Can consciousness be explained by quantum physics?
theconversation.com
theconversation.com
However, I believe that consciousness itself is an emergent phenomenon. I believe you could set up a huge neural net (as much as all computation hardware extant in 2021), give it sensors and servos much like the human body, and AGI would be the result after decades of learning.
First it would have to figure out how to hold up it's head, and work with those low resolution cameras that flitter about under servo control. And do all the other things babies do.
This is actually one of the most important things "hard-wired" into us that I am talking about! Language acquisition seems to be natural for humans, even if the languages are quite different.
Can we devise an AGI that feels intense pleasure when doing any kind of work?
For example, David Chalmers, who originated or developed many favorite anti-materialist positions (the 'hard problem', the explanatory gap, p-zombies) holds many positions that would normally be considered materialistic - for example, that if one were to replace a person's neurons with functionally-equivalent artificial devices, the resulting entity would still be conscious in the same way as the person we started with, and therefore, presumably, capable of feeling pain and pleasure. It would seem tendentious to hold this position and still assert that artificial conscious entities (complete with subjective experiences) are impossible.
Why isn't the internet conscious?
There is nothing magical about this: consciousness is obviously a complex computational process that has evolved over millions or billions of years. It is a good solution to the problem of navigating the world as a living thing.
Saying 'you need an extremely complex computational system to have consciousness' does not entail that any extremely complex computational system is conscious (incidentally, this is also the mistake that the integrated information people make).
And so about your physical body, a collection of atoms can 'try' to do something?
'Trying' is a function of conscious intent. So you can't 'try' to be conscious, that would be a prerequisite.
"consciousness is obviously ..."
Glib statement of the century.
"It is a good solution to the problem of navigating the world as a living thing."
Only conscious, living things can have 'problems', ergo, your statement is again circuitously referential.
If as you say 'there is nothing magical' then there can never be 'intent' or 'problems' to solve - it's a pretty big giant contradiction.
There would just be 'bags of particles bouncing around, randomly' that's it. In the context of materialism, then there isn't even anything such as 'evolution' - just the appearance of it. Atoms bounce around from one thing to the next, randomly, we just call it a 'process' - but it's not, so long sticking to materialism (i.e. no magic).
" 'you need an extremely complex computational system to have consciousness' does not entail that any extremely complex computational system is conscious"
Sure. But the totality of technology on the planet is already very well connected and 'sufficient' (according to the OP's indicated scale) so as to form the basis of some kind of 'network', which, while may not be 'conscious', should nevertheless exhibit some of the emergent properties of such a thing, if emergence does come out of those kinds of configurations.
Things in the world behave the way they behave regardless of our interpretations of them. Electrons were bound to nuclei long before consciousness observed and named these phenomena. Bacteria and archaea looked for sustenance and reproduced and suffered changes and adapted the same way billions of years before any consciousness existed in the world. They were 'trying' to survive just as much as we are - you can express this in principle in terms of a physical law just as much as the motion of the Electron wave, but we don't know that law and use a human word, "trying", instead.
As evolution happened according to the laws of chemistry and statistics, biological beings became more and more complex, and found more and more complex ways of interacting with the world. At some point(s), simple agents capable of planning were born from a previous generation, and these created more offspring than previous organisms, so this trait became more and more pronounced. After eons of this process, these planning abilities started including information not just about the environment, but about other living things, agents (the behavior of such things is not well explained by basic mechanical laws). Over more time, this same logic circuit probably started being applied to model the behavior of the entity doing the planning, which is probably what we recognize as self-reflection, experiences, consciousness.
Of course, there are many things in my account that may turn out to be inaccurate. I'm not claiming to understand the exact evolution of consciousness - I'm only giving a somewhat vaguely plausible explanation of consciousness as a physical, evolved phenomenon. We don't know how to model evolution, we have no idea how life began from anorganic compounds, we don't know to what extent evolution could have produced different results than we see today.
Also, while concepts like 'life' and 'agents' and 'objects' are human categories imposed on a fuzzy world, that does not mean that these concepts are arbitrary or that they are purely fabrications of the human mind. It is an objective fact of the world that the laws that describe the motion of a rock tumbling down a slope are different from the laws that describe the motion of a gazelle running down a slope. Any being or computer that wanted to predict the speed or position of the gazelle and the rock would fail to do so if it didn't account for this difference.
