On 2/15/08, Eric B. Ramsay <[EMAIL PROTECTED]> wrote: > > I don't know when Lanier wrote the following but I would be interested to > know what the AI folks here think about his critique (or direct me to a > thread where this was already discussed). Also would someone be able to > re-state his rainstorm thought experiment more clearly -- I am not sure I > get it: > > http://www.jaronlanier.com/aichapter.html
I take it the target of his rainstorm argument is the idea that the essential features of consciousness are its information-processing properties. If you have a physical system that processes the same information as a brain does, but with a physical substrate different from biological neurons -- e.g. silicon chips, raindrops or asteroids -- then that physical system ought to be attributed the same[1] properties of consciousness that we attribute to the brain. Let's call this view Functionalism. His first rainstorm example posits a rainstorm and a possible computer such that when the computer takes the rainstorm as input, it performs the same information-processing as your brain does. Functionalism is committed to the idea that were that computer to be actualized and actually take that rainstorm as input and thus actually perform the same information-processing as your brain does, then that system would be conscious. Lanier suggests that Functionalism would say that simply the rainstorm taken by itself would be conscious, that this is ridiculous and that Functionalism ought to be rejected. Then he realizes that's a crappy Straw Man argument because the rainstorm by itself is just a passive potential program and not processing any information and goes on to his next example. Next, he seems to posit another possible computer that would "treat" a bigger rainstorm as both the rainstorm and computer of the last example. He seems to rely on the following implicit premise to conclude that there now really is an information-processing system isomorphic to your brain: If there is a possible computer that "treats" the bigger rainstorm as an information- processing system, then that rainstorm is in fact such an information-processing system. He tweaks this example in response to some imagined worries to make sure the rainstorm is treated as an information-processing system that changes over time isomorphic to your brain's dynamics. He expects us to conclude that according to Functionalism, such a rainstorm would in fact be an isomorphic information-processing system, hence conscious, and that Functionalism should therefore be rejected for delivering such an absurd result. Given his next tweak, it seems clear that whatever it is to "treat" something as an information-processing system, it is some sort of rather superficial correlation of various data that does not take into account any genuine causal influences. In his next example, he tries to fix this by positing asteroids which do have some sort of causal influence on each other and then performing the same trick of correlating asteroid data with brain data with a possible computer. He seems to think that the worry Functionalists would have is simply *that* the raindrops don't causally interact, which would be fixed by merely positing a system with *some* kind of causal interaction. But presumably, any sensible Functionalist would not just care about having some kind of causal interaction, but ones of the right kind, e.g. of the same complex sort that biological neurons have on each other. In order to really fix this example, it seems he would have to posit asteroids whose gravitational effects on each other are genuinely isomorphic to all the causal interactions the physical particles making up our brain have on each other. There'd be no need in this example to posit any possible computer to correlate data from asteroids with computations because the asteroids would have formed an actual computational system. But if there really were such an intricate network of interacting asteroids, it seems to me the Functionalist would no longer treat it as an absurd result and happily concede that the asteroids have miraculously formed into an actual computer, or at least an information-processing system, that is relevantly similar to a brain and therefore conscious. (I imagine the chances of this are vanishingly small (at least for most finite regions of space) since the dynamic information-processing we are interested in with respect to positing consciousness would probably require this complex network of asteroids to persist for some time with its functional integrity intact.) To be fair, some AI theorists and perhaps even philosophers (maybe Daniel Dennett?) do seem to have embraced the crucial implicit premise he relied on that all there is to being an information-processing system is that it be treated as such or even possibly treated as such. However, I see no reason to think Functionalism is committed to such a stupid view. I certainly don't (yet) have a fully worked out theory of this, but the most systematic attempt to address these issues can be found in philosopher Fred Dretske's book "Knowledge and the Flow of Information." He doesn't explicitly address issues of consciousness, but on plausible views of consciousness, understanding information-processing and representation is at least an all-important necessary first step (and the focus of Dretske's book). I'm sure many of you are familiar with Shannon's work on information theory, but that was mostly just studying quantity of information. Dretske is concerned not with how much information is (or can be) communicated but with providing a systematic *semantic* theory of *what* information is communicated and what the conditions are for such communication and processing of information.[2] I am beginning to suspect that understanding these philosophical issues (often called a theory of intentionality) is absolutely crucial to computationalizing meta-ethics and conceptual analysis, which I take to be prerequisites to the development of Friendly AI Theory and therefore, a safe Singularity. But that would be a whole other post/essay and I have a more traditional ethics dissertation I need to work on at the moment. ------- [1] The plausible view here is that "same" should be read as qualitative identity rather than numerical identity. If I buy a Toyota Corolla with my spouse, we have the numerically same car, that is, there is one car that we both own. If my friend then goes and buys a Toyota Corolla of the same model, then there's a sense in which we might say we own the same car, but what we mean is that there are two cars which we respectively own and they share roughly the same properties. It seems to me that Lanier equivocates on these two meanings of sameness when he asks,"Is it conscious as being specifically you, since it implements you?" and perhaps "You should realize by now that your brain is simultaneously implemented everywhere," seeming to suggest that, for instance, the existence of a rainstorm somewhere that is qualitatively identical in its information-processing as your brain is just as good as the continued existence of an information-processing system spatio-temporally and perhaps psychologically continuous with your brain, e.g. the normal way in which our selves persist into the future. [2] Unfortunately, I think Dretske's work is incomplete at best and from what I hear, he has lately converted to a theory of intentionality more focused on adaptive history. On such a view, if it turns out that you did not in fact have the evolutionary and psychological history you thought you did but rather spontaneously formed a moment ago in the exact same physical configuration, you would not in fact have any representations (or consciousness to the extent that mental representations are a prerequisite to consciousness) no matter how much it seems to you that you really would be doing such things as forming a representation of (and being conscious of) reading these words. ------------------------------------------- singularity Archives: http://www.listbox.com/member/archive/11983/=now RSS Feed: http://www.listbox.com/member/archive/rss/11983/ Modify Your Subscription: http://www.listbox.com/member/?member_id=4007604&id_secret=96140713-a54b2b Powered by Listbox: http://www.listbox.com