Consciousness speed-dating

datingWhy can’t we solve the problem of consciousness? That is the question asked by a recent Guardian piece.  The account given there is not bad at all; excellent by journalistic standards, although I think it probably overstates the significance of Francis Crick’s intervention.  His book was well worth reading, but in spite of the title his hypothesis had ceased to be astonishing quite a while before. Surely also a little odd to have Colin McGinn named only as Ted Honderich’s adversary when his own Mysterian views are so much more widely cited. Still the piece makes a good point; lots of Davids and not a few Samsons have gone up against this particular Goliath, yet the giant is still on his feet.

Well, if several decades of great minds can’t do the job, why not throw a few dozen more at it? The Edge, in its annual question this year, asks its strike force of intellectuals to tackle the question: What do you think about machines that think? This evoked no fewer than 186 responses. Some of the respondents are old hands at the consciousness game, notably Dan Dennett; we must also tip our hat to our friend Arnold Trehub, who briefly denounces the idea that artefactual machines can think. It’s certainly true, in my own opinion, that we are nowhere near thinkng machines, and in fact it’s not clear that we are getting materially closer: what we have got is splendid machines that clearly don’t think at all but are increasingly good at doing tasks we previously believed needed thought. You could argue that eliminating the need for thought was Babbage’s project right from the beginning, and we know that Turing discarded the question ‘Can machines think?’ as not worthy of an answer.

186 answers is of course, at least 185 more than we really wanted, and those are not good odds of getting even a congenial analysis. In fact, the rapid succession of views, some well-informed, others perhaps shooting from the hip to a degree, is rather exhausting: the effect is like a dreadfully prolonged session of speed dating: like my theory? No? Well don’t worry, there are 180 more on the way immediately. It is sort of fun to surf the wave of punditry, but I’d be surprised to hear that many people were still with the programme when it got to view number 186 (which, despairingly or perhaps refreshingly, is a picture).

Honestly. though, why can’t we solve the problem of consciousness? Could it be that there is something fundamentally wrong? Colin McGinn, of course, argues that we can never understand consciousness because of cognitive closure; there’s no real mystery about it, but our mental toolset just doesn’t allow us to get to the answer.  McGinn makes a good case, but I think that human cognition is not formal enough to be affected by a closure of this kind; and if it were, I think we should most likely remain blissfully unaware of it: if we were unable to understand consciousness, we shouldn’t see any problem with it either.

Perhaps, though, the whole idea of consciousness as conceived in contemporary Western thought is just wrong? It does seem to be the case that non-European schools of philosophy construe the world in ways that mean a problem of consciousness never really arises. For that matter, the ancient Greeks and Romans did not really see the problem the way we do: although ancient philosophers discussed the soul and personal identity, they didn’t really worry about consciousness. Commonly people blame Western dualism for drawing too sharp a division between the world of the mind and the world of material objects: and the finger is usually pointed at Descartes in particular. Perhaps if we stopped thinking about a physical world and a non-physical mind the alleged problem would simply evaporate. If we thought of a world constituted by pure experience, not differentiated into two worlds, everything would seem perfectly natural?

Perhaps, but it’s not a trick I can pull off myself. I’m sure it’s true our thinking on this has changed over the years, and that the advent of computers, for example, meant that consciousness, and phenomenal consciousness in particular, became more salient than before. Consciousness provided the extra thing computers hadn’t got, answering our intuitive needs and itself being somewhat reshaped to fill the role.  William James, as we know, thought the idea was already on the way out in 1904: “A mere echo, the faint rumour left behind by the disappearing ‘soul’ upon the air of philosophy”; but over a hundred years later it still stands as one of the great enigmas.

Still, maybe if we send in another 200 intellectuals…?

Aliens are Robots

BISASusan Schneider’s recent paper argues that when we hear from alien civilisations, it’s almost bound to be super intelligent robots getting in touch, rather than little green men. She builds on Nick Bostrom’s much-discussed argument that we’re all living in a simulation.

Actually, Bostrom’s argument is more cautious than that, and more carefully framed. His claim is that at least one of the following propositions is true:
(1) the human species is very likely to go extinct before reaching a “posthuman” stage;
(2) any posthuman civilization is extremely unlikely to run a significant number of simulations of their evolutionary history (or variations thereof);
(3) we are almost certainly living in a computer simulation.

So that if we disbelieve the first two, we must accept the third.

