Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!thunder.mcrcim.mcgill.edu!snorkelwacker.mit.edu!apple!sun-barr!rutgers!aramis.rutgers.edu!athos.rutgers.edu!nanotech From: lovejoy@alc.com (Alan Lovejoy) Newsgroups: sci.nanotech Subject: Re: Is uploading suicide? Message-ID: Date: 15 Jan 91 22:16:33 GMT Sender: nanotech@athos.rutgers.edu Organization: Ascent Logic Corporation; San Jose, CA Lines: 293 Approved: nanotech@aramis.rutgers.edu In article franz@cs.washington.edu (Franz G. Amador) writes: >It seems to me that the concept of uploading has a fundamental and >unavoidable flaw. Namely, there is no way to tell if an uploading has >been successful. By "successful" I mean that one's consciousness has >been copied into the computer, not merely one's behavior patterns. How do I know that YOU are conscious? What objective measurement can I undertake to reveal that ANY individual is conscious? Is it not possible that what I experience as consciousness is unique to me--and different in kind from what any other "intelligence" in the whole universe experiences? What proof can you give me that you are not merely a machine that appears to mimic my internal mechanisms and external behavior, but in which no "consciousness"--as I experience it--actually exists? And why do you believe that other human beings are conscious? Just because we all say we are? Doesn't the fact that we all have so much trouble defining what consciousness *IS* in a formally rigourous way cause you *ANY* doubt whatsoever that all or most of us are lying--or mistaken--about being conscious? If we all experience pretty much the same thing, shouldn't we be able to agree on its description/definition? You and I don't share the same molecules, the same body, the same memories or the same identity. Yet we accept--mostly--the idea that both of us are "conscious"--whatever that is. Why? Essentially, because of the similarities between us. We consist of atoms, organized into the same types of molecules. Our structure, in fact, is VERY highly co-analogous/similar from atoms to molecules to cells to tissues to organs to our bodies as a whole. Not only that, our behavior is highly similar--again from atoms to bodies. And we observe that, as a general rule, functional/behavioral similarity correlates very highly with structural similarity. I accept that others are conscious because I know that they are very, very much like me, and my experience demonstrates to me that such a high degree of similarity of observable properties correlates highly with a high degree of similarity of ALL properties, whether they are observable or not. Lacking a "consciousness detector"--a la Star Trek--this "proof by analogy" is the only way available for deciding who--or what--is conscious. >Whether these points bother the reader depends, I suppose, upon >whether he or she believes that a "perfect" simulation will >necessarily create consciousness. I do not see why it should, but the >answer is immaterial to my argument. If you decide to risk your >consciousness by uploading, it will be an act of faith without a >guarantee, because there is no empirical means of determining if any >prior uploading has been successful. Even if you decide to return to >material reality after a fixed time limit, and your uploaded self >doesn't change its mind, you still won't be able to know if you were >conscious while in the computer. > >So if you really can't tell the difference while made of flesh and >blood (not meat - meat is dead muscle tissue), and you can return to >physical form whenever you or your simulation wants to, what >difference does it make whether you are really conscious while >uploaded? It is the difference between life and death. If your >consciousness is not there while uploaded, then so long as you stay in >the computer, you are dead. If existence in there really is >preferable to that in the material world, then your simulated self >might choose never to return, and by uploading you are committing >suicide. > >Franz Amador >franz@cs.washington.edu >{rutgers,cornell,ucsd,ubc-cs,tektronix}!uw-beaver!june!franz The key point at issue is this: is consciousness a function of the physical composition of an intelligent machine, or is it a function of the logical operation of an intelligent machine (that is, the semantic content or meaning of what the machine does)? Specifically, this is what we want to know: is consciousness a unique effect of neural networks implemented in a protein molecule medium? Or could it be a unique effect of neural networks? Or, might it simply be a general effect of intelligent machines whose degree depends upon the level of intelligence? What we are really seeking to discover is the level of abstraction at which the consciousness effect operates. That is, what transformations in the properties/structures/functions/components of a conscious machine preserve consciousness, and which do not. What are the minimal necessary and sufficient conditions for creating the consciousness effect? We know, for instance, that introducing certain molecules into the brain suspends the consciousness effect produced by that machine--as do other even more dramatic changes such as halting the blood flow. We also know that consciousness is largely independent of the physical and/or temporal locations at which a brain machine operates--among other things. By inductive analogy, we accept that it is also independent of the identity of the atoms used as the composition medium of a brain machine (one hydrogen atom appears to work just as well as another; in fact, there appears to be complete interchangeability of parts at the molecular level: any molecule of the same substance is as good as any other). However, the structures in which those molecules occur--where each molecule is in relation to the others--is observably rather important to the consciousness effect. And yet there is more than one molecular structure that a brain machine can have and still produce consciousness. There is, apparently, an infinite set of molecular structures of brain machines that produce consciousness--and