--- In Wittrs@xxxxxxxxxxxxxxx, "SWM" <SWMirsky@...> wrote: > What makes an action "about something"? Perhaps it's hard to give a general characterization, but it often seems evident. A cat playing with yarn, or a cat playing with a mouse, or a dog with a bone. It gets harder with insects, etc. > On my view, "aboutness" is when we can relate some symbol or > indicator to something else, when we can see meaning. To tie it to symbols might be a bit too restrictive. > Perhaps the issue between us hinges, to some extent at least, on > your focusing on the nature of computational programs themselves > (nothing conscious about them!) vs. my focusing on the nature of > computational systems (i.e., many different programs running many > different processes to accomplish many different functions in a > kind of orchestral arrangement)? That could be. When you get to the level of asking "how do I program that" you begin to see some difficulties that were not so obious before. Hubert Dreyfus says "I was particularly struck by the fact that, among other troubles, researchers were running up against the problem of representing significance and relevance <http://leidlmair.at/doc/WhyHeideggerianAIFailed.pdf> ". While those are not the words I would have chosen, it's a pretty good assessment of the kind of problem I ran into. > Yes, the human eye is constantly moving about and the picture it > captures consists of many distinct imprints or partial images which > the brain somehow sees as a whole, a complete pattern. (Hawkins > uses this model quite a bit in his book On Intelligence.) I am quite skeptical of that view. It's a top down designer view of how to do vision, rather than a bottom up evolutionary view. I think it more likely that it is similar to a single cell scanning back and forth, and looking for sharp signal transitions to find a boundary. However, it is being done a billion times in parallel by the different retinal cells. > What is a visual feature but a pattern within a larger pattern, > a picture within a larger picture? No, I disagree with that. The features are marked by boundaries. And the thing about boundaries, if you are using a scanning method, is that you can locate boundaries with higher resolution than you can locate other things. In any case, the visual part is guesswork. However, the standard AI approach is too dominated by top down thinking. From an evolutionary perspective, you need to find a use for a single retinal cell, and then an evolutionary benefit for proliferating that into many retinal cells. > As I noted, Hawkins suggests the brain develops and retains templates > and that when a remembered image is called up we get more of an > adumbration which we then use to plug in details, presumably by > recognizing subsections and using this to call up detailed images > within the larger one. The templates part is okay, if intended as a recognizer. J.J. Gibson (the "direct perception guy") would have used the term "transducer" rather than "template." However, I am doubtful about the "call up detailed images" part. I doubt that there are any stored images to call up. Sure, we can have imagery in our thought, but it doesn't seem to be a called up image and is more likely a reconstruction. > Isn't that true of us too? We only see in the world what we are > built to see. If we had been built differently the world might seem > entirely different to us, no? Do you really think that we were built to see (in the sense of "comprehend") jet aircraft, HIV virus, electron microscopes? Regards, Neil