Toward machine consciousness

I am a bit sceptical of this, based on my reading of the paper itself and my admittedly limited experience of Professor McFadden's theories, on which it is based. Now, I should be clear that I know I could be wrong. I may need to change my mind in future. I should also admit that everyone has their own ideas on how AI should work, creating an obvious tendency to be critical of other people's. Nevertheless, right now, I have some issues.

My main problems are not with the experimental work in the paper itself, but with Professor McFadden's theories on which it is based.

Professor McFadden proposes that both human consciousness and evolution are based in a very important way on quantum mechanics - specifically on Everett's many-worlds interpretation of quantum mechanics. He claims the existence of "free will" and proposes that computation in brains is not sufficient to explain our actions - that we are are "not automatons" - and he proposes "adaptive mutation" - that single celled organisms can kind of "sniff out" the possibilities of different mutations in different worlds and somehow select the best mutation - so that there is a statistical bias towards favourable mutations.

Before I give some criticisms on this, my criticisms are not based on his usage of the many-worlds interpretation. I do not have any problem with that. Now, onto the criticism:

I think that "free will" is an incoherent idea, if it is supposed to be anything more than another term for complex systems making decisions. The usual AI view of human decision making is that some computational system is in some computational state. The system then enters its next computational state, possibly affected by inputs which it received, and makes outputs. It then enters its next computational state and so on. Every computational state follows from the last. Right now I am typing this and not something else because right now my brain is in a state that makes me type it. I may feel that I choose to type it, but that in itself does not mean that there is any real choice, independent of the computation. The choice itself was made that way because of the computational state of the brain. "I" did not choose this computational state. It followed from previous computational states.

There is no escape from this. I can try to escape. In fact, I will try to escape now. I refuse to accept my path down the rails of typing an article to criticise Professor McFadden. Instead, I will step outside my "programming" and write part of a James Bond novel.

James Bond removed his shoelaces. Q had made them and they contained a sharp wire made of an advanced carbon material with many times the strength of steel. Bond used it to decapitate the guards. Then he went and had a martini.

Now, was that convincing? For a moment there, did I really step off the rails that my brain had laid out for me? Of course I did no such thing. All we just showed is that the rails I stepped off - those of writing my criticism - were not the true rails. The real rails, all along, were those of running down the criticism of Professor McFadden rails, stepping off those rails to write James Bond fiction, and then getting back to my criticism. I stayed on the real rails all along and all that you, and I, just found is that those real rails allowed some strange behaviour for me.

I think there is no escape from this. Anything that I do must always be done because it was in the nature of the physical system that "owns" me to do it. It is this physicality that makes any hope of free will misplaced. Ultimately, you can try to escape in lots of little ways, but there is no escape from what you are. For this to happen there would have to be an "I" which is somehow separate to what you are and which is making the decisions in some way, yet how could this be? There is no "I" that exists separately to you.

Professor McFadden says that free will is accounted for by quantum mechanics and that we are "not automatons". What I just said above is why I have problems with this. Even if we somehow give the brain the capability to "sniff out" things in a multiverse, how does this change what I just said? All it would change is the substrate. It could be that some special physics is needed for me to function at all, that computation in a classical world would be much to inefficient to cause me, or that somehow formal systems like computers cannot even describe me (Penrose makes an argument like that), but even if some sort of Everett substrate is needed for me, how does that make my actions determined by anything other than my physical nature? Do I somehow step outside my substrate and physical nature to control what this multiverse processing is going to decide? Why would using a multiverse for processing make me any more or less an automaton (if that is why I am) than any other way in which I might exist? Even if Professor McFadden is right, that rather poor James Bond fiction that you read earlier was probably not the product of "free will"; it was merely the result of physical events happening in a reality described by the Everret many-worlds interpretation of quantum mechanics. This then, is my first problem: free will seems incoherent anyway, and finding different physics or substrates does not save it unless you can show why it makes things different.

