X-Message-Number: 30207 From: Date: Mon, 24 Dec 2007 11:37:57 EST Subject: again, feelilng vs. computing Flavonoid wrote in part: >R.E. asserts that an animal is "not programmed in the way that a computer >is. The computer is language-based and digital, which is very >different." And he appears to conclude based on this assertion, that >therefore no computer can be programmed with a true intelligence >surpassing that of humans. Clearly he is not really paying attention. I have said repeatedly that, in principle, a computer might eventually surpass human intelligence, if we define intelligence to mean prediction or description or problem solving. In fact, in some areas computers have already surpassed humans. Also: > I see no reason to believe a computer could not be programmed >in the same manner that human/animal brains are. The reason, yet again, is that brains and digital computers are VERY DIFFERENT, and in fact we already KNOW that a computer could not be programmed in the same way that brains are. For a simple example, you cannot (with certain obvious exceptions) program a computer to give different outputs for the same input, whereas this happens often with brains. And you cannot program a computer to act in a certain way when it feels a certain way, because it doesn't feel. You cannot, for the foreseeable future, even program it to act in a certain way when it computes that a human would feel in a certain way, because we don't yet understand feeling. > He of course does not admit "feeling" [in computers] either. It shouldn't be much of a strain to understand that, since we cannot yet characterize feeling (subjectivity) in physical terms, it is premature to assume that this will ever be available to a digital algorithmic computer. As a crude example, if subjectivity depends on unique properties of carbon, then it cannot be duplicated in silicon. (And remember that "emulation" doesn't count. A description of a quale is not a quale.) He goes on to mention some current research on the capabilities of neurons, e.g. that touching a single neuron can induce sensation. Not relevant to the problem of subjectivity. Sensation can result from a touch to the skin or a photon on the retina, but this information, while useful, does not answer the basic question of the nature of qualia. He also continues to assume that, somehow, the dangerous computer will have its own agenda, and might be "hostile" or "indifferent." Yet again, computers don't have agendas in the sense we do, and are always indifferent to our value judgments, except as those can be expressed in very specific, detailed, and unambiguous terms, which means the computer is "satisfied" when it registers a particular set of numbers at a particular address, or one of a particular set of sets. He also ignored my suggestions for simple controls which surely any programmer would want. One example was a programmed requirement for human review before any "execute" order. Another possibility would be a requirement for specific predictions of some of the results of an execute order. He also ignored Kennita's note that there will be lots of frontier computers and programmers, and most of those will value life and be alert to danger. He complains that not enough is being done presently to face the dangers, ignoring the fact, which seems obvious to me, that we are leagues and leagues away from singularity-level computers. R.E. **************************************See AOL's top rated recipes (http://food.aol.com/top-rated-recipes?NCID=aoltop00030000000004) Content-Type: text/html; charset="US-ASCII" [ AUTOMATICALLY SKIPPING HTML ENCODING! ] Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=30207