quantum state is intact then the AI has split and made many-worlds-type observations.

I think we agree that if the final quantum state is intact then in fact no observation has taken place at all.

As all computations can be expressed as reversible computations then it seems to me that quantum measurement must be non-computable.

 

[PeterDonis]

David Deutsch proposed this thought-experiment in order to test whether Copenhagen or many-worlds is true.

Then you should give a reference to whatever paper Deutsch published with this proposal.

An artificial intelligence makes a measurement of a quantum system and then reverses its measurement in order to restore the system to its initial quantum state.

This is obviously inconsistent with the basic QM concept of measurement: the basic QM concept of measurement is that measurement is irreversible, so you can't "reverse a measurement".

 

[jcap]

Then you should give a reference to whatever paper Deutsch published with this proposal.

Quantum Theory as a Universal Physical Theory
International Journal of Theoretical Physics vol.24, no.1 1985.

https://link.springer.com/article/10.1007/BF00670071

The observer AI thought experiment is described on pages 32-36.

Does providing a link to the article via Google Scholar break copyright?

 

[PeterDonis]

Does providing a link to the article via Google Scholar break copyright?

It depends on what the link is to. I don't think Google Scholar checks copyright for the links it shows.

 

[Nugatory]

a reversible computer cannot function as an observer.

For all X, a reversible X cannot function as an observer.

It doesn't matter which interpretation we choose, and it doesn't matter whether X is a computer, a human brain, or even just the air through which a particle moves (consider a cloud chamber and the Mott paradox) - if the interaction is reversible it's not an observation.

 

[sysprog]

As all computations can be expressed as reversible computations then it seems to me that quantum measurement must be non-computable.

How is e.g. the computation a=a-a expressible as a reversible computation?

 

[jcap]

It depends on what the link is to. I don't think Google Scholar checks copyright for the links it shows.

The link Google scholar gives is certainly not the paywalled journal link I give above.

 

[PeterDonis]

The link Google scholar gives is certainly not the paywalled journal link I give above.

That doesn't necessarily mean it's ok. Generally, if the journal link is paywalled, I look on arxiv.org to see if there is a preprint; that's the safest place to look for a copy that can be shared.

 

[jcap]

How is e.g. the computation a=a-a expressible as a reversible computation?

I don’t know. However as far as I understand it all classical computations can be performed by a quantum computer. If that is the case then the quantum computer can be reversed as it must obey unitarity by definition.

 

[PeterDonis]

However as far as I understand it all classical computations can be performed by a quantum computer. If that is the case then the quantum computer can be reversed as it must obey unitarity by definition.

I don't follow your logic here. First, whether or not a quantum computer can be reversed has nothing to do with whether or not a classical computer can be reversed. Second, you seem to be assuming that all classical computations can be reversed, which is false. And third, a quantum computer can only be reversed if it hasn't been asked for a result yet; once the quantum computer gives a result, that's a measurement and is irreversible. Before the quantum computer gives a result, there is no measurement yet and so everything is reversible; but, as already noted, that is true in any QM interpretation, since it's just part of basic QM.