dmo@turkey.philips.com (Dan Offutt) (06/30/89)
> ... The brain is not magic. If it can evolve, it can be >purposely designed. There can be no credible refutation of this logic. The proposition "if X can evolve via biological evolution, then X can be purposely designed" is generally false. The human brain or immune system might be copied or imitated, but neither human beings nor "fast" AI could possibly design such systems. The proposition is false for two reasons. First, purposeful design presupposes a design problem definition. We know very little about the design constraints for biological systems. The human brain has been shaped by gravity, the ratio of oxygen to nitrogen in the atmosphere, the availability of various raw materials, the capabilities of various predators, evolutionary pressures to economize on resource consumption, and innumerable other influences that we do not understand well enough or are not even aware of. Most of the constraints that shaped the human brain are unknowable in principle because most acted in the distant past, and information about those influences has been lost. For example, the structure of the human brain today may reflect, in some subtle way, the predatory capabilities of saber-tooth tigers, which are now extinct. Saber-tooth tigers are just one influence out of trillions. In short, systems that evolved by biological evolution cannot be designed by people or AI because it is impossible to specify more than a minute fraction of the design constraints. It may be possible to *copy* or *imitate* these evolved systems (as Mr. Drexler points out), but that is not the same thing as designing them. The second reason the proposition is false is that all human designers and all "fast" AI systems put together will never collectively possess the parallel "computational resources" that biological evolution has had to work with. Moreover, biological evolution has been searching for more-viable forms of life for billions of years. Presumably, those who argue that evolved systems could be designed do not have in mind billion-year-long design processes. The products of biological evolution conservatively cost Nature at least trillions of billions of processor years. (If there have been at least a few trillion living things -- each representing a "processor" -- on earth continuously for a billion years.) And this characterization ignores both the complexity of individual organisms and the nonliving environment around all of these creatures. Human designers and "fast" AI over the next 1000 years will not amount to anywhere near a trillion billion processor years. So the raw computational power available to design processes over the next 1000 years is an insignificant fraction of the raw computational power that biological evolutiuon comsumed to produce today's population of life on earth. Thus there isn't enough time available in the next 1000 years to design the human brain, even if the design constraints could all be specified. Dan Offutt dmo@philabs.philips.com
truesdel@prandtl.nas.nasa.gov (David Truesdell) (07/01/89)
dmo@turkey.philips.com (Dan Offutt) writes: [Mostly nonsense] >> ... The brain is not magic. If it can evolve, it can be >>purposely designed. There can be no credible refutation of this logic. >The proposition "if X can evolve via biological evolution, then X can >be purposely designed" is generally false. The human brain or immune >system might be copied or imitated, but neither human beings nor >"fast" AI could possibly design such systems. I fail to see a significant distinction between duplication/emulation, and "design". To duplicate, you need to know what's there, to emulate, you need to know how what's there works. >First, purposeful design presupposes a design problem definition. We >know very little about the design constraints for biological systems. >The human brain has been shaped by gravity, the ratio of oxygen to >nitrogen in the atmosphere, [etc., etc.] > ... > For example, the structure of >the human brain today may reflect, in some subtle way, the predatory >capabilities of saber-tooth tigers, which are now extinct. >Saber-tooth tigers are just one influence out of trillions. All of which is would be nice to know, BUT, are they really necessary for the design of a (possibly NON-biological) human equivalent "brain". One principle of "design" is to know what you DON'T need. How to deal with an extinct creature holds little value. Mr. Offutt obviously assumes that the human brain is the *ultimate* design and that a "proper" design specification would produce an *exact* replica. I'm afraid I don't share this opinion. The *current* design of the brain is simply a *local* optima. There's no rational reason to assume that the job could not be done better, with a little thought :-). Just remember, the same "trillions" of influences, also built lice. >The second reason the proposition is false is that all human designers >and all "fast" AI systems put together will never collectively possess >the parallel "computational resources" that biological evolution has >had to work with. Moreover, biological evolution has been searching >for more-viable forms of life for billions of years. The statements reflect a basic ignorance of what biological evolution is, and how it operates. A "billion-year-long design processes" requiring "trillions of billions of processor years" is only necessary if the "design process" is completely random, guided only by the viablity of each revision as determined by a changing environment. Mr. Offutt, are you a programmer? Do you develop your programs by blindly making changes, and testing the results? Would you care to estimate how long it would take to make a word-processor from the source of (lets say) a compiler using this method? >Dan Offutt >dmo@philabs.philips.com Just remember, the inforation required to build a human brain, it's "design" if you will, is tiny compared to the information capacity of the brain itself. There's no reason to assume that a brain couldn't improve on it's on design. -dave truesdell (truesdel@prandtl.nas.nasa.gov) "Character consists of what you do on the third and fourth tries." -dave truesdell (truesdel@prandtl.nas.nasa.gov) "Character consists of what you do on the third and fourth tries."
