--- In Wittrs@xxxxxxxxxxxxxxx, "jrstern" <jrstern@...> wrote: >> I think it poses a problem for the more physicalist views that >> come up in AI discussions - the views of Eray, for example. >> For it seems clear that much of what is described in CTM occurs >> at the level of convenient fictions, rather than at the physical >> level. > Isn't that a good thing, if a physicalist theory of computation > gives rise to a CTM that covers all of these "convenient fictions"? Does the resulting CTM cover these? Or are people just depending on wishful thinking? The problem I see, is with learning. The theories of learning that physicalists hold, mostly amount to discovering patterns in physical signals. But if the world we live in is based on convenient fictions, then the needed patterns don't exist in the physical world. Learning needs to be more about invention than about pattern discovery. > I just want to point something out, regarding rules, agents, > fictions, and the world. I don't know if a highway or a bank is > a thing, or what sort of thing it is. I'm not trying to know, > and perhaps I'm trying not to know. What it is, whether it is a thing, does not so much matter. The question is how could it ever arise if learning is just the discovery of patterns in physical signals? > The rules, or what may be rules, and the following, or what > may be following, that *I* am talking about, are all within a > computer program, and/or perhaps in a mind - computer and/or mind > I abbreviate as "agent". An agent is free to construct convenient > fictions. They (rules) are not being asserted as science or norms > or ontologically real outside of the agent. I don't have a problem with that. But the issue is in the "construct" part, vs. the pattern discovery that is often assumed to be the basis of learning. > LW's discussions about rule following are also of this sort, > though he wanders into mathematical proof theory, and a logicist > may mistake that for being the same thing. That is, LW talks > only about humans as agents, and assumes machines are in some > other category. And, just to be clear, I don't have a problem with saying that computers follow rules. I happen to think that was not what Wittgenstein was talking about in his arguments on rule following. Regards, Neil GOOGLE: http://groups.google.com/group/Wittrs FREELIST: //www.freelists.org/archive/wittrs/09-2009 TODAY: http://alturl.com/whcf 3 DAYS: http://alturl.com/d9vz 1 WEEK: http://alturl.com/yeza YAHOO: http://groups.yahoo.com/group/Wittrs/