Another important note that many reductionists seem to not accept is that a computer computes regardless of a conscious mind interpreting this computation. If we find a physical process that moves a block of wood in response to certain patterns of light hitting a seminconducting crystal, that system is a computer using a camera to open a door, regardless of whether it happened naturally or was designed by a designer. This was the great insight that Turing and Church came up with: that computation is a mathematical object that is isomorphic to any system with certain simple properties.
This observation is so fundamental that we can actually deduce limits of physical processes by modeling them as computations: for example, we know that certain crystals need some minimum time to grow a certain way because the physical laws that cause the crystals to grow are isomorphic to a known optimization problem. Since we know the minimum complexity of a computer solving that problem (the minimum number of steps that need to happen) we also know that the physical processes must take at least [time(step) * number of steps] time to grow the crystals.
For the internet, the explanation is simple again: there is no equivalence between the evolutionary laws that led to consciousness evolving and any process used in the design of the internet, so we have no reason to expect the internet to be conscious. Again, some level of complexity is necessary but not sufficient for consciousness to emerge, in the materialist theories of consciousness. There are very few people (notably, the IIT people) that equate complexity with consciousness.
If you have bred animals for ability, the folk wisdom is the sire throws certain traits, the dam attenuates them, and then develops others through socialization. This qualitative unscientific view obviously overlooks the dam genome, but through the lens of DNA molecules encoding models to run on a consciousness via quantum operations, the metaphor I would use is DNA encodes compressed model data.
In this view, the consciousness evolves over its runtime until it ceases to cohere information, but anything past the basic "lift head, solve for food, security, reproduction" models that run on a consciousness may originate at conception via genes.
Why do you say that?
The brain and it’s cells certainly have a lot of structure that’s baked in from inherited genetic and epigenetic information. I don’t see what that has to do with whether conciseness is dependent on quantum effects.
e.g. Consciousness is not some higher dimensional super or substrate, it's code, compressed into and transmitted by dna, running on a quantum computer in a brain, learning and processing according to initial weights in the dna model.
We evolved the way we did due to evolutionary constraints.
The constraints of the evolution of our silicon computing devices are completely different.
Why would you hook all that gear up to a single automaton, with a head, two eyes, two arms and two arms. Those are completely arbitrary.
Why not hook it up to 100's or millions of cameras? Thousands of 'limbs'? i.e. ever camera and operable apparatus connected to the internet?
Our 'silicon' will have a recall advantage that our pure-NN brains do not, which is consistent lookup of data and 'perfect recall' of every bit of human data every recorded, which means that it's evolution will be quite different as well.
We'll have distributed systems that go way past the 'Turing Test' decades before anything remotely approaching AGI.
'Siri' is almost there, in 10 years I'll bet it becomes difficult to tell if Siri is a person or not. And Siri isn't even 'one entity' - it's a generic interface to other, distributed systems. 'Siri' in 10 years will start to resemble AGI but with other, more fantastic abilities (information access and synthesis).
Imagine in 50 years your Tesla is deeply integrated with satellite GPS location data, weather data, it's 'talking' to the 'local road controller' for control data with ms latency, and also 'talking' to every other car within 500m to get relevant positioning and other sensor data. Maybe some of the relevant AI processing requires far too much power, and it's done in real-time at the closest data centre via networking. At that point you have to say the Tesla isn't really the vehicle in front of you, it's an entire system of technology.
Sun Microsystem's motto was 'The Network Is The Computer' implying to the extent that all computers are comprehensively connected ... they form at least some kind of coherent entity.
The concept of AGI is so abstract that it almost doesn't make any sense.
By the time we are able to assemble a discrete group of computing components, to a discrete bipedal robot (i.e. a Metropolis-like automaton, i.e. Data from Star Trek) we will already be many decades ahead of that in terms of understanding how all of that maps to consciousness. We'll already have intelligent systems doing things way beyond what any human could do and probably what we could imagine right now.
The research described in this article is similar. Even interpreting it charitably, it has nothing to do with the human brain. The research shows that it may be possible to build quantum devices with certain properties that could have high communication complexity. It doesn't show that human brains have any of these properties or that these properties are in any way necessary to explain consciousness.
String theory, "not even wrong" as it may be, was at least inspired by an empirically adequate way to explain certain experimental data.