My second objection is that the many-worlds interpretation of quantum mechanics does not really allow communication between "worlds". According to the many-worlds view things like the double-slit experiment are not really caused by communication between worlds, but rather because reality, at a small scale, does not immediately split into different worlds when quantum events happen. When an event starts to have any significance at all, "decoherence" of waves is supposed to occur: different parts of reality, even though in a sense they exist at the same time and in the same space, are, in a way "too out of step" to interact with each other and separate worlds exist. This is why we, as humans, are supposed to experience specific things like atoms. If any communication is supposed to occur across worlds in Professor McFadden's view then it is up against the problem of decoherence for a start. On the other hand, it could be argued that this communication is not really across worlds, that it really occurs before the worlds splits, in that weird superposition of states, or more accurately (since separate events could not be happening in such a situation) that what is going on in some thinking system could be based on some kind of superposition of states which then gives a single state with the answer. I would have no problem with this in principle. In fact, if the many-worlds interpretation is true this would not be too different from what quantum computers are doing, and there seems to have been some success in this area. However, I would remain to be convinced that it would be anything more than the brain tapping into quantum mechanics for faster computation and my objections about "free will" and "we are not automatons" would still stand. Furthermore, I would need convincing that such a level of coherence is even possible for neurons. Quantum computers can only work under very specific conditions. What would prevent the decoherence that we should expect in a brain and that, if it occurs, would prevent quantum thinking like this from working? Decoherence is easier to hold off at low temperature. I have not seen any evidence of cooling systems in my skull.

Another criticism I have is that quantum mechanics is a computable theory. If you are prepared to wait long enough you can model quantum systems on conventional computers. It could be that a "quantum thinking" brain does things vastly more efficiently than a conventional computer (and you already know I am sceptical about that), but whatever it did could still be modelled by a computer using just classical physics. How does this allow the quantum system to be somehow more profound than a normal computer, so much so that Professor McFadden uses words like "soul"? The many-worlds interpretation itself is an utterly computable model, violating the Church-Turing hypothesis in no respect when you consider the system as a whole. Maybe there is some non-computable aspect to quantum mechanics. Maybe the current description of the theory is wrong and people like Penrose are right. Until someone persuades me of that t, then I am going to be sceptical.

A further criticism I have is that some of this appears based on a kind of argument from ignorance. Professor McFadden has declared computation, in itself, inadequate for causing our minds, but how can he really know this? This seems to me a subjective view based more on his feeling that computation as an explanation for us is implausible. What is wrong with the idea that your thinking and mine is based on some unknown type of computation? Why is that any worse than invoking parallel worlds and (possibly) unknown physics?

My next criticisms are not based on Professor McFadden's views of minds and consciousness in itself, but on his theory of quantum evolution. I think this is still relevant because it is part of the same sort of idea, and any problems with it are indicators of possible problems with the work on minds and consciousness as well. Professor McFadden proposes that "quantum evolution" occurs - that the DNA somehow sniffs out different mutations, in different worlds, to select the right one. The big problem I have with this is decoherence. How would this process know that a mutation is harmful or beneficial? A mutation, in itself, is just that. It only has any effects - beneficial or harmful - when it causes a change in how an organism is built and how it works, and ultimately how that organism interacts with its environment. Suppose a particular mutation makes an organism more likely to survive. This benefit is not realised in some tiny fraction of time after the mutation, and it is certainly not realised during some weird quantum superposition of states. It is realised much later, when the mutation is out there with an organism, struggling to survive in the world. In any view of the many-worlds interpretation that would have been recognised by Everett, this is an age after decoherence has occured. As an analogy, this seems to me almost like proposing quantum politics, where different political ideas can be tried out in the multiverse and the best one can somehow be accessible to us. Decoherence would seem to stop such things - and before you try to dismiss decoherence it is supposed to be why we even experience a world like ours, with specific events, at all.