an infinite set that do not. We also know that there is a very high correlation between changes in the molecular structure of a brain machine and changes in the precise nature of the consciousness ("mental state") produced by that brain machine. It is an experimentally reproducible fact that changes in conscious state correlate strongly with changes in the molecular structure of the brain, and that induced changes in the molecular structure of a brain modify the consciousness of that brain. It is even predictable to various degrees--depending on the nature of the change--how a change in the molecular structure of a brain will affect the consciousness, or how a change in consciousness will be reflected by a change in brain molecular structure. So why is brain molecular structure so special? For that matter, why is the molecular structure of anything significant? Or even better, why is the structure of anything significant? Now THAT's a fundamental question! Structure is simply the relations between things. A relationship is determined by the similarities and differences between two things. The full relationship between two things is a function of the differences in the values of all the attributes of those things. The distance relation between two objects is a function of the difference in the values of their space-time coordinates (location attributes). To ask how X relates to Y is to ask what are all the differences and similarities between X and Y. A system is a structured collection of interacting components. The definition or description of each component includes the factors that distinguish each component from every other component, as well as the factors that determine how each component interacts with--if at all--every other component. In other words, each component of a system is defined by its relation(ship)s to all the other components in the system. So asking why the structure of something is important is equivalent to asking about the importance of that thing's components and their interrelationships--which is another way of talking about the decomposability/decomposition of a system, and the differences/similarities between the system's components. The point I am trying to make is that structure, (de)composition, relationship, similarity and distinction must be understood as different "aspects" (ways of looking at) the same thing. And what is that thing? Now that's ANOTHER fundamental question! The answer, obviously enough, is reality itself. Ah, but what is reality? That, of course, is the ultimate question. All that we really do when we attempt to define reality is decompose it into components based on our ability to detect similarities and/or differences. We even do this merely by decomposing the state of the Universe into "separate" attribute values! To say that the universe has a temperature and a size is to assume that size and temperature are distinct attributes. It is just as valid to turn things upside down and claim that Nature is actually a set of inherently distinct things, which only appears to form an integrated whole--a "universe"--as a result of our mental abstraction processes. It is neither possible--nor necessary--to determine which of these completely inverse views is correct. The distinction is probably just as meaningless as the one between the waves and particles of atomic physics. One may either take unity to be the true state of things--in which case decomposition (division into components based on similarities and/or differences) is merely a semantic device, or one may take individuality as the true state of things--in which case abstraction (factoring out invariants by finding similarities and parameterizing based on differences) is merely a semantic device. Are you able to see these two inverse scenarios as the same thing viewed from a different perspective? I hope so. Did you notice that we are getting rather close to a recursive, self referential argument? That tends to happen with really fundamental questions. Perhaps it indicates that the question is wrong, meaningless or stupid. Perhaps it is enough to say that Reality is What Is. Period. Our investigation of What Is is necessarily conducted by observing it. To observe, we use our senses to detect--and our minds to interpret--signals from our enviroment. These signals convey information about the state of the world. Physical law determines a relationship between the state of the world and the signals generated by the world when in that state. Actually, physical law determines the future state(s) of the world as a function of the present state. It is our own arbitrary semantic system that differentiates certain aspects of the state of the world as "signals" that indicate the value(s) of other aspects of previous states of the world. This needs to be emphasized: the physical mechanism by which information is conveyed to us from our environemnt is actually a process of the universe transitioning to new states as determined by the previous state and natural law. We extract information from "signals" by deducing the previous state(s) of the universe that are implied by the present state of the universe--as represented by those objects we label "signals." The "information" carried by the "signal" is whatever can be "deduced" about previous universal states given the state of the "object(s)" or phenomen(on)(a) we use as the basis of our deductive reasoning. Let us call this the transmission of information by the interpretation of the states of objects/phenomena as tokens of the previous states of (perhaps other) objects/phenomena. The important characteristic of information conveyed in this way--that is, by "interpretation of tokens"--is that it must be true. It cannot be mendacious. It can certainly be misinterpreted, but that is an entirely different thing. Why? Because natural law guarantees what the predecessor states can be for any state of the universe--perhaps not always in an unambiguous way, but in a deterministic way. (Yes, I know about chaos--that's why I mentioned that things can be ambiguous. Chaotic behavior is still