pd (01/31/83)
I would like to make a case/arguement against the feasiblity of constructing a machine that exhibits conciousness/idenitity. The frame work of the argument will be as follows: 1) Conciousness is a non-linguistic, non-mathematical experience. 2) Thus, one cannot convey the nature of one's experience of conciousness to another (by mathematical or linguistic means). 3) Thus one can never disprove or prove the existence one's own or any one else's conciousness thro mathematical or linguistic arguments. 4) Since one cannot mathematically model, or linguistically communicate one's experience of conciousness to another, it is impossible to build a machine that has conciousness. Furthermore, one can never prove or disprove that a machine has conciousness. I will only defend statement 1) above as follows: Any statement one makes about oneself has an active and a passive: (eg) consider the statement "'I' am a good person" 'I' makes a descriptive statement about him/her/itself. This duality will always be the case, what ever the statement the entity makes about itherimself. Itheshe may make meta, philosphical, recursively defining statements, but this will still prevail. (Go ahead, try it). Trying to do otherwise will be like chasing one's own tail. One's experience of one's own identity or conciousness is essentially uncommunicable, since all mathematical and linguistic descriptions/models are dualistic, whereas Conciousness is not. It is a monadic experience of oneself. Above is a justification of 1); hence 2), 3), and 4) Any takers ? Prem Devanbu
ka (02/01/83)
4) Since one cannot mathematically model, or linguistically communicate one's experience of conciousness to another, it is impossible to build a machine that has conciousness. Furthermore, one can never prove or disprove that a machine has conciousness. The two sentences above are contradictory. Kenneth Almquist
neiman (02/14/83)
Let me ramble for a moment--- Global statements like "It is impossible to define conciousness" bother me. "It is currently impossible to describe a conciousness exactly" would be more accurate. "It is impossible to define conciousness because it is non-linguistic and non-mathematical." So...do what any other scientists do when they are dealing with a concept too hairy to be defined explictly; create a model with enough simplifying assumptions so that it can be represented. A concious machine need not be aware in all the ways that a human being is aware in order to be concious. I would define conciousness ( were I not afraid to boldly go, etc) as the ability to examine one's own motivations and internal states. I am hungry, I am depressed, I am performing this action to obtain this result. Current computer systems do not have this capability; their instruction set/program is not available for examination and/or modification. A rule-based or script-driven AI program can examine its state and is therefore "more" concious then its predecessors. A Gedanken experiment: Suppose it were possible to take a human mind and copy it atom for atom so that you have two identical pieces of wetware. Would you have created another conciousness? Probably not, what you would have is one mind, and one fairly useless pile of organic matter. The difference would be analogous to the difference between a running VAX and a VAX taken down for maintenaince (their normal state). Suppose you were to repeat the experiment a little more carefully storing potentials also so that the state of the created object is *exactly* the same as that of the original. Is this concious? Is it an artficial intelligence? a. Don't say that this is impossible. Brains are being created all the time. I suspect that a (fearfully advanced) fabrication device could do the job as well as any genetic mechanism. b. A better argument might be that the construction was done without any real understanding. Well, yes, but once we've proved that conciousness can be achieved by a duplication of state, then the device which records that state is immaterial. One more random thought... Evolution took three billion years of fumbling in the dark to make something intelligent enough to be cocky about it. Computers have been around for about thirty years of directed evolution and have already gotten to a point where the hairless apes are getting nervous. My guess is that anyone who turns off a model-year 2083 IBM/DEC/XXX will probably get thirty years to life. dann (who, like cognitive scientists, ought to know better)