X-Message-Number: 11626 From: Thomas Donaldson <> Subject: More for Mike Perry: simulated persons and real persons Date: Tue, 27 Apr 1999 23:53:56 +1000 (EST) More for Mike Perry: The important point behind what I was saying in my previous message (why Turing machines, or indeed any sequential computer, are quite inadequate as machines able to act in the world just as well as persons can) is that persons are much more than computational devices. Not only are persons more than computational devices, but if you try to see them ONLY in terms of computation you will very likely miss the construction of a real person by light years. At the lowest, brute level, acting in the world involves such things as being able to name and manipulate real objects --- which necessarily fail to resemble any ideal object at all closely. Not only that, but even to use real language (I claim) will ultimately require much more than just the ability to compute things (I will allude here to the major point I made a few weeks ago: primarily, we deal with the world, not with language, and it is very unlikely that any device can be built able to use English --- or any other human language --- as well as humans if that device consists only of a computer, a keyboard, and a video screen(*)). You cannot simply assume such an ability, as I said before. If you believe such a machine is possible, you must prove it... and by so doing prove, as I said before, that given the vocabulary and grammar of English then no other mapping of English onto reality is possible other than the one we learned by learning to name things in the world. We did not learn the definitions of elementary words in terms of other words, we learned them by seeing how the world behaved and acting in that world. And it is this relation to reality which makes Turing machines, as means to analyze how humans think, so ludicrously inadequate. For such purposes as we use our brains, it is essential and not at all just a side issue that we are so highly parallel. Simply to learn to talk (and hear) we use that parallelism; when I said that much more is needed than just a computer to make a person (artificial or not) I was not referring to drives such as hunger or thirst. Yes, we would need such drives too, but the FUNDAMENTAL problem shows up even for a machine able to think about the real world and plan actions to be done in it (rapidly enough that they might really be done, of course). I know that various people will point out how fast computers have become in doing various calculations etc etc. Yes, and WE HUMAN BEINGS built them to be so fast on tasks for which our own evolution has not caused us to be fast (except in terms of vision and graphics, which is known to require very great --- but also highly specialized --- computer power). That point emphasizes just how parallel we really are, while at the same time pointing out that for evolutionary reasons we simply did not evolve the ability to do the calculations needed to design a jet plane or understand chemical reactions at similar speeds. That is why we built computers to help us do those things --- while no one yet has needed a computer to allow them to look at the world. And yes, we may build artificial systems to allow those to see who are not only blind because they lack eyes, but blind because they lack the brain areas involved in our visual cortex. Those systems may be built on biological principles using biological methods, or by some other method. Not only that, but our past evolution says nothing about our future needs: if someone really needs in their brain (rather than in handy computing devices which can be changed for others at will, just as with other tools) an ability to design wings for jet planes, then that ability can be produced as part of them too (**). Again, by biological or by other methods. There isn't even any BIOLOGICAL reason why our neurons could not be redesigned to work faster --- they work as fast as they do, again, because our evolutionary history has not required them to work faster. In short, if you adopt Turing machines as models for how our brain works, you miss the entire point of what constitutes a person rather than a computer. And the very most you could produce would be a simulated person (NOT aware) in a simulated world --- and hardly a very well simulated person, either. As for awareness, they will be no more aware than simulated rainfall can make anyone wet. Best and long long life to all, Thomas Donaldson (*) as is the normal setup for the Turing test. (**) there are enough such abilities that separating them from us as tools seems to me much the most efficient means Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=11626