The other criticism I have of the quantum evolution idea is based on this text by Professor McFadden:

"Consider those chemicals sloshing through the primordial sea: billions of molecules colliding, combining, breaking and reforming to make trillions of new molecules. The likelihood of their random collisions generating a molecule, or even a microbe, complex enough to replicate itself is minute, far too low to have occurred on this planet or even a billion planets. But … and this is a crucial but, that probability is not zero. So long as the events took place in the quantum multiverse then everything that could happen would happen. If there is some non-zero probability that a fantastically improbable combination of molecular collisions could lead to life, then it WILL lead to life, in the multiverse. Life will emerge. And life, by replicating itself to generate larger more complex structures will crash out of the multiverse to yield the world we know. Any small pond could generate life, if it had access to the quantum multiverse."

I actually have no problem with this! if Everett's many-worlds view is correct then yes - it seems that anything that can theoretically happen will happen - "somewhere". It would therefore be useless to suggest that the origins of life, or evolution hitting on the right sequence of mutations, would be unlikely. These things would happen somewhere in the multiverse and the anthropic principle could then be used to explain we we are here: we would have to be down one of those branches where intelligent life evolved. I think this, in itself, is fine. It may not be needed or it may be: it depends on how likely you think the origin of self-replication is and some of the later events in evolution. The problem I have with this is that it is not the same as his quantum evolution argument. At one stage, Professor McFadden seems to be proposing that DNA can somehow "sniff out" different mutation options in the multiverse simultaneously and that such an idea is needed to explain evolution's success because it would otherwise be too unlikely for life to be here. Then he proposes that everything happens and that the Everett view guarantees that even unlikely sequences of events must happen somewhere. If true, this would actually remove a lot the need for these clever, adaptive mutations to explain life! If we are here because "everything happens" then it would seem that the origin of life, and the evolution of early life would be inevitable anyway, no matter how unlikely even with a conventional view of mutation in an Everett multiverse, merely as a result of its vastness and scope for things to happen. What, then, is Professor McFadeen suggesting?

In view of these issues, and given that the research paper is based on Professor McFadden's work, and given that I have not seen any evidence of actual artificial intelligence emerging from this - in the sense of artificial systems doing intelligent things - I remain sceptical until someone can set me straight on my objections.

My take on McFadden's theory and of the paper presented in this article is that it may well be an important contribution to neuroscience, in the sense that field interactions may prove to contribute importantly to the overall dynamics of the human brain. Such field interactions may turn out to explain some mental phenomena, which would be important.

However, I think that's as far as it can go. Calling EMF interactions the seat of consciousness is just hand-waving, because it doesn't actually explain anything "hard" about consciousness (in the sense of Chalmers' "Hard Problem of Consciousness"), although it can potentially solve the "easy" binding problem (solving the binding problem was the impetus behind McFadden's theory). In other words, McFadden isn't able to explain qualia, or how subjective experience arises from a physical system. Talking about field interactions only moves the problem into a different physical aspect of our brains.

That's my initial objection. I also wonder about how far you can take the idea of a global field of interaction because field effects would be so contingent on the physical layout of the brain, and affect all neurons in a given area indiscriminately. Field effects in one region of the brain will have a more significant effect on neighboring regions of the brain than in remote regions, and yet the perceptual elements that we routinely bind together often come from remote areas of the brain. In other words, it's easy to see how field interactions could help synchronize the phases of neurons that were close to each other, but much harder to see how such influence could be selective, which would be a key component of any real solution to the binding problem.

My next objection stems from an unconventional point of view, which is that the binding problem is not really a problem, in the sense that we look for a particular place in the brain where the neural correlates to our experience all come together. It appears to be a serious problem, because we make one very important, but flawed assumption: that our perception is actually unified. That's a fiction that we are hypnotized into believing as small children when we are given our names and language and the idea that we cause our own actions.

The unified feel of our experience is a magic trick, and the magician is the ego. It really does unify our perception in the sense that we relate everything in our experience to "I". But "I" is a fiction, and cannot be expected to be located in some center of the brain. In other words, my concept of self has no physical basis in reality. It's a convenience, a social convention, and it would be unnecessary if we weren't social creatures. Yet, we treat the ego as if it were real, and as such, are blind to the possibility that our perception is ultimately a patchwork of competing stimuli pulled together - with effort - by mental processes we are given by society at a very young age.

