darren@cs.city.ac.uk (Darren Whobrey) (01/30/90)
Despite being somewhat off the mark, Searle's Chinese Room problem does raise some interesting points. Firstly, the room does not understand Chinese and, secondly, that non-self-referencial algorithms cannot embed symbols with internal meaning. Thus he concludes that computers can never understand or think, and that strong AI is, well, lacking. Before we make such bold claims don't we have to define what we mean by 'understanding' and 'consciousness'? What is it in our minds that understands, or is conscious? If you introspect on this for a few moments you'll no doubt start thinking in circles, or that you don't really understand anything at all. What if we have a self-referencial algorithm that imparts meaning to its symbols in an internally consistent manner. 'Consciousness' is deliberately programmed, or modelled, via a self-referencial algorithm, and is not some quirk arising from a very complex system as some would suggest. In Searle's terms the mind-model that the algorithm simlutes 'understands' or is 'conscious'. Not the room or person in the room, or the rules they are following, but the model itself understands, has internal meaning etc. The Whirlpool in the Room ------------------------- Suppose we have a whirlpool in the room and our little helper (Searle say) is standing in it's eye. Now this whirlpool is rather special for its made from ping-pong balls, which for our purposes are symbolic of molecules (water, nitrogen, whatever). In order to maintain the whirlpool our helper must apply some rules to each ball. Specifically, move each with a certain velocity. What prevents him getting wet (however you want to define getting wet by ping-pong balls)? Suppose further that these ping-pong balls are pretty sly, they've been lisening to Searle's argument that syntactically manipluated symbols cannot have meaning, so all at once they think to themselves 'Hey, we're not a whirlpool and thus shouldn't be wizzing around this person'. Immediately they collapse in upon our helper drenching him to the skin. The above scenario could quite as easily have been carried out on a computer of any sort, as Searle reminds us. The point is though, that firstly the model the algorithm is simulating is what's important, e.g. the whirlpool, and not the ping-pong balls, the helper, the rules he's following, or the room. Secondly, the room and it's contents has it's own internal semantics. Compare the whirlpool to consciousness, it's derived from rules applied to symbols which have meaning steming from their interaction and relation to each other. I know the above scenario isn't air-tight, but I think it's amusing anyway. Basically we have to consider self-referencial systems with either direct feedback, or some decaying recursive feedback process, if we are ever to create artificially conscious machines. This is what gives the system meaning. Darren Whobrey, e-mail: Janet: darren@uk.ac.city.cs City University, God was satisfied with his own work, London. and that is fatal. Butler, 1912.