loren@tristan.llnl.gov (Loren Petrich) (03/12/91)
In the current issue of "Discover" magazine, there is an article about attempting to construct robots with insect-level intelligence. It featured the work of Rodney Brooks, who has had great success with such robots. He and his colleagues have built several kinds of such robots, including some six-legged walkers and some wheeled ones. They are programmed with simple heuristics which compete with each other to make responses. For instance, short-range responses like "back off", when triggered by the prospect of a collision, can override long-range goals like "track prey". Is this type of setup an example of "Fuzzy Logic"? The article reports that Pattie Maes has derived a program that can help one of Brooks's six-legged robots learn to walk. At first, the robot flails its legs around uselessly. But it then learns what leg motions both keep the robot up and make it go forward. Before too long, the robot has learned an insect gait -- to move the two outer legs on each side in sync with the middle leg of the other. The article did not describe the algorithm, but it was probably some kind of Neural Net algorithm. Brooks got the idea from considering that insects are awfully dumb animals, but ones which are nevertheless able to walk and fly and see. He had the idea of building up from simple reflexes. He decided to leave behind the traditional AI model of symbolic reasoning about the world, because it had yielded only very limited success in controlling robots. For instance, such models have trouble with unfamiliar circumstances, circumstances that Brooks's robots have no trouble with. And they require a LOT more processing power than Brooks's robots, which only need simple microprocessors. Not surprisingly, he has looked down on symbolic-modeling AI, which has been the mainstream of the field. And he has gotten a lot of flak from those who have worked in the symbolic-AI field, who have accused him of scaling down his goals from human-level intelligence to insect-level intelligence. My response to that would be that insect-level intelligence is better than level-zero intelligence. There are a lot of analogies with the Neural Nets field -- a system not based on traditional, symbol-manipulating, AI that is much simpler than it and easily outperforms it, but which seems to have much less in the way of ultimate potential. Is that fair? And does anyone else know of any details of Brooks's work? $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$ Loren Petrich, the Master Blaster: loren@sunlight.llnl.gov Since this nodename is not widely known, you may have to try: loren%sunlight.llnl.gov@star.stanford.edu
demers@odin.ucsd.edu (David E Demers) (03/12/91)
In article <92995@lll-winken.LLNL.GOV> loren@tristan.llnl.gov (Loren Petrich) writes: > In the current issue of "Discover" magazine, there is an >article about attempting to construct robots with insect-level >intelligence. It featured the work of Rodney Brooks, who has had great >success with such robots. [...] > Is this type of setup an example of "Fuzzy Logic"? No, but there may be a way to express subsumption architectures with Fuzzy logic. [...] > There are a lot of analogies with the Neural Nets field -- a >system not based on traditional, symbol-manipulating, AI that is much >simpler than it and easily outperforms it, but which seems to have >much less in the way of ultimate potential. > Is that fair? Not really - they are different tools. Although work is being done on higher-level cognitive tasks with NNs (addition/subtraction, for example), symbolic systems work better on many tasks. They are really two different methods, and comparing them is probably not useful except with respect to a particular problem. AI should be inclusive; let's use the right tool for a particular task. > And does anyone else know of any details of Brooks's work? He has published quite a bit. You may want to look at the recent collection of papers "AI at MIT: Expanding Frontiers" edited by Patrick Winston (MIT Press, 1990) ISBN: 0-262-23150-6 The first two chapters of vol. 2 are by Brooks; one discusses the subsumption architecture in general terms, the second shows the application to six-legged locomotion. -- Dave DeMers demers@cs.ucsd.edu Computer Science & Engineering C-014 demers%cs@ucsd.bitnet UC San Diego ...!ucsd!cs!demers La Jolla, CA 92093-0114 (619) 534-8187,-0688 ddemers@UCSD
barryf@aix01.aix.rpi.edu (Barry B. Floyd) (03/12/91)
Is Brook's on the net (USENET)? barry -- +--------------------------------------------------------------------+ | Barry B. Floyd \\\ barry_floyd@mts.rpi.edu | | Manager Information Systems - HR \\\ usere9w9@rpitsmts | +-Rensselaer Polytechnic Institute--------------------troy, ny 12180-+
karln@uunet.uu.net (03/13/91)
In article <92995@lll-winken.LLNL.GOV> loren@tristan.llnl.gov (Loren Petrich) writes: >flak from those who have worked in the symbolic-AI field, who have >accused him of scaling down his goals from human-level intelligence to >insect-level intelligence. > I was at the pub the other night, after which I am not sure there is a difference[;-}> <-A sly sort of smile from a square head with beard. Very interesting subject. Although I have heard repeated claims that neural nets are supposed to learn, as in the learning to walk example, this is the first I personally have heard of such success. It seemed to me that Neural Nets were mostly only good for Statistical output, in other words, given a number in, the Neural Nets Statistics 'Database' shows that this is probably the number out. This is from seeing many Cartesian to Polor converters and such. I suppose what is involved in a neural net learning to walk is a "Feels good" circut or subroutine for feedback. If the bug is not moving, the neural net gets negative feedback. If the bug is moving, then the net gets positive feedback. The level or severity of feedback would depend on haw well the bug is moving as told. If the "Barin" wants to go foward, but the neural net drives the bug slightly sideways, then the feedback is bad. The feedback gets better as the bug goes straighter, and worse as the bug goes more sideways. I wonder what other type of feedback is required to get the business working properly. Do all the legs current positions need to be feedback? How about the current leg positioning command? This seems like something to keep track of around here. Karl Nicholas karln!karln@uunet.uu.net
mullen@evax.arl.utexas.edu (Dan Mullen) (03/13/91)
I find it rather ironic that after almost a half century of research in "artificial intelligence" we have humbly accepted our limitations and assumed the task of modeling insects. I remember reading once that a mosquito with its paltry 10,000 neurons is infinitly more intelligent than our fastest super-computer. I'm not argueing on either side of that. My point is simply that modeling insects is a good place to start. Maybe soon there will be the small rodent robot and then the farm animal robot. Maybe an entire robot farm, and then ..... d.m.