But as we all know: magic works, because quantum something or whatever ...
Consciousness has, as yet, no empirical explanation. Only a variety of theories that, so far, have failed to make any headway in providing an explanation. That's why it's called the "hard problem of consciousness."
If you think there is an easy answer, then you do not yet understand the question.
Just like you do not need quantum physics to explain how muscles work.
But that is the opposite of the case here. We came up with those ideas after we had an understanding of the system. With consciousness, people casually use the term all the time and think they have a vague idea of what it is, but we have little understanding of the system. It may be that it simply does not exist for example.
This is so even if it turns out that the phenomenon does not exist, such as in the case of the mythical causal effects of the four humors.
Because nobody has a definition and when it turns out that computers can fit the definition people will move the goalposts.
The hard problem of consciousness is getting a priest to agree to a definition that a scientist can test. Because everything that touches the real world can probably be modelled with distressing accuracy by matrix multiplications and a little non-linearity.
That's not the only reason though.
> The hard problem of consciousness is getting a priest to agree to a definition that a scientist can test.
Scientists are free (assuming free will) to define their own, which I think they (some) actually have.
There is no easy answer that people find satisfactory, but it very well might be that people's intuitive understanding of their own consciousness is incoherent and there are in fact no answers at all that don't violate one or more of the arbitrary axioms that populate the thoughtspace regarding that concept.
However, if we don't start with such assumptions, it is easy to accept that our attention mechanisms have some way of interpreting the stimuli that they receive, and that similar computational architectures will interpret these stimuli the same way. It is entirely possible that all humans experience pain almost identically, aand that it is even more or less identical to how dogs experience pain; while at the same time, an alien or AGI might experience these things entirely differently.
It is also quite obvious that a being without self-reflection, such as a microprocessor, will not have this experience mechanism, almost by definition.
With enough study of the computational structure of the brain, we will likely come to even understand the precise advantage that experiences (self-reflection) have.
To give just one personal reason why I don't believe in the idea/importance of qualia, I feel very clearly that there are certain mental states that I experience (such as colors and smells etc), but there are also mental states that don't have an associated experience. for example, I don't think there is such a thing as 'the feeling of thinking about the number 277`. This to me suggests strongly that experiences are a specialized part of our computational apparatus, with specific roles that we don't understand precisely now (like so much about our minds), but with limited applicability - not some be-all, end-all of consciousness and thought.
I will also note that my position is entirely in line with philosophers like Daniel Dennett, it is not some dismissal of mainstream philosophical thought.
What does it mean "to know" as opposed to contain, or have access to information? A computer program, in one sense, "knows" what the current value of a variable is. But it has no awareness of its own knowing.
There is no doubt that we are creatures of stimuli that is computationally processed in the nervous system. However, there is an ontological difference between the the ability to process stimuli, and the experience of said stimuli. The presumption is that with a sufficiently complex system, the ability to experience stimuli will self-reflective awareness, i.e., consciousness will emerge. But that's a whole lot of hand-waving. What is the mechanism? None is offered. It is only assumed.
As others have noted, we cannot even agree on what the term "consciousness" means, and because of this, some would like to posit that consciousness does not actually exist, i.e., that it is illusory. The irony of this conjecture is that they are trying to convince self-aware beings that they are not self-aware. The intent to convince the other is itself a function of self-awareness. Intent cannot exist without self-awareness, and Persuasion of "the other" also cannot exist without self-awareness.
Denying a phenomenon does not make it go away. I could just as easily respond to the person making this argument that if, indeed, self-awareness is an illusion, then so is their argument. It is mere data, the product of a series of calculations leading to a linguistic expression that intends to communicate nothing. It is merely the output from an extremely complex set of interlocking algorithms.
Despite the difficulty of agreeing to a definition of consciousness, we each, to a person, know what it is because of our own experience. The reason all current attempts to explain consciousness are unsatisfactory is because these attempts contradict our common experience. In fact, every attempt so far denies experience itself. They are unsatisfactory because they are illogical, or because they do not provide any explanation at all, but merely presume that somehow the conclusion is achieved, despite the lack of any mechanism in the proposition that can bring us to that conclusion.
