Is Many-Worlds even falsifiable?

David Deutsch seems to think that a general intelligence running on a quantum computer ought to be able to observe itself existing in parallel worlds if many-worlds is true.

I'm skeptical that Deutsch actually meant that, but even smart people have crazy ideas.

He wrote a pretty well known article about it that as far as I can tell wasn’t written off-hand and hasn’t been retracted or disavowed. I’m pretty confident he meant it.

There has to be some misinterpretation going on here (or David Deutsch just hasn't thought this through). You can have a quantum computer in a superposition of states, but that isn't many worlds; to have many worlds, the computer has to interact with the outside world and split the outside world into a corresponding superposition of states. Such interaction with the outside world prevents a quantum computer from functioning - and it's precisely such information leakage to the outside world that is a major difficulty when trying to scale up quantum computers.

Which sadly means nothing, so many details being hidden behind the words "general", "intelligence", "running", "quantum computer", "ought", "oberve" and "existing".

If you could refine what these words all mean, I guess we could understand it as something more than "Deutsch seems to think that yes", which can replace your sentence entirely, "yes" describing it all as precisely.

I am myself, in some ways, a general intelligence running on a quantum computer but I dont feel like I can observe myself existing in a novel special way. Let alone being able to then express it for you in a way that is novel as well.

Most of those words don’t cause definitional problems here any more than they do for all other scientific tests which involve general intelligences (humans) making observations about things that exist. “Quantum computer” is a unique term that doesn’t show up in every description of a scientific experiment, but as far as I know it doesn’t have a particularly ambiguous definition.

And while you are indeed a general intelligence, I don’t think you’re running on a quantum computer.

> observe itself existing in parallel worlds if many-worlds is true

If many-worlds is true, we are already observing ourselves in each of those universes in which we can observe, we just don't communicate with our replicas. And if the AGI can communicate with its other instances, the exponential replicas will quickly overload their channel.

I think the point of the claim is that a general intelligence running on a quantum computer would make different observations if many worlds is true versus if many worlds is false.

Discovery of non-linearity in the evolution operator would falsify MWI. Like e.g. this germanium experiment gave a positive result.

Yes, because it falsifies every interpretation of quantum mechanics not just MWI itself. I think in context here, "is superdeterminism falsifiable" is as silly a question as "is logic falsifiable" or "is causality falsifiable". Specific mmathematical, causal or superdeterministic models are always falsifiable, but the idea that superdeterminism as a whole is not falsifiable should be no more surprising or interesting than the fact that logic is not falsifiable.

If superdeterminism allows specific models. The idea is that the experimenter's behavior is fine tuned so he can't discover superdeterminism, which means falsification would be a paradox by definition.

Given we can enumerate all possible models because we can enumerate all possible Turing machines, we clearly have enough degrees of "freedom" in this universe that that isn't an issue.

> Specific mmathematical, causal or superdeterministic models are always falsifiable

I request further elaboration here, please. Examples would be great.

> than the fact that logic is not falsifiable

What?

There is non-linearity in the evolution operator though - it's called gravity. It's why MWI in its trivial form is falsified - though with a linear theory of everything (potentially string theory) - it might be possible.

Hopefully this is on point https://www.lesswrong.com/posts/DFxoaWGEh9ndwtZhk/decoherenc...

Yes. You could, for example, allow a measured particle to interact with an isolated measuring apparatus of varying size - from a few atoms to a macroscopic object - and see how the size affects the system - so you get evidence of decoherence?

Also, many worlds interpretation as it currently exists is already falsified by its incompatibility with gravitational nonlinearity.

't Hooft's version of superdeterminism, according to his book, can be falsified by demonstrating a large scale quantum computer. I don't know if Hossenfelder's version is something different or how she would account for this.