gordon@idca.tds.philips.nl (Gordon Booman) (07/04/89)
In article <Jun.29.22.43.09.1989.29501@athos.rutgers.edu> dmo@turkey.philips.com (Dan Offutt) writes: >... >The proposition "if X can evolve via biological evolution, then X can >be purposely designed" is false. >... If we're talking about designing an **exact** copy of the human brain, Dan is right. No one would ever design, on purpose, something so inefficient. :-) But seriously, folks. I think that Dan is misinterpreting the original suggestion: "that which evolved can be designed" as something more like: "a perfect copy of the brain can be designed", when it was clear that the intent was: "any function which evolved can be designed". I think there's a parallel with flight. Man tried for a long time to fly by gluing feathers to his arms and flapping. We've been able to fly now for close to a century; but we don't use feathers or flapping. "Man can't fly like a bird". Right. Who wants to? This is the danger of the Turing test. Intelligence doesn't have much to do with what people do (particulary some people :-). Intelligent machines, which we **already** have, don't act like people. Good thing, too. I certainly hope that when I download, I get some nice new algorithms to replace the mistakes of evolution. -- Gordon Booman SSP/V3 Philips TDS Apeldoorn, The Netherlands +31 55 432785 domain: gordon@idca.tds.philips.nl uucp: ...!mcvax!philapd!gordon
alan@oz.nm.paradyne.com (Alan Lovejoy) (07/07/89)
In article <Jun.29.22.43.09.1989.29501@athos.rutgers.edu> dmo@turkey.philips.com (Dan Offutt) writes: >I wrote: >> ... The brain is not magic. If it can evolve, it can be >>purposely designed. There can be no credible refutation of this logic. >The proposition "if X can evolve via biological evolution, then X can >be purposely designed" is generally false. The human brain or immune >system might be copied or imitated, but neither human beings nor >"fast" AI could possibly design such systems. The proposition is false >for two reasons. > >First, purposeful design presupposes a design problem definition. We >know very little about the design constraints for biological systems. >... >Most of the constraints that shaped the human brain are unknowable in >principle because most acted in the distant past, and information >about those influences has been lost. >...In short, systems that evolved by biological evolution cannot be designed by >people or AI because it is impossible to specify more than a minute >fraction of the design constraints. It may be possible to *copy* or >*imitate* these evolved systems (as Mr. Drexler points out), but that >is not the same thing as designing them. None of this is relevant. We don`t NEED TO KNOW these things. We only need to know and understand WHAT FUNCTIONS HUMAN BIOBRAINS PERFORM TODAY. If this were not true, then--by the same reasoning--most of our biological and medical knowledge would have been unobtainable. In fact, we have designed many devices, processes and substances whose functions mimic, replace, outdo and substitute for various biologically-evolved mechanisms and substances. And we did this in all cases without the knowlegde you claim is necessary. This constitutes disproof by massive counterexample. >The second reason the proposition is false is that all human designers >and all "fast" AI systems put together will never collectively possess >the parallel "computational resources" that biological evolution has >had to work with. This argument, although it is an interesting point which is not completely false, is not relevant to my original statement. My statement made no assumptions about the amount of time that might be required to design something--or about the necessary computational resources. The reason for this is that the idea which my statement is intended to refute is the following: "Artificial intelligence on the level of human capabilities is completely and forever impossible." In that context, arguments that it might require 20 billion years to design an artificial human brain already concede my point. Technicalities aside, it is appropriate to consider the implications and relevance of Mr. Offut's second point. After all, if a synthetic human brain will not be built for 20 billion years, then its relevance to us personally--and even to all present and future humanity--is in very serious question. As a sanity check, we should compare the length of time required for natural evolution to design intelligent systems with some quantifiable degree of intelligence with the length of time required for human scientific and engineering processes to achieve similar milestones. The evidence to date not only does not support Mr. Offut's thesis, it flatly and violently contradicts it. By his reasoning, we are wasting our time doing scientific research, since evolution will "compute" or discover this information so much faster than we can. These observations suggest not only that something is very wrong with Mr. Offut's argument, but even point to the source of the fallacy: "natural" evolotion is not the most efficient methodology of discovering truth. There are much better algorithms. In fact, natural evolotion is not some static process whose methods never change. Evolution also evolves. And we are living examples of that fact. Alan Lovejoy; alan@pdn; 813-530-2211; AT&T Paradyne: 8550 Ulmerton, Largo, FL. Disclaimer: I do not speak for AT&T Paradyne. They do not speak for me. ______________________________Down with Li Peng!________________________________ Motto: If nanomachines will be able to reconstruct you, YOU AREN'T DEAD YET.
landman@sun.com (Howard A. Landman) (07/07/89)
In article <Jun.30.22.25.34.1989.15323@athos.rutgers.edu> truesdel@prandtl.nas.nasa.gov (David Truesdell) writes: >I fail to see a significant distinction between duplication/emulation, and >"design". To duplicate, you need to know what's there, to emulate, you need >to know how what's there works. Anyone with a copier can duplicate Escher. Emulating Escher (creating a new artwork in his style) is *MUCH* harder. Designing (creating an artwork in your own style, and creating your own style) may be harder still. The information content of a copy, given the original, is very small. You can choose to make a copy by pressing one button. The information content of a *de* *novo* design is very large. There is no one button to push. You must make many decisions. Howard A. Landman landman@sun.com