If we can even go so far as to agree: "Consciousness is a property of a being that is aware of its own existence, and is able to contemplate its own existence as an entity distinct from its environment," this does not make it any easier to explain how any computational model will arrive at such a property.
One does not have to be a theist to acknowledge this. Some of the best minds in consciousness research are atheists.
Unfortunately this definition doesn't help weeding out misconceptions such as philosophical zombies. All you have to do is to say that a system may claim it's conscious without really being "aware" of it's own existence. This definition doesn't help because it doesn't say what being aware of it's own existence mean, offering arguably a circular definition.
Quite contrary to solipsism, I'd posit that what matters is behaviour. If a cognitive process is able to communicate with us and convincingly recount its reflections about itself, for all intents and purposes that system is conscious. It doesn't matter if it differs from our consciousness in significant ways.
Let's consider a machine that is able to reflect on itself quote effectively but it's also to turn on and off that module "at will". This is quite different from our experience of consciousness, in that we're not able to turn off our own consciousness, to the extent that we believe we're "there" only while we're conscious of it, and thus we end up equating our existence with those surface experiences. But we're way more than that. Were made of a thousand brains and not all processing reaches what we call our consciousness.
A system that exhibits a different mixture of those mechanism shouldn't be ruled out as not meeting the arbitrarily anthropomorphic bar of consciousness.
One key aspect that people that talk about qualia and p-zombies and Chinese rooms assume is that it is possible, in principle, to behave as if you are conscious without really being conscious. The fact that a p-zombie is conceivable does not entail that it can exist in the world.
It is entirely possible, I believe even likely, that as we understand the evolution of consciousness we will discover it is in fact a necessary property of an agent with certain abilities. The reason I believe this is a simplistic evolutionary argument: consciousness would be unlikely to have evolved if it had not been beneficial or even necessary for the beings which possess it.
The tendency to say that people are in fact not conscious beings and that the consciousness itself is an illusion seems to me to come mainly from people who reject free will, and point out that our experience of cosnciously choosing our next action is illusory - our unconscious mind is certainly 'in charge' of many processes, and our consciousness often is only an observer of these processes. For a basic example, I can notice how my heart is beating, and I can sometimes feel like my heart is beating faster because I am scared, but this is jusg an interpretation, which mag be wrong - as anyone who has suffered a panic attack can tell you. More interestingly, when observing people with altered states of consciousness, such as Alzheimer's disease, it is often possible to observe them emitting wildly wrong theories about their own actions, such as claiming that they are dressed up because a relative came to visit, when in fact they were dressed up because they were preparing for an appointment. These examples show at the very least that our consciosuness is to some extent a mechanism that comes up with theories about our own actions, without being directly the cause of those actions, which contradicts our experience of consciousness.
The fact that we can't yet explain these processes and how they come about from computation is not surprising to me, given the youth of the field of computer science and the complexity of the human brain and mind.
(This is a reference to Fermat's Last Theorem in which Fermat confidently proclaimed in a margin of his book that he had a simple elegant proof which the margin was too small to contain)
IF our choices and actions are free, quantum particles' behavior is free in the exact same sense, assuming three postulates, FIN, SPIN, and TWIN. Neither random nor deterministic, but free.
I take its lack of being discussed by the usual heavy-hitters in quantum physics as a sign it is not well received, but there is still a lack of discussion for being a theorem developed by Conway and Kochen, both very highly regarded.
"Free choice" is so visceral, and it is hard to imagine how experimental outcomes would be determined in heavily contrived ways such as the 9182nd word in a some book chosen by some other contrived way, and so on and so on to choose which spin axis to measure. And all done after the particles are spacelike separated
What does that mean, though? It seems to me that a mixture of randomness and determinism covers all logically possible systems. I don't understand what "free" is supposed to represent at all. The Free Will Theorem is about certain events not being a function of the past, but does it exclude randomness? How does it do that?
My thought about this is that the feeling of free choice that we have stems from a form of irreducible epistemic uncertainty: we can't know what we are going to do before we do it, because our knowledge would influence our action. Furthermore, when choosing between A and B, we may use our self-concept to simulate our choices, meaning that by necessity our self-model is non-deterministic: it indeed performs several different choices during deliberation.
I think our error is that we mistake epistemological uncertainty for metaphysical uncertainty, and the non-determinism of our self-model for the non-determinism of our real self.