pat@cs.strath.ac.uk (Pat Prosser) (03/13/91)
In article <1991Mar12.201920.18088@evax.arl.utexas.edu> mullen@evax.arl.utexas.edu (Dan Mullen) writes: >I find it rather ironic that after almost a half century of research in >"artificial intelligence" we have humbly accepted our limitations and >assumed the task of modeling insects. I remember reading once that >a mosquito with its paltry 10,000 neurons is infinitly more intelligent >than our fastest super-computer. I'm not argueing on either side of that. >My point is simply that modeling insects is a good place to start. Maybe >soon there will be the small rodent robot and then the farm animal robot. >Maybe an entire robot farm, and then ..... d.m. and then .... Do Robots Dream Electric Sheep?
zane@ddsw1.MCS.COM (Sameer Parekh) (03/17/91)
In article <1991Mar12.201920.18088@evax.arl.utexas.edu> mullen@evax.arl.utexas.edu (Dan Mullen) writes: >I find it rather ironic that after almost a half century of research in >"artificial intelligence" we have humbly accepted our limitations and >assumed the task of modeling insects. I remember reading once that >a mosquito with its paltry 10,000 neurons is infinitly more intelligent >than our fastest super-computer. I'm not argueing on either side of that. >My point is simply that modeling insects is a good place to start. Maybe >soon there will be the small rodent robot and then the farm animal robot. >Maybe an entire robot farm, and then ..... d.m. I agree that trying to create low intelligence is a good start, yet there is no purpose in having 80 artificial cows. They can't do anything better than us, so why use them? (Or were you joking?) -- zane@ddsw1.MCS.COM
johnc@ms.uky.edu (John Coppinger) (03/17/91)
In article <1991Mar12.201920.18088@evax.arl.utexas.edu>, mullen@evax.arl.utexas.edu (Dan Mullen) writes: > I find it rather ironic that after almost a half century of research in > "artificial intelligence" we have humbly accepted our limitations and > assumed the task of modeling insects. When you try to run down the path before first learning to crawl, you'll soon find yourself sprawled out in the mud only a few yards from where you began. I believe this happened to AI. Perhaps it's not an acceptance of limitations. Rather, it's the realization that crawling is the important first step towards running. > I remember reading once that > a mosquito with its paltry 10,000 neurons is infinitly more intelligent > than our fastest super-computer. Ah, the power of parallel... -- -- John Coppinger "You'll find that your left cuff link -- -- University of Kentucky will be communicating with your right -- -- johnc@s.ms.uky.edu cuff link via satellite" -- -- johnc@graphlab.cc.uky.edu [NeXT] -- Nicholas Negroponte --
karln@uunet.uu.net (03/20/91)
In article <1991Mar17.063506.28939@ddsw1.MCS.COM> zane@ddsw1.MCS.COM (Sameer Parekh) writes: > I agree that trying to create low intelligence is a good start, >yet there is no purpose in having 80 artificial cows. They can't >do anything better than us, so why use them? (Or were you joking?) > I do not think that is true. An artificial cow could be 'programed' to eat more at the right time to produce more meat come market time. An artificial cow might be able to have two calves. The error of that statement is glaring. Anyway we should ask a farmer about that one ... "Mr Farmer, what would you do with a programable cow?" "Well, could I run my Nintendo games on it?!?!" .... Also, I do not think that many people would argue that the first computer, made from spinning wheels and brass bars, couldn't do math better than us .... you gotta start somewhere. I keep waiting for someone to wonder why there is a problem with two types of AI. I seem to get the impression that the insect neural net application is like the motor control center of the brain. Why ask a symbol processor to move legs when it could just 'think' I wanna go there and let the neural net solve the specifics of motivation, which it is doing quite well. Karl Nicholas karln!karln@uunet.uu.net
usenet@cs.utk.edu (USENET News Poster) (03/21/91)
