Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!uwm.edu!ogicse!moe!maxwebb From: maxwebb@moe.cse.ogi.edu (Max G. Webb) Newsgroups: comp.ai.philosophy Subject: Re: Conciousness Summary: Let's try for better definitions. Message-ID: <20909@ogicse.ogi.edu> Date: 2 May 91 19:39:35 GMT References: <2102@seti.inria.fr> <2124@seti.inria.fr> <11611@uwm.edu> <74193@eerie.acsu.Buffalo.EDU> Sender: news@ogicse.ogi.edu Organization: Oregon Graduate Institute - Computer Science & Engineering Lines: 81 o>From: oistony@ubvmsd.cc.buffalo.edu (Anthony Petro) o>but then, i'm not sure what it would mean to be conscious but not o>intelligent; seems to me that it might be possible to be intelligent o>and not conscious, but not possible to be conscious but not o>intelligent... o>anthony m. petro I am going to try to sketch definitions of 'intelligence', that don't hinge on the mental operation of _putting yourself in another's shoes_ an operation that doesn't work too well with nonhumans. (I am not hoping to settle the issue once and for all [HAH!] but to get your comments). Later, I have some questions about 'intentionality') Define 'emergence' as global computation by local adaptation (meaning local, simple rules). Vague, but at least something an engineer could start with. It seems to me that intelligence is what we call emergent adaptive behavior. As alan watts once said, 'thoughts grow in our brains like grass'. They have an apparent autonomy and independence, while possessing utility (that is, maxim- izing some internal reward or measure). It is this that con- ventional computers lack, and in lacking, are asserted to lack intelligence. (they don't write their _own_ programs, do they, Dad?) Neural nets have this autonomy, to some degree. But a level of emergent adaptive behavior comparable to our own, requires internal representations of the world and our- selves. Having language, we attach the word 'self- awareness' to the self-observed interplay between the model of ourselves and our self-observed behavior. Having language, we attach 'consciousness' to the percieved free and independent way in which perceptions, patterns and thoughts grow and die (c.f. 'stream of consciousness') Finally, pain. certain stimuli have a way of killing thoughts, and behaviors they are associated with (reducing their frequency). Any field of emergent computation/behavior w/ a similar mechanism should, I think, be said to have intelligence _and_ consciousness _and_ pain, to at least some degree. o>this is most interesting, particularly in that the exact o>opposite may be surmised. dennett's intentionality proposes o>that everything has some degree of intelligence or "inten- o>tionality," the extent determined by the degree of internal o>representation of its environment the thing has. Based on the spelling of the word, this sounds like an attempt to define in a verifiable way the concept of 'having purpose'. This sounds like a great definition, but there seem to me to be lot's of ways it breaks. 1) Any backprop network which learns to approximate some mapping must have *some* kind of representation of it's surrounding environment, but you'll never be able to tell whether it has one, or how detailed it is by look- ing at the weights. In other words, you have to fall back on some behavior based definition here. 2) Does 'internal representation' mean that by looking at the innards of the entity, we can get information about it's surrounding? _That_ is true of many objects we would not care to associate purpose with. E.G. does a camera w/ it's shutter opened briefly become more 'intentional' than it was before? If it does, then I submit the definition doesn't sound like it describes any- thing useful. If not, then you must start talking about the _behavior_ based on the internal representation to augment the definition - and again, we are forced into a criterion that is at least partially based on behavior. So, maybe I'm missing a subtlety here, but it sounds like yet another useless philosophical definition. Max