X-Message-Number: 22374 Date: Mon, 18 Aug 2003 00:59:12 -0700 From: Mike Perry <> Subject: "Simulations", Time Isomorphism Robert Ettinger, #22364, and my comments: >Getting back to descriptions and simulations--a description IS a simulation, >and a simulation IS (merely) a description. It would be good to clarify what is really meant by a "simulation." If people were clockwork machines and perfectly predictable, it might be practical to have an exact simulation, say in different materials, or as a running computer program that evolved a succession of state descriptions over time. (Or we might just have a static record of the successive states.) But people--and processes more generally--are unpredictable, so what I think is mainly of interest would be systems that exhibit certain similarities in their functioning to people, modeling the brain and its functioning for example. (For now I limit consideration to actively running systems rather than static records.) The modeled brain in this case would not, except in very unlikely circumstances, correspond to any real brain, but would perhaps correspond to a "typical" hypothetical brain. You could say the modeling is a "simulation" but in general how would you conclude even that? Supposing for a moment that it did exactly model some organic brain, you could say the organic brain simulated *it*. Why not? The organic brain could actually be constructed after the other system is built, as a simulation of it. Again, I doubt if anything approaching a true, event-for-event simulation would occur. But you could imagine making an artificial brain first, not an attempt to copy any existing brain but just to make something very like existing brains in certain basic ways. Then you make an organic brain that models the artificial brain as closely as you can. The organic brain is very like other organic brains but is not a twin of any. But it is a twin of what the inorganic brain models. So which is the "original" and which the "simulation"? >Dr. Badger and others reject time >isomorphism, but I know of no good reason to reject isomorphism for time >while >accepting it for other purposes. I have expressed what I think is a good way to deal with the time isomorphism problem, and others have made their tries. Here I will try again. But first I have to assume an isomorphism is possible. >Also, as I have said many times with many >examples, a running computer simulation is NOT fully isomorphic to a >person, and >cannot be. I'm not sure why in principle this could not be so. It seems that significant events in the real world are discrete, just as they are in a computer, and are Turing computable at the description level. (Certain complications need to be dealt with such as unpredictability but I think they can be.) This should be enough for any finite time's worth of full isomorphism. Present physical theories may be inadequate but something that rests on discreteness should be adequate. As long as the significant, discrete events (finite in number) are all modeled in their appropriate interrelationships, we could say our isomorphism is complete. (Of course this is only in a thought experiment and far beyond the abilities of present day computers.) Let's go back now to the time isomorphism problem. Imagine you have a static description of happenings at the subatomic, discrete-events level. All significant events are covered (all particle interactions that is). The description extends over a large volume of space-time, and includes many years' worth of the life of some conscious, human agent. The agent, that is, and the world he is part of, are embedded in your description. This world, really a history of a hypothetical world, becomes the "frame of reference" for that agent, and it is reasonable to say (in my view) that he is conscious relative to that frame of reference. In particular we could tell what that agent is doing, sensing, thinking, and feeling at different times by inspecting the record. We could observe that certain parts of the agent's brain are active at certain times, and so on, just as we would expect with a human in our world. Needless to say, this would have to be a very big record and we could imagine it is very far away from anything we can now observe, yet still extant somewhere. (Maybe you would have to tweak cosmology a fair amount for this, but let's allow it for the sake of argument.) Our own world, then, is clearly separate from this static-record world. To say that the agent is conscious relative to the static world does not have to imply he is conscious relative to our world. On the other hand, we can assume that there is some straightforward modeling of time in the static record, in the form of saying that event such and such has the following space-time coordinates (with respect to some chosen coordinate system) so time is represented isomorphically. So here we don't reject time isomorphism but it does not force us to conclude that the modeled agent is conscious, as we usually understand consciousness. To further clarify, we can now imagine an *active* system somewhere, in which events unfold in time rather than just being statically represented. So time is represented as we understand time, perhaps with a slowdown or speedup factor, but nothing beyond that. Again a world is modeled and our agent is conscious relative to it. Perhaps again this world is very far away but maybe we can reach it in Star-Trek fashion. At this point we should be able to interact with the agent, carry on a conversation and such. So this agent's frame of reference really is now our own. If in fact the "simulation" is being done in a certain way, we might immediately conclude the agent is conscious, as when the brain is organic and very similar to known cases of human brains. But even otherwise, if we accept the information paradigm, we would conclude on the basis of the isomorphism *and* the essential coincidence of the frames of reference, that the agent is conscious as we usually understand the term. In other words, when you can talk to the guy and he seems to be conscious by various tests, you have good ground to accept his consciousness as real. If, on the other hand, a space voyage uncovered the big static record instead, it still would not be possible to talk to the agent embedded in the description, and his frame of reference would still, by reasonable criteria, be different from and incongruent to ours. So again you would not have to consider him conscious as we usually understand the term, though still conscious in some other sense. >Once more, the assertion that a simulation would "be" a person is >nothing but dogma, with nothing whatever to back it up except the perceived >elegance of the concept. I don't see it that way, at least if we allow for possible future developments. It seems reasonable that a future, advanced but still inorganic robot could *seem* very conscious and human, both in its behavior and in the way its artificial brain was carefully modeling an organic counterpart at a deep and detailed level. In such a case I see no way in principle of establishing, through observational tests, that the robot was *not* conscious and really experiencing the feelings it seemed to have. So I would have no trouble accepting it as a person. Mike Perry Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=22374