Friday

consciousness = complexity/time

i'd like to return to my thoughts on the mind briefly, if i may.

it occurred to me while reading the emperor's new mind that no-one seems to seriously be taking time into account as a factor of consciousness. penrose seemed to be hinting in towards this, perhaps even directing my thoughts, but i haven't read it anywhere explicitly; i think time-scale is key to consciousness. think about it; think about how fast an animal's internal biology moves, in terms of enzyme reactions, nerve impulses and the like. our consciousness-generating neurons fire perhaps billions of times a second, yet consciousness is something that takes place over individual seconds - which is surely why seconds are the length they are. if we were to run a simulation of consciousness on a machine at a rate we can monitor, it would not just have to be scaled down in terms of complexity, but also slowed down so much that it would take a very long time to be recognised as conscious - perhaps only by another mind running at the same rate.

in discussions about consciouness, the differences and similarities between brains and computers are always brought up. the problem that i don't think is addressed with enough weight is that each neuron of our brain is a computer, not the whole thing. to replicate, or even create, consciousness, we need a large number of linked-in machines, each connected to many others. a human brain has what, 100 billion neurons, 100 trillion synapses. the plot of 'the moon is a harsh mistress' almost sounds plausible, although the internet has got a long way to come to achieve this; but what it also needs, is a reality as rich as the one a human inhabits. as the nazis demonstrated, lock a baby in a room with enough food &c. to live on, and it will still die through lack of stimulation and affection [citation needed]. without an external reality to stimulate us, our minds rot, or if we are young enough, fail to develop.

of course, any one machine can simulate the running and the commumication between several other machines at a lower rate; a classic computing time-space pay off. so one immensely powerful - perhaps impossibly powerful - machine could concivebly model all the neurons in a brain talking to each other. we can do this now with simple artificial neural networks. penrose, however, does make the quite good point that consciousness might not just be an accurate model of neurons firing, but something to do with the underlying laws of physics. let's say you could model those on a machine (although quantum theory might have other implications), but the complexity in doing anything remotely near the scale of a real brain, even a simple one, would be probitively horrendous. consider a machine modelling the laws of physics required to run itself; at what fraction of it's own potential speed would it be running the simulation at? i think that's quite a good way of telling, and perfectly illustrates my point about time being the crucial factor. it would ultimately be a perfect representation of itself, just massively slower.

so the internet _is_ like a brain; so is society. so is any graph of nodes. it is only a matter of scale - as in the number of nodes and the number of connections - and time - as in the relative time based on the frequency of communications. but also, i conject, a matter of clear division between environment and self. which is not something you'd often hear me say, and i might have to think about that a bit.
Post a Comment