I will have much more to say about this in an upcoming article. I will leave it at that for now, and just say that McFadden's cemi theory tries to explain something - the binding problem - that goes away when looked at from a different perspective. Therefore, I think the very basis of its formulation as a theory of consciousness is flawed, even if cemi field interactions do prove to be a useful and innovative contribution to the physical dynamics of the brain.

I only needed to read the 3rd paragraph of your post to know I shouldn't read this guy's work. This is god of the gaps type of nonsense. The many worlds interpretation of QM is just that, an interpretation. It does not really bring anything new in the form of consequences to the physics we can observe. It does however leave open the option for what you suggest, namely that even unlikely states can come to exist.

Quite frankly I am really confused that someone would think such things. It works great for scifi, no doubt, but coming from a professor and proposed as an actual theory, ah .. no. Where is this information stored and how is it processed? If a molecule has access to this information it needs to exist, does it not? Where does this magical life force reside?

We know quite little about the conditions that led to the creation of life. It undoubtedly took a lot of time and chemistry to get basic life started on earth( true even if seeded from space) so we have no real idea what the probability of it happening really is.

I would agree that this does not seem to contribute to any solution of the "hard problem of consciousness". This is very similar to the way in which it fails to contribute to the issue of "free will" (which some people feel exists in some profound way and I don't - I don't have any problem with the idea that something I call "I" exists as some kind of abstraction of lots of smaller processes in my brain, but I suspect that when I have the experience of "choosing" a course of action it would probably be more accurate to say that what I really experience is observing my brain making a choice). All that is being proposed, it seems to me, by Professor McFadden is a different mechanism. At best it would be a mechanism that can do things that Turing machines cannot do (which we should need a lot of persuading of, to put it mildly). Failing that, it would help if it could do what Turing engines can do, but vastly more efficiently than Turing engines that could be built in other ways. Even if it achieves something like that (and that is a fairly big "if"), the idea would just be a formal description of how certain things happen in nature, with no indication of how this maps onto human subjective experience. Even if the theory accounts for every aspect of why I do what I do, even if something called "consciousness 1" is described in the theory, this does not explain why this "maps onto" my subjective experience of consciousness (let's call that "consciousness 2") anymore than any other formal description of things in nature would. Are we supposed to just assume it would? If so, why not just assume that a formal description of a computer doing something could "map onto" our subjective experience? A philosophical gulf - rather an abyss - has to be crossed between attempted explanations of how brains work and answers to the hard problem of consciousness and I am aware of no reason why these ideas achieve this. In fact, I have serious doubts that it is even a question that can be answered. If all science can do is try to make accurate formal descriptions of nature I do not see how science can do anything other than tell us that a particular formal description of some part of nature appears to be a candidate for our mental experience. Science may tell me that my brain is doing things that appear to "match" my mental experiences and it may tell me that a piece of rock isn't, or that a houseplant isn't. For example, some theory may say:

"Such and such a description of some part of nature has something called S in it. S's working involves some features s1, s2, s3, s4, etc. Now, observe that your experience of consciousness involves some characteristics that you call c1, c2, c3, c4, etc. If we say that s1=c1, s2=c2, s3=c3, s4=c4, etc, then there is no disagreement between S and C (you could substitute c1 for s1, etc in any description and it would still work) so S=C is reasonable." - rather like showing someone Jaws in a movie theatre and then showing them the reel of film from the projector and showing how individual bits of the film map onto their memories of what they saw.

and I suspect that is all you could ever say - and some people would find that very unfulfilling and say, "Yes, but please answer the hard problem of consciousness!" - but what else could we ever hope to do? From the point of view of the hard problem of consciousness, S agrees with C is just not good enough. People want to know why S is C: I think the futility of trying to deal with this is clear.

Whatever the answer, if any, is, I am fairly sure that this is not it and this makes me somewhat suspicious of the whole idea. If the theory is claimed to solve the "hard problem of consciousness" and explain "free will", then someone like me is going to have doubts about the entire thing - extreme philosophical claims and more mundane technical claims.

If it is true, it may really be advisable for me to keep my mobile phone away from my head.