Scott Aaronson takes down Roger Penrose’s nonsense about human brain having an uncomputable superpower beyond known physics.
If you want to give your AI model quantum noise, because you believe that a source of unpredictable random noise is key to a ‘true’ intelligence, well, ok. You could absolutely make a computer chip with some analog circuits dependent on subtle temperature fluctuations that add quantum noise to a tensor. Does that make the computer magic like human brains now?
Or does it have to be the specific magic of ions and ion channels? Well, what if we made a machine with tiny containers with water and ions and synthetic ion channels, and conductivity sensors. We plug that ion movement noise into the neural network as a source of noise. Does that do it?
Why base your human chauvinism on quantum noise? Because quantum stuff seems mysterious and cool and seems to do some funky magic stuff at very small scales (while the large scale combinations of it are just the normal workings of the universe which is all we’ve ever known). Ok, meat-worshipper, but then where will you run if I add ion channels to my AI?
Hurumph. Nice to hear Scott Aaronson tackling this irrational meat-chauvinism.
Scott Aaronson takes down Roger Penrose’s nonsense
That clip doesn’t address Penrose’s ideas at all (and it’s not meant to, Penrose is only mentioned at the end). Penrose’s theory is that there is a subquantum determinism with noncomputable equations of motion, the noncomputability being there to explain why humans can jump out of axiom systems. That last part I think is a confusion of levels, but in any case, Penrose is quite willing to say that a quantum computer accessing that natural noncomputable dynamics could have the same capabilities as the human brain.
Ok, fair enough, I overstated the case Scott Aaronson was making. It is my position however that Roger Penrose is wrong about the importance of quantum effects in neurons, so I was excited to find support in what Scott was saying. Here’s a specific passage from Roger Penrose and Stuart Hameroff’s paper “Consciousness in the universe; A review of the ‘Orch OR’ theory”:
“MT [microtubule] automata based on tubulin dipoles in hexagonal lattices show high capacity integration and learning [61]. Assuming 10^9 binary tubulins per neuron switching at 10 megahertz (10^7 ) gives a potential MT-based capacity of 10^16 operations per second per neuron. Conventional neuronal-level approaches based on axonal firings and synaptic transmissions (10^11 neurons/brain, 10^3 synapses/neuron, 10^2 transmissions/s/synapse) give the same 10^16 operations per second for the entire brain!”
Roger and Stuart make this claim in earlier papers as well. They also make claims about the microtubules being used in various ways for information storage. I think the information storage claims they make are likely exaggerated, but I’m not confident that there isn’t non-negligible information storage occurring via microtubule modification.
I will, however, state that I believe the microtubule computation hypothesis to be false. I think 10^15 − 10^16 operations per second is the correct estimate for the human brain.
If you want to try to figure out a bet to make on this, we could make a Manifold Market.
I think (on “philosophical” grounds) that quantum entanglement probably has a role in the brain, but if the microtubules are involved, I think it’s far more likely that each microtubule only contains one or a few logical qubits (stored as topological quantum information, entanglement that resists decoherence because it is wound around the cylinder, as in the Kitaev code).
Penrose is quite willing to say that a quantum computer accessing that natural noncomputable dynamics could have the same capabilities as the human brain.
Hmm, we’re still talking past each other. I’m saying I don’t believe any quantum weirdness is behaviorally relevant to the human brain or simulations thereof.
Just ordinary analog and digital computer chips, like in ordinary consumer electronics. Nothing special but the neural architecture and learning rules set up by the genome, and the mundane experience of life.
Right, and I disagree with the usual computational theory of mind (at least with respect to “consciousness” and “the self”), according to which the mind is a kind of virtual state machine whose microphysical details are irrelevant. There are sorites problems and binding problems which arise if you want to get consciousness and the self from physically coarse-grained states, which is why I look for explanations based in exact microphysical properties and irreducible complex entities instead.
https://youtube.com/clip/UgkxowOyN1HpPwxXQr9L7ZKSFwL-d0qDjPLL?si=JT3CfNKAj6MlDrbf
Scott Aaronson takes down Roger Penrose’s nonsense about human brain having an uncomputable superpower beyond known physics.
If you want to give your AI model quantum noise, because you believe that a source of unpredictable random noise is key to a ‘true’ intelligence, well, ok. You could absolutely make a computer chip with some analog circuits dependent on subtle temperature fluctuations that add quantum noise to a tensor. Does that make the computer magic like human brains now?
Or does it have to be the specific magic of ions and ion channels? Well, what if we made a machine with tiny containers with water and ions and synthetic ion channels, and conductivity sensors. We plug that ion movement noise into the neural network as a source of noise. Does that do it?
Why base your human chauvinism on quantum noise? Because quantum stuff seems mysterious and cool and seems to do some funky magic stuff at very small scales (while the large scale combinations of it are just the normal workings of the universe which is all we’ve ever known). Ok, meat-worshipper, but then where will you run if I add ion channels to my AI?
Hurumph. Nice to hear Scott Aaronson tackling this irrational meat-chauvinism.
That clip doesn’t address Penrose’s ideas at all (and it’s not meant to, Penrose is only mentioned at the end). Penrose’s theory is that there is a subquantum determinism with noncomputable equations of motion, the noncomputability being there to explain why humans can jump out of axiom systems. That last part I think is a confusion of levels, but in any case, Penrose is quite willing to say that a quantum computer accessing that natural noncomputable dynamics could have the same capabilities as the human brain.
Ok, fair enough, I overstated the case Scott Aaronson was making. It is my position however that Roger Penrose is wrong about the importance of quantum effects in neurons, so I was excited to find support in what Scott was saying. Here’s a specific passage from Roger Penrose and Stuart Hameroff’s paper “Consciousness in the universe; A review of the ‘Orch OR’ theory”:
Roger and Stuart make this claim in earlier papers as well. They also make claims about the microtubules being used in various ways for information storage. I think the information storage claims they make are likely exaggerated, but I’m not confident that there isn’t non-negligible information storage occurring via microtubule modification.
I will, however, state that I believe the microtubule computation hypothesis to be false. I think 10^15 − 10^16 operations per second is the correct estimate for the human brain.
If you want to try to figure out a bet to make on this, we could make a Manifold Market.
I think (on “philosophical” grounds) that quantum entanglement probably has a role in the brain, but if the microtubules are involved, I think it’s far more likely that each microtubule only contains one or a few logical qubits (stored as topological quantum information, entanglement that resists decoherence because it is wound around the cylinder, as in the Kitaev code).
Hmm, we’re still talking past each other. I’m saying I don’t believe any quantum weirdness is behaviorally relevant to the human brain or simulations thereof. Just ordinary analog and digital computer chips, like in ordinary consumer electronics. Nothing special but the neural architecture and learning rules set up by the genome, and the mundane experience of life.
Right, and I disagree with the usual computational theory of mind (at least with respect to “consciousness” and “the self”), according to which the mind is a kind of virtual state machine whose microphysical details are irrelevant. There are sorites problems and binding problems which arise if you want to get consciousness and the self from physically coarse-grained states, which is why I look for explanations based in exact microphysical properties and irreducible complex entities instead.