Robot Maker Mark Tilden: All Life is Analog
simpl3x points to this New York Times article on master robotsmith Mark Tilden, writing: "It is interesting what makes a good toy." My favorite line is Tilden saying "I want to sell millions of toys, but what I really hope is that a bunch of kids who open them up use the motors and things to build something else ... They are my colleagues of the future."
For those who don't wish to register to NY Times:
Toyland Is Tough, Even for Robots
By BARNABY J. FEDER
MARK TILDEN recalls being a lonely child, repeatedly uprooted by his family's moves around Canada. He took comfort in his gift for constructing toys, especially mobile toys.
"I was born a compulsive builder," Mr. Tilden said. "I made my first robot out of sticks and rubber bands when I was 3."
Mr. Tilden, now 41 and a resident of Los Alamos, N.M., figures he has made thousands more since then. His designs have included machines to explore other planets, mine-clearing devices, toilet bowl cleaners and, more recently, a line of toys called B.I.O.-Bugs. The footlong creatures, which vaguely resemble roaches despite having just four legs, were a hit at the 2001 Toy Fair in New York and were brought to market last fall by Hasbro (news/quote).
Mr. Tilden's specialty has been designing robots with little or no brainpower. Instead, they are built around networks of simple sensors, switches and mechanical systems that respond to analog signals like lightwaves, heat or sounds without any need to convert them into a digital code of ones and zeros for analysis by a microprocessor.
Colleagues marvel at the dexterity and speed with which Mr. Tilden builds devices, noting that such finesse seems unexpected in a man so large and rotund that he jokingly describes himself as "big enough to create my own ozone layer." Then there is his ingenuity. Many a Tilden robot consists largely of components harvested from cameras, videocassette recorders and other devices retrieved from junk bins.
"Tilden is unique in his ability to intuit and hack analog circuits," said Rodney Brooks, head of the Massachusetts Institute of Technology's Artificial Intelligence Laboratory. "You just cannot find anyone else with his virtuoso skills in that area."
But if Mr. Tilden has become widely known, even admired, among robotics experts, his views have not won him a large following. Nor has his recent plunge into the toy business played out as he hoped. Simpler is not always better for toy makers looking for unique products, he learned, and unexpected events, like domestic terrorism, can change perceptions of even a toy.
Mr. Tilden has been arguing with little success for well over a decade that progress in robotics would be much more rapid if researchers concentrated on designing relatively dumb robots rather than devices stuffed with increasingly powerful programmable electronic brains. The trick, in Mr. Tilden's view, is to equip simple-minded but physically robust robots with mechanical variations on animal nervous systems.
Nervous networks do not organize and process information digitally as computers do. Nonetheless, he points out, every second of life on earth is filled with millions of types of dim-witted creatures using nervous systems to respond instantly to environmental challenges that stump the powerful digital brains of today's computer-driven robots.
"All life is analog," Mr. Tilden said.
Many other robotics experts are also interested in nervous networks. And many are just as convinced as Mr. Tilden of the value of designing robots from simple building blocks. But most believe that without digital brainpower -- lots of it -- machines will have little potential to learn from experience and be far too limited in their ability to interact usefully with humans or other machines.
The robotic design wars that have preoccupied Mr. Tilden since the late 1980's have largely been waged in university laboratories, obscure journals and government-financed research projects. Mr. Tilden's main livelihood since 1993, for instance, has come from research at the federal government's Los Alamos National Laboratory.
In recent years, though, the toy industry has emerged as a new playground for the robotics theorists. In this sector, as in the others, the advocates of programmable robotics clearly have the lead and the upper hand. Products like the Sony (news/quote) Aibo (which cost $2,500 when it was introduced in 1999), Furby and Lego Mindstorms have been huge hits. Robotics and virtual pets accounted for only $160 million of the $2.3 billion toy industry's revenues in 2000, but Poochi and Tekno, both robotic toys, were individual best sellers.
The novelty of Mr. Tilden's approach and some of his inventions caught the eye of executives at WowWee just over a year ago, shortly before the company was acquired by Hasbro, the second-largest toy company after Mattel. Mr. Tilden said he was thrilled by the invitation to become a consultant.
"You build something for NASA and you only build two of them," Mr. Tilden said. "You build for the military and they might want 50. But here it could be millions."
Mr. Tilden's fondest dreams were battered a bit by his first year in the toy business, though. B.I.O.-Bugs, priced at $39.95, reached toy stores last September. There were four bugs in the line, each with slightly different behavioral tendencies. The red Predator was the most aggressive, the blue Stomper the noisiest, the green Destroyer slightly more suited to moving in rough terrain and the yellow Acceleraider the speediest. The battery-driven bugs operate on their own or under remote control.
Mr. Tilden had originally hoped for a broader line including some bugs intended to appeal to girls rather than the 4- to 9- year-old boys Hasbro had in mind. Mr. Tilden also wanted to make B.I.O.-Bugs easy to dissect and alter, a starkly different attitude from that of Sony, which has threatened to sue customers who publish information about how to alter its Aibo dogs or the software that runs them.
"I want to sell millions of toys, but what I really hope is that a bunch of kids who open them up use the motors and things to build something else," Mr. Tilden said. "They are my colleagues of the future."
Hasbro had a more commercial and conservative perspective than Mr. Tilden's, of course. Before mass production began last year in Hong Kong, he said, Hasbro told him that a chunk of the "neural network" engineering needed to be converted into digital functions executed by a microprocessor so that B.I.O.-Bugs would be harder for competitors to reverse-engineer and duplicate.
"It ended up with about 80 percent of what I wanted," Mr. Tilden said.
Hasbro ended up feeling similarly unfulfilled. B.I.O.-Bugs sold well -- they were, for example, the best-selling robotic toy at F.A.O. Schwarz during the Christmas season, said Steven Benoff, the toy retailer's chief buyer for electronics, action figures, video games and vehicles. But overall sales added up to "a double or a triple" rather than a home run, according to Loren T. Taylor, the Hasbro executive who oversees WowWee. In the toy industry, only a smash hit guarantees a line's survival beyond its first year.
Mr. Tilden and some independent experts are convinced that B.I.O.-Bugs would have done much better had Hasbro not been forced to abandon a portion of its