Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!decvax!decwrl!ucbvax!JPL-VLSI.ARPA!larry From: larry@JPL-VLSI.ARPA.UUCP Newsgroups: mod.ai Subject: "Proper" Study of Science, Conservation of Info Message-ID: <8608010557.AA11269@ucbvax.Berkeley.EDU> Date: Thu, 24-Jul-86 06:30:03 EDT Article-I.D.: ucbvax.8608010557.AA11269 Posted: Thu Jul 24 06:30:03 1986 Date-Received: Fri, 1-Aug-86 08:44:37 EDT Sender: daemon@ucbvax.BERKELEY.EDU Organization: The ARPA Internet Lines: 81 Approved: ailist@sri-ai.arpa I have to start with materialism. What we mean today when we say the word may have a common core with its use in previous centuries, but the details are vastly different. Today we recognize not only wind and wave, steam and steel as physical realities, but also quanta and field effects (and virtual particles!)--subjects that pre-modern physicists and engineers would consider downright mystical. And that would have been exactly true--in their time. But we can precisely define these things now, quantify them, experiment with and measure them. An even more radical difference is that information--pattern, form--is now a part of physics, "a metric as important as time, space, charge, etc." The ability to quantify and measure pattern and shape has profound implica- tions for the study of formerly mystical topics such as intelligence. It means we can develop conservation laws for information, without which you can't construct an essential ingredient of mathematics, equations. I'm not implying I know what they are in any detail; people with other qualifica- tions than mine must provide that. But the shape of the research seems to be clear; cybernetics and information theory provide the basis. For instance, there are several links between information and energy. Higher frequency radiation has more bits per unit time. Mutation is the result of external energy pushing genes beyond the ability of their binding energies to maintain a stable structure. The impressing of information on media (diskettes, molecules, brains) requires energy which can be measured. Organization of information in structures (indexed or random files, percepts, concepts) has time/energy trade-offs for different kinds of accesses. In a way, the information content of an entity is more important than its material content. A decade from now it's likely that none of our bodies will contain EVEN A SINGLE ATOM now in them. Even bones are fluid in biological organisms; only when we die does matter cease to flow into and out of us. We are NOT matter, or even energy, in the Antique sense. We are patterns, standing waves in four (or more) dimensions. Maintaining these patterns within safe parameters, or learning new safe parameters, requires that our very molecules input data, store it, process it--often in a recursive or self-referential or time-dependent fashion--and act. (RNA is an excellent model for an advanced computer, for instance.) And we can be thought as a number of layers each with its unique informa- tion needs: cells, tissue, organs, organisms, tribes. One feature common to all intelligences, however rudimentary, is the ability to create and manipulate analogs of the environment and of themselves. Simulations are much cheaper and safer than experiments. This also gives a clue as to how will impresses itself on the universe despite its immaterial nature--because it isn't truly immaterial. Patterns are no more independent of their matter/energy base than matter can exist without pattern. (That is, the pattern of binding is what makes the difference between an atom and a burst of radiant energy.) Because intelligence is a pattern of energy it can affect matter and through triggering have effects enormously greater than the triggering stimulus. A whim and a whistle can destroy a city--with an avalanche. The point of all this is that life and intelligence are no longer supernatural--beyond the reach of formalism and experiment. What is still a mystery to me is consciousness, but the understanding doesn't seem beyond practical realization. It seems reasonable that con- sciousness arises as a result of time-binding, recursion, and self- reference. Perhaps multiple layers of vulnerability and adaptability are important, too. (Our current robots and computers don't have any of these and are thus poor candidates for models of intelligent mechanisms, much less conscious ones. Thus I'd agree with one recent critic of some AI research.) I can't agree that consciousness is an improper subject for scientific study. Our inability to observe it directly (in a public as opposed to subjective way) is shared by many other scientific fields. In fact the most crucial subjects in the "hard" sciences must be studied indirectly: radia- tion, atoms, viruses, etc. The difficulty of defining terms shouldn't be a deterrent either. All developing research shares the same problem as the underlying ideas change and solidify. Some people object on emotional grounds. Many of them only succeed in revealing their own limitations, not those of the rest of us. They are too emotionally stunted to have the strength of humility; they must somehow be above nature, superior. And too intellectually crippled to see the magic and mystery in star-shine and bird flight, in ogive curve and infinitesimals and the delicious simplicity of an algorithm. Larry @ jpl-vlsi.arpa