deterministic. The past and the future are still constrained by the present deterministically--just not uniquely). It would seem that the distinction we make between such "information" and the "tokens" that convey that information is rather artificial. The distinction serves no purpose. One can think of these "different" things as actually being the "same thing" without suffering any negative consequences. To prove this, one need only observe that *INFORMATION CANNOT EXIST SEPARATELY FROM THE TOKENS--OR SYMBOLS--THAT "ENCODE" IT*. ("Aha!", the reader says. "Now I begin to see what the last 15 paragraphs might have to do with the supposed subject of this article!!!". The reader might want to ponder the fact that Knowledge Representation is a very important subcategory of Artificial Intelligence research. Perhaps there is some deep reason for that...). Fortunately, the same information can be redundantly encoded by "different" tokens and/or symbols. But to say that the information is no longer encoded by any tokens anywhere is equivalent to saying that some prior state of the universe is no longer deducible from the present state. However, it IS possible to lie, to convey false information. To prove this, it is merely necessary to compare the public statements of any government with the facts :-). Humor aside, it is a well known fact that human beings, and perhaps some apes/chimpanzees, deliberatey send false messages. Distinguishing between the "necessarily true" information conveyed by physical tokens and the "NOT necessarily true" information conveyed by intelligent/conscious machines is rather important. At least, I think so :-). What is it that physically distinguishes tokens--which convey necessarily true information about the state of the world--from symbols--which convey not-necessarily true messages from one intellect to another? We have already determined that they are different logically (functionally) in terms of their dependability. What is the mechanistic, physical difference that motivates the difference in behavior? The difference does NOT lie in the physical medium in which the tokens or symbols are manifested. The words on this page--which are symbols--are composed of matter/energy. You read the words by using your eyes to detect the photons that are reflected off the page--or perhaps emitted by excited phosphors from a CRT screen. Your eyes/nervous system/brain deduce what letters are on the page by interpreting the information conveyed as tokens by these photons. But the information encoded by those letters and words is NOT deduced by reliance on natural law to compute all possible previous states of the universe that are implied by the present state. The meaning of these symbols does not depend on natural law, but rather on mutual agreement between the sender and the receiver of the message. In effect, they create their own "law" which defines the relationship between the symbols and their referents. Such "invented" laws only hold to the extent that parties to the agreement that creates the laws abide by the agreement. The universe enforces its own laws, because its laws are really defined by whatever it does. Fortunately, those laws allow both for predictability and unpredectability. Laws invented by agreement always involve adding more "predictability" or "symmetry" than nature does to the dynamics of some system. This can be done by taking local action, which is possible because some symmetries of nature can be broken locally. Local symmetry breaking means that "invented" symmetries can be locally enforced. So symbols can refer to that which is not, unlike tokens, because the "symmetries" (agreements) on which they depend are only valid locally. There exist regions of space-time where those symmetries do not hold. The map is not the territory. A computer simulation of a hydrogen molecule is NOT a hydrogen molecule, but merely a manipulation of SYMBOLS that represent the mathematically significant components of a hydrogen molecule. Consider, however, this question: when a video camera records a picture on video tape, is the information on the video tape recorded symbolically (possibly false) or as tokens (necessarily true)? Now consider the same question with respect to memories recorded in a brain machine. The relevance of all this to the original subject should now be starkly clear, I think. Although the fact that brain machines communicate--even internally--using symbols which can signal falsehoods proves that brain machines do store at least some information symbolically, that fact is not sufficient to prove that the mechanism of consciousness is a function of symbolic information processing. However, if it can be shown that all information stored and processed by brain machines is symbolic, it irrefutably follows that consciousness is a function of symbolic information processing--just like the processing that digital computers do! So then, are memories symbols or tokens? If memories can be false, they are symbolic--by definition. Need I say more? -- %%%% Alan Lovejoy %%%% | "Do not go gentle into that good night, % Ascent Logic Corp. % | Old age should burn and rave at the close of the day; UUCP: lovejoy@alc.com | Rage, rage at the dying of the light!" -- Dylan Thomas __Disclaimer: I do not speak for Ascent Logic Corp.; they do not speak for me! [It should be pointed out that lying (and cheating and stealing) are not necessarily functions of consciousness, as witness the viceroy butterfly, the cuckoo, etc. More specifically, sending false signals is likely to evolve in any situation where there is an agent which (a) does significant cognitive reduction and (b) has interests opposed to those of the agent evolving the deception. (a) simply means that you can't fool, e.g., plants, but only things which make what we might call decisions. You could arguably call that symbolic, but I don't. A housefly has a fairly well-understood neural circuit that connects its visual inputs directly to a stimulus to flight. This circuit can be "fooled"--that doesn't mean the fly is doing anything symbolic. --JoSH]