But is it the simpler explanation? That is hard to answer. On the one hand Many Worlds is "simple" and deterministic, on the other hand the complexity of the universe is so much greater. A stochastic model like GRW is also simple in the sense there is only one world, and all particles have a regular chance of random collapse. Also Many Words doesn't just provide the (classical picture)*(number of worlds), so it isn't even that "simple". There is nonlocal structure to the global wavefunction. The worlds don't split into any possible worlds; if one splits one way the others have to split into the other possible choices. These global correlations are nonlocal AFAIK - I know the MWI is nonlocal but I have a hard time understanding David Wallace, Lev Vaidman, and Tim Maudlin who all say it is nonlocal. The precise nature of the nonlocality is hard to understand (for me), but it is nonlocal!
> The worlds don't split into any possible worlds; if one splits one way the others have to split into the other possible choices.
My (possibly flawed) understanding of MWI is that there isn't really any splitting. There is just a quantum superposition of all possible worlds, and the amplitude corresponding to each one fluctuates over time. A "split" from A to possibilities B and C would just correspond to the amplitude of A draining to 0, half going to B, the other half going to C, but the branches might have had nonzero amplitudes to begin with (they already existed, but now they are "stronger"). Different "branches" may also pour amplitude into the same other "branch" (i.e. they join rather than split).
In other words, every moment of possible universe has always existed and will always exist, but over time they switch from "off" (amplitude 0) to various degrees of "on", or to off again, and so on, accordingly to the wavefunction. It's nonlocal in the same way that quantum entanglement is nonlocal in any interpretation, it's just that now we're working on the gargantuan quantum entanglement of every possible state of the entire universe.
MWI is arguably the simplest interpretation from the standpoint of rule simplicity: it doesn't involve a notion of collapse, so it has less rules than other interpretations. It is less simple if you look at the sheer mass of what exists according to it, but I find that notion of simplicity less compelling than rule simplicity. Either way, I don't think it matters so I'm noncommittal.
This also seems to imply the MWI forces us into block time, then having to recover the experience of time flowing for finite observers like us. The MWI-verse has always existed, as you say, so I'm having a hard time imagining what you mean by dynamical "switching", unless you mean the feeling/appearance of. It would have to be the appearance of something happening, an appearance of time flowing when it is block. I'm sure we can imagine MWI can explain how observers like us in the MW universe experience a flow of time. This still may be simple in the long run.
The idea that something does not "exist or not exist", but rather "exists at amplitude 0, amplitude 0.5, amplitude -i, etc." is kind of a mind screw, but that's the suggestion.
This being said, there might be a steady state or fixed point where all amplitudes remain constant. Not sure how we could know if there is one or if we're in it, but that would be interesting.
This isn't entriely true. MWI still needs to explain why the observed behavior of a measurement is non-linear. So, it comes with a rule about how to deduce the observed probabilities from the wave function amplitudes: you have to do the calculation in a single 'world', you can't do them for the entire wave function; and then the amplitudes give you some kind of count for each 'world', so that if we then perform an analysis on all observers in all worlds, the chance of us being one of the observers who measure state A is proportional to the amplitude.
Even worse, the MWI needs to find or posit some way for the universe to choose, for each quantum measurement, a particular basis that gives rise to the many worlds - as that is also something missing from the Schrodinger equation. The CI has an advantage here, as the basis is explicitly chosen by the measurement device, whereas in MWI it has to be a fundamental property of the universe (this is usually suggested to happen through de coherence, that there is actually one unique 'right' basis, the one that is stable under de coherence effects, but this is not entirely satisfactory).
Every few years a story about how quantum physics is releated to our brain..
I'm not a physicist or a consciousness researcher, but I'm not sure what the discovery is here. Unless they used fractal patterns to produce quantum effects at macroscales? But the article doesn't say what the scale is, and 'printed on a chip' + 'injected a photon' sounds pretty small.
this is analogous to saying that 'all hexagons turn to ice when frozen' and then freezing a hexagon made of water. we already knew that water freezes to ice, we haven't discovered a novel property of hexagons
There is value in demonstrating that our suppositions hold, but, by their very nature, such confirmations are unlikely to herald breakthroughs. They do not seem to have demonstrated anything specific to the Penrose - Hamerof hypothesis.
Drive-by downvotes aren't an argument, by the way.