In fact there are plenty of reasons to argue that the first two propositions are true. The first evokes ideas of nuclear catastrophe or an unexpected comet wiping us out in our prime, but equally it could just be that no post human stage is ever reached. We only know about the cultures of our own planet, but two of the longest lived – the Egyptian and the Chinese – were very stable, showing few signs of moving on towards post humanism. They made the odd technological advance, but they also let things slip: no more pyramids after the Old Kingdom; ocean-going junks abandoned before being fully exploited. Really only our current Western culture, stemming from the European Renaissance, has displayed a long run of consistent innovation; it may well be a weird anomaly and its five-hundred year momentum may well be temporary. Maybe our descendants will never go much further than we already have; maybe, thinking of Schneider’s case, the stars are basically inhabited by Ancient Egyptians who have been living comfortably for millions of years without ever discovering electricity.

The second proposition requires some very debatable assumptions, notably that consciousness is computable. But the notion of “simulation” also needs examination. Bostrom takes it that a computer simulation of consciousness is likely to be conscious, but I don’t think we’d assume a digital simulation of digestion would do actual digesting. The thing about a simulation is that by definition it leaves out certain aspects of the real phenomenon (otherwise it’s the phenomenon itself, not a simulation). Computer simulations normally leave out material reality, which could be a problem if we want real consciousness. Maybe it doesn’t matter for consciousness; Schneider argues strongly against any kind of biological requirement and it may well be that functional relations will do in the case of consciousness. There’s another issue, though; consciousness may be uniquely immune from simulation because of its strange epistemological greediness. What do I mean? Well, for a simulation of digestion we can write a list of all the entities to be dealt with – the foods we expect to enter the gut and their main components. It’s not an unmanageable task, and if we like we can leave out some items or some classes of item without thereby invalidating the simulation. Can we write a list of the possible contents of consciousness? No. I can think about any damn thing I like, including fictional and logically impossible entities. Can we work with a reduced set of mental contents? No; this ability to think about anything is of the essence.

All this gets much worse when Bostrom floats the idea that future ancestor simulations might themselves go on to be post human and run their own nested simulations, and so on. We must remember that he is really talking about simulated worlds, because his simulated ancestors need to have all the right inputs fed to them consistently. A simulated world has to be significantly smaller in information terms than the world that contains it; there isn’t going to be room within it to simulate the same world again at the same level of detail. Something has to give.

Without the indefinite nesting, though, there’s no good reason to suppose the simulated ancestors will ever outnumber the real people who ever lived in the real world. I suppose Bostrom thinks of his simulated people as taking up negligible space and running at speeds far beyond real life; but when you’re simulating everything, that starts to be questionable. The human brain may be the smallest and most economic way of doing what the human brain does.

Schneider argues that, given the same Whiggish optimism about human progress we mentioned earlier, we must assume that in due course fleshy humans will be superseded by faster and more capable silicon beings, either because robots have taken over the reins or because humans have gradually cyborgised themselves to the point where they are essentially super intelligent robots. Since these post human beings will live on for billions of years, it’s almost certain that when we make contact with aliens, that will be the kind we meet.

She is, curiously, uncertain about whether these beings will be conscious. She really means that they might be zombies, without phenomenal consciousness. I don’t really see how super intelligent beings like that could be without what Ned Block called access consciousness, the kind that allows us to solve problems, make plans, and generally think about stuff; I think Schneider would agree, although she tends to speak as though phenomenal, experiential consciousness was the only kind.

She concludes, reasonably enough, that the alien robots most likely will have full conscious experience. Moreover, because reverse engineering biological brains is probably the quick way to consciousness, she thinks that a particular kind of super intelligent AI is likely to predominate: biologically inspired superintelligent alien (BISA). She argues that although BISAs might in the end be incomprehensible, we can draw some tentative conclusions about BISA minds:
(i). Learning about the computational structure of the brain of the species that created the BISA can provide insight into the BISAs thinking patterns.
(ii) BISAs may have viewpoint invariant representations. (Surely they wouldn’t be very bright if they didn’t?)
(iii) BISAs will have language-like mental representations that are recursive and combinatorial. (Ditto.)
(iv) BISAs may have one or more global workspaces. (If you believe in global workspace theory, certainly. Why more than one, though – doesn’t that defeat the object? Global workspaces are useful because they’re global.)
(v) A BISA’s mental processing can be understood via functional decomposition.

I’ll throw in a strange one; I doubt whether BISAs would have identity, at least not the way we do. They would be computational processes in silicon: they could split, duplicate, and merge without difficulty. They could be copied exactly, so that the question of whether BISA x was the same as BISA y could become meaningless. For them, in fact, communicating and merging would differ only in degree. Something to bear in mind for that first contact, perhaps.

This is interesting stuff, but to me it’s slightly surprising to see it going on in philosophy departments; does this represent an unexpected revival of the belief that armchair reasoning can tell us important truths about the world?

Illusions for robots

robot illusionsNeural networks really seem to be going places recently. Last time I mentioned their use in sophisticated translation software, but they’re also steaming ahead with new successes in recognition of visual images. Recently there was a claim from MIT that the latest systems were catching up with primate brains at last. Also from MIT (also via MLU) though, has come an intriguing study into what we could call optical illusions for robots, which cause the systems to make mistakes which are incomprehensible to us primates. The graphics in the grid on the right apparently look like a selection of digits between one and six in the eyes of these recognition systems. Nobody really knows why, because of course neural networks are trained, not programmed, and develop their own inscrutable methods.

How then, if we don’t understand, could we ever create such illusions? Optical illusions for human beings exploit known methods of visual analysis used by the brain, but if we don’t know what method a neural network is using, we seem to be stymied. What the research team did is use one of their systems in reverse, getting it to create images instead of analysing them. These were then evaluated by a similar system and refined through several iterations until they were accepted with a very high level of certainty.

This seems quite peculiar and the first impression is that it rather seriously undermines our faith in the reliability of neural network systems. However, there’s one important caveat to take into account: the networks in question are ‘used to’ dealing with images in which the crucial part to be identified is small in relation to the whole. They are happy ignoring almost all of the image. So to achieve a fair comparison with human recognition we should perhaps think of the question being not ‘do these look like numbers to you?’ and more like ‘can you find one of the digits from one to six hidden somewhere in this image?’. On that basis the results seem easier to understand.

There still seem to be some interesting implications, though. The first is that, as with language, AI systems are achieving success with methods that do not much resemble those used by the human brain. There’s an irony in this happening with neural networks, because in the old dispute between GOFAI and networks it was the network people who were trying to follow a biological design, at least in outline.  The opposition wanted to treat cognition as a pure engineering problem; define what we need, identify the best way to deliver it, and don’t worry about copying the brain. This is the school of thought that likes to point out that we didn’t achieve flight be making machines with flapping, feathery wings. Early network theory, going right back to McCulloch and Pitts, held that we were better off designing something that looked at least broadly like the neurons in the brain. In fact, of course, the resemblance has never been that close, and the focus has generally been more on results than on replicating the structures and systems of biological brains; you could argue that modern neural networks are no more like the brain than fixed-wing aircraft are to birds (or bats).  At any rate, the prospect of equalling human performance without doing it the human way raises the same nightmare scenario I was talking about last time; robots that are not people but get treated as if they were (and perhaps people being treated like machines as a consequence.

A second issue is whether the deception which these systems fall into points to a general weakness. Could it be that these systems work very well when dealing with ‘ordinary’ images but continue go wildly off the rails when faced with certain kinds of unusual ones – even when being pout to practical use? It’s perhaps not very likely that  system is going to run into the kind of truly bizarre image we seem to be dealing with, but a more realistic concern might be the potential scope for sabotage or subversion on the part of some malefactor.  One safeguard against this possibility is that the images in question were designed by, as it were, sister systems, ones that worked pretty much the same way and presumably shared the same quirks. Without owning one of these systems yourself it might be difficult to devise illusions that worked – unless perhaps there are general illusions that all network systems are more or less equally likely to be fooled by? That doesn’t seem very likely, but it might be an interesting research project.  The other safeguard is that these systems are not likely to be used without some additional safeguards, perhaps even more contextual processing of broadly the kind that the human mind obviously brings to the task.

The third question is – what is it like to be an AI deceived by an illusion? There’s no reason to think that these machines have subjective experience – unless you’re one of those who is prepared to grant a dim glow of awareness to quite simple machines – but what if some cyborg with a human brain, or a future conscious robot, had systems like these as part of its processing apparatus rather than the ones provided by the human brain?  It’s not implausible that the immense plasticity of the human brain would allow the inputs to be translated into normal visual experience, or something like it.  On the whole I think this is the most likely result, although there might be quirks or deficits (or hey, enhancements, why not) in the visual experience.  The second possibility is that the experience would be completely weird and inexpressible and although the cyborg/robot would be able to negotiate the world just fine, its experience would be like nothing we’ve ever had, perhaps like nothing we can imagine.

The third possibility is that it would be like nothing. There would be no experience as such; the data and the knowledge about the surroundings would appear in the cyborg/human’s brain but there would be nothing it was like for that to happen.  This is the answer qualophile scpetice would expect for a pure robot brain, but the cyborg is more worrying. Human beings are supposed to experience qualia, but when do they arise? Is it only after all the visual processing has been done – when the data ariive in the ‘Cartesian Theatre’ which Dennett has often told us does not exist? Is it, instead, in the visual processing modules or at the visual processing stage? If so, then we were wrong to doubt that MIT’s systems are not having experiences. Perhaps the cyborg gets flawed or partial qualia – but what would that even mean..?