The suggestion for a "rodent robot" brings to mind s pair of articles in the magazine _INK_, edited, I believe by Steve Ciarcia of Circuit Cellar fame (or his brother, I don't remember which). Anyway, these two articles in the 1990 (year) volume describe how to build "MityMouse II", a rodent robot. I've taken these magazines home, so I don't have the exact citations. From: sfp@mars.ornl.gov (Phil Spelt) Path: mars!sfp BTW, there are numerous other approaches to robot behavior. We here at ORNL hjave produced two autonomous robots which are capable of doing a variety of "intelligent" things -- see my article in _IEEE_Expert_, Winter, 1989. Phil Spelt, Cognitive Systems & Human Factors Group sfp@epm.ornl.gov ============================================================================ Any opinions expressed or implied are my own, IF I choose to own up to them. ---------------------------------------------------------------------------- MIND. A mysterious form of matter secreted by the brain. Its chief activity consists in the endeavor to asscertain its own nature, the futility of the attempt being due to the fact that it has nothing but itself to know itself with. -- Ambrose Bierce ============================================================================ Phil Spelt, Cognitive Systems & Human Factors Group sfp@epm.ornl.gov ============================================================================ Any opinions expressed or implied are my own, IF I choose to own up to them. ============================================================================
carroll@ssc-vax (Jeff Carroll) (03/22/91)
In article <1991Mar21.123604.284@ulrik.uio.no> espen@math.uio.no (Espen J. Vestre) writes: > >In these Patriot times, don't you think there'd be a market even for >things simpler than artificial cows? Sharks .... or swarms of >insects.... yech. > I can see it now. Artificial locusts... frogs... gnats... flies... Maybe we could start out with artificial hailstones. Yeah, that's it... We could call them "brilliant pebbles". -- Jeff Carroll carroll@ssc-vax.boeing.com
zane@ddsw1.MCS.COM (Sameer Parekh) (03/26/91)
In article <1991Mar19.225113.15536@uunet.uu.net> karln@karln.UUCP () writes: >In article <1991Mar17.063506.28939@ddsw1.MCS.COM> zane@ddsw1.MCS.COM (Sameer Parekh) writes: >> I agree that trying to create low intelligence is a good start, >>yet there is no purpose in having 80 artificial cows. They can't >>do anything better than us, so why use them? (Or were you joking?) >> > > I do not think that is true. An artificial cow could be 'programed' >to eat more at the right time to produce more meat come market time. An artificial >cow might be able to have two calves. The error of that statement is glaring. The idea of artificial cow, I had thought, was a robotic being with the intelligence of a cow. Making an artificial cow that is EDIBLE, in my belief, would be tougher than designing a human-level intelligence. -- The Ravings of the Insane Maniac Sameer Parekh -- zane@ddsw1.MCS.COM
karln@uunet.uu.net (03/28/91)
In article <1991Mar25.173925.21895@ddsw1.MCS.COM> zane@ddsw1.MCS.COM (Sameer Parekh) writes: >In article <1991Mar19.225113.15536@uunet.uu.net> karln@karln.UUCP () writes: >>In article <1991Mar17.063506.28939@ddsw1.MCS.COM> zane@ddsw1.MCS.COM (Sameer Parekh) writes: >> >> I do not think that is true. An artificial cow could be 'programed' >>to eat more at the right time to produce more meat come market time. An artificial >>cow might be able to have two calves. > > The idea of artificial cow, I had thought, was a robotic being >with the intelligence of a cow. Making an artificial cow that >is EDIBLE, in my belief, would be tougher than designing a human-level >intelligence. > True. I was thinking that an artificial cow would be a cow with a brain transplant. Perhaps in the future putting the desired `behavior patterns` into a cow via `brain transplant' would be better than trying to breed this `behavior pattern` into cows. If this ever were the case the small Neural Net to handle walking and eating could be very usefull. Really just some way out there sort of thoughts. My appologies for seeming nasty. BTW, I have recieved the program from Pat and Greg Williams that models the insect behavior of Dr. Beers insects. Sure enough the insects walk around and are show a certain ummunity to neural damage or removed without failing completely. However, the bugs come with the neural nets already programmed. The bugs come knowing how to walk. I was sort of hopeing to see these bugs flounder around a bit then eventually learn to walk. Does anybody have any suggestions on how I should set about setting this up. Is this set of programs capable of it? any comments. anybody else have this program? karl nicholas karln!karln@uunet.uu.net