A Novel Theory

Picture:Dan Lloyd . I mentioned Dan Lloyd’s book “Radiant Cool” earlier. As I suggested then, linkages between philosophy and literature are not exactly unknown, but Lloyd’s is certainly the only book I know which splits neatly between a story in the front half and some really heavyweight stuff at the back.

The story is about Miranda Sharpe, who finds her Professor, Max Grue, slumped over his desk (unconscious?) when she creeps in to reclaim a folder. She leaves him there, only to find he has disappeared. Miranda has a series of encounters, with: Grue’s class; the arch-computationalist Clare Lucid; nerdy neural networker Gordon Fescue; Porfiry Marlov the former Soviet policeman and multidimensional scaling enthusiast; the menacing Zamm and Addit, who zap sections of her brain with Professor Cronkenstein’s transducer/activator; and finally, with ‘Dan Lloyd’. The mystery of what happened to Grue, the origins of the Chaos Bug which is currently infecting computers everywhere, and a few thoughts about consciousness, are cleared up along the way. It’s a readable narrative, with limited literary ambitions: you wouldn’t have been utterly surprised if Professor Plum had turned up in the library with a length of pipe at some point.

Some of Lloyd’s ideas get an exposition in ‘the thrill of phenomenology’, the first part of the book; a more systematic treatment, covering some additional ground appears in the second (and inevitably less readable) part, ‘ the real firefly’.

So what are these ideas? According to Lloyd, we have been too inclined to view people as ‘detectorheads’. Seeing the main function of neurons as detection has worked very well for a number of cognitive functions, but it won’t do for consciousness. Why not?

Lloyd presents a version of that old classic, the brain in a vat. If our brains were taken out of our bodies, they could still be fed with cunningly contrived signals which would give us the same experiences, phenomenologically, as we get from real life. We wouldn’t know the difference. We could be having the same experience of seeing a firefly whether in fact there is a real firefly out there or not. It follows that what we actually experience consciously isn’t the real world ‘out there’ at all (and consequently consciousness isn’t a matter of detection). I think this reasoning is mistaken. It’s true we can misinterpret our experiences, but that doesn’t mean that if our interpretation is wrong, our experiences are not really experiences of anything external.

Consider the mad scientists feeding our brains the signals which convince us we are still moving normally in the real world. Where do they get these extremely complex signals? By far the easiest way to get suitable signals would be to read them direct from reality with a camera and other suitable equipment. But in such a case, we obviously still are experiencing the real world – it’s just coming to us via the scientists’ recording apparatus. The scientists could use a small model world and tiny cameras instead, but that wouldn’t make a fundamental difference. We might be lulled into believing we were seeing the full-sized world, but the model is still a real, external thing. Now they could go further and use a computer model which existed only as data, they might even be able to devise a program which could generate appropriate signals without an explicit model: but however far they go along the path of abstraction we’ll still be detecting something outside our own brains. Our detectors may have been bamboozled and we may be quite wrong about what they’re detecting. But they’re not detecting nothing.

This doesn’t invalidate the rest of Lloyd’s account, however. He mentions two particularly important characteristics of consciousness – superposition and temporality – and offers a way of explaining them.

Superposition is the curious quality perceived objects have of being many different things at once. When you see a cup of coffee, you also see a container of liquid, a drink, somebody’s property, a cause of insomnia, and so on – and you see it as all these things simultaneously and immediately. How can this be? This is where the multidimensional scaling come in. If we like, we can treat each characteristic of a thing as a dimension. If we have the characteristic of redness, we imagine a line stretching from not red at all to utterly red: every object can be placed somewhere on this one-dimensional line according to its redness. Then we can construct an imaginary space out of these dimensions. Each position in this space will define a different combination of qualities, and hence, a different possible object. Now there are going to be an awful lot of dimensions involved in this imaginary space – Lloyd allows for billions (I think in fact that an infinite number of dimensions is required – and to complicate matters further, some characteristics are obviously related to others, rather than being capable of arbitrary independent variation). But that’s OK. The technique of multidimensional scaling apparently allows this very complex space to be boiled down to a much more comprehensible three dimensions, while preserving the distance relationships between salient points (with a certain amount of compromise). What this leaves (we hope) is a grouping of objects according to their resemblances and relationships. Hovering in the simplified 3-d space , dogs cats and fish will be close together because they are all animals; but fish will be a bit further off because they aren’t mammals, and they will also be part of another group, with apples and pancakes, because they are normal food items. Now, when you recognise something, you trigger some neural equivalent of this space which means you automatically see the thing you recognised in several different potential contexts at once. As you can tell, I have some reservations about the complex space apparatus here, but it does seem there is some gleam of light in the underlying idea.

When it comes to temporality, the influence of Husserl bulks large. According to Lloyd, every experience contains within it a strong sense of both past and future – retention and protention. To those with a good Husserlian background, this may seem evident, but things just don’t seem like that to me: some experiences have a strong temporal element, others don’t. According to Lloyd, the neural patterns which encode each experience of a given moment also encode, in a weakened form, the moment before and the moment after. Since the moment before itself contained a reflection of the moment before that, we have a kind of nesting effect.
To prove his point, Lloyd trained a neural network to predict the arrival of a ‘boop’ a fixed period after a ‘beep’, and then investigated it in considerable detail. It proved possible (using another neural network) to reconstruct earlier and later states of the network from any given point in the beep-boop process, a property not evident when it was simply running at random. Lloyd claims there is definite empirical evidence from scanning data that similar properties apply to the ‘neural networks’ in the brain. This tends to validate his view of the nature of consciousness as something suffused with temporality. Of course, using a task which is inherently about time – ie waiting the right length of time for the ‘boop’ – might be held to have biased the results.

Lloyd acknowledges scope for some reservations, and does not expect to have provoked the much-sought ‘aha!’ reaction of the reader who suddenly understands the whole thing at last. But he feels he’s made some pretty good progress. I’m not convinced he’s altogether on the right track, but the book is genuinely a useful prod towards some novel ways of thinking.

A lot of related material including some 3-d models and (apparently) an abortive exchange with Miranda Sharpe, can be found on Lloyd’s own website.