7 ms·
> if the "friend" is a human-level artificial intelligence running in a massive quantum computer. Well, there's your problem right there. No AI can run on a q
by lisper 6y ago
> if the "friend" is a human-level artificial intelligence running in a massive quantum computer.
Well, there's your problem right there. No AI can run on a quantum computer. Intelligence is inherently a classical phenomenon because it requires copying information, and quantum states cannot be copied, only classical states can be copied.
The classical world "emerges" from the quantum world when you take a quantum system and choose to consider only a subset of its degrees of freedom. When you do that, what results is a "mixed state" which behaves classically. This is not a reflection of any (meta)physical reality, it's just a consequence of your point of view. But here's the thing: you yourself are a subset of a quantum system, and so you yourself have no choice but to take this point of view. You cannot ever take the "god's-eye view" and "see" the whole system. That is fundamentally impossible because "seeing" requires copying information, quantum states can't be copied yada yada yada. Even God cannot "see" the god's eye view!
- Zenst 6y ago> quantum states cannot be copied Wouldn't quantum entanglement enable a comparable ability?
- gjhan 6y agoEntanglement means that there are systems with highly (more-than-classically) correlated state - that the combined state is something that can’t be represented by considering the systems independently - not that the states are the same. The no-cloning theorem prohibits a process that takes the state of a system and replicates it on another system. It’s certainly possible to have or to generate entangled sets of systems that have identical separable state, but not to copy an arbitrary state.
- boomboomsubban 6y agoI may be wrong, but I'm fairly sure you need to entangle the particles before they become a quantum state.
- amelius 6y ago> Intelligence is inherently a classical phenomenon because it requires copying information, and quantum states cannot be copied, only classical states can be copied. Ok, so make classical pointers to quantum states and copy the pointers instead of the quantum states :)
- adrianN 6y agoI'm not convinced that intelligence requires copying information in the way that the no-cloning theorem forbids. Approximate cloning is not forbidden for example.
- lisper 6y agoThe problem is not copying internal states, the problem is I/O. A given instance of a quantum computation can only be queried once before it decoheres and goes "poof" because the querying process itself decoheres the system. So a quantum computer can never pass the Turing test because it cannot even participate in the Turing test without destroying itself.
- Strilanc 6y agoRequiring the the quantum algorithm have literally no interaction with the outside world to count as quantum is a strawman. An AI could use Grover search when trying to find a good enough response. That would be like part of the AI going into and out of superposition for an efficiency boost. Many quantum tasks are more efficient if certain catalyst states are present [1][2]. That would be like a part of the AI staying permanently in superposition while still ultimately contributing to classical results, again for an efficiency boost. 1: https://en.wikipedia.org/wiki/Quantum_catalyst https://en.wikipedia.org/wiki/Quantum_catalyst 2: https://arxiv.org/abs/1812.01238 https://arxiv.org/abs/1812.01238
- lisper 6y agoLet's not lose the plot here: we're talking about a Wigner's-friend-type thought experiment. If you're not going to require that the quantum computer be isolated, then you may just as well use an actual human.
- Strilanc 6y agoAgreed. The measurement at the end of the experiment requires the agent's state to stay under superposition with no leaks (or perhaps only trivial leaks that can be patched over).
- c1ccccc1 6y agoA good point, but quantum computation can do anything that reversible classical computation can. If you do your entire computation while keeping all the qubits in the {|0>, |1>} basis, then performing a CNOT onto a qubit initialized to |0> will work as a classical copy. (It will fail to truly copy the quantum state, but that's fine.) You just need a large enough supply of qubits initialized to |0> for all the copies you plan to do.
- lisper 6y ago> quantum computation can do anything that reversible classical computation can Well, yeah, but that "reversible" hedge is quite significant. Reversible means you have to retain all of your intermediate states, and it is far from clear that you could build a reversible AI that would run for more than a few milliseconds before consuming all of the available resources in its light cone. > You just need a large enough supply of qubits initialized to |0> for all the copies you plan to do. Yes, and that produces exactly the same problem: it is far from clear that there are enough qbits in the universe to build a quantum AI this way. You also have a more fundamental problem, which is that neither quantum computers nor reversible classical computers can do I/O, so how are you going to demonstrate that this thing is in fact an AI? It can't even participate in a Turing test, let alone pass one.
- c1ccccc1 6y agoIf your goal is human level AI, there are definitely enough qubits. Humans can subsist for quite a while on just a sandwich and a large room full of air. And that is with very inefficient usage of the available negentropy. I guess one would have to demonstrate that the program implements an AI by running it on a computer with I/O, then move the same program over to an isolated quantum computer once it's passed that test. Maybe not completely philosophically satisfying, but should work in practice.
- nicoffeine 6y ago> Intelligence is inherently a classical phenomenon because it requires copying information, and quantum states cannot be copied, only classical states can be copied I don't agree with this. The recent research in quantum biology[1] suggests that biology at least can read from quantum states -- birds may use it to navigate, our sense of smell may be a quantum interaction, and plants appear to use quantum properties to achieve almost perfect efficiency for photosynthesis. The human eye may be able to detect a single photon.[2] General quantum activity also seems possible, since lithium-6 and lithium-7 appear to produce different pharmacological results despite being only one neutron apart[3]. I also fall into the camp of supporting the Orch-OR theory of consciousness[4], which claims that our interpretation of a classical system for brains is incorrect, and that neurons have internal structures called microtubules which can maintain quantum states. Microtubules have been known to exist for decades, but it was thought that their purpose was purely structural. Anirban Bandyopadhyay demonstrated[5] that these structures appear capable of storing information at the quantum level. Orch-OR would explain how anesthesia works, and how biology with classically simple brains are capable of extremely complex lifecycles and self-organization. I don't know enough about neuroscience to have any real insight on its validity, but there seems to be a huge disconnect between where we think we are with AI and the results so far. I have watched a butterfly navigate through a chain link fence on a windy day, and that kind of computation doesn't seem possible in a classical model, from a space or an energy perspective. > Even God cannot "see" the god's eye view! The strongest argument I see for a concept of God is the fact that the universe doesn't break. Quantum brains provide some interesting possibilities for the privacy of thought co-existing with some sort of perfect originator. In any case, given the amount of computation it would take just to model the biology of one person and the underlying quantum states, I am almost positive that we are not in a simulation. [1] https://royalsocietypublishing.org/doi/10.1098/rspa.2018.0674 https://royalsocietypublishing.org/doi/10.1098/rspa.2018.067... [2] https://www.nature.com/news/people-can-sense-single-photons-1.20282 https://www.nature.com/news/people-can-sense-single-photons-... [3] https://www.quantamagazine.org/a-new-spin-on-the-quantum-brain-20161102/ https://www.quantamagazine.org/a-new-spin-on-the-quantum-bra... [4] https://en.wikipedia.org/wiki/Orchestrated_objective_reduction https://en.wikipedia.org/wiki/Orchestrated_objective_reducti... [5] https://aip.scitation.org/doi/abs/10.1063/1.4793995 https://aip.scitation.org/doi/abs/10.1063/1.4793995
- robocat 6y ago> No AI can run on a quantum computer Each quantum calculation can be collapsed to a state, and that state could then be fed to the next quantum calculation etc. etc. Would you call a series of such quantum calculations a quantum computer, or not?
- lisper 6y agoFor the purposes of being a stand-in for Wigner's friend, no, obviously not. The whole point of the thought experiment is that the "friend" observer exists in a superposition.
- narag 6y agoSo in your opinion, does this thought experiment can be translated to a "real" experiment now or in a not-so-distant future, like in our lifetimes?
- lisper 6y agoThat depends on what you mean. If you mean using a simple system (like a single photon) as a stand-in for Wigner's friend, that has already been done. That's what the OP was about. If you mean using a complex entity like an actual human or an AI that can pass a Turing test, I'll give long odds against.
- narag 6y agoThe later... I ask because every time I read about some thought experiment, it turns out to be an impossible experiment. In a recent story I asked in general and others told me that there were thought experiments that became proper experiments. But it seems that they were somewhat different, more like pre-experiments. When something gets stuck in the thought experiment category, it's more like a no-experiment. I have no problem admitting that this kind of ellucubration is useful to anticipate problems. But extracting conclusions from an impossible setup? That's like asking what would happen if an irresistible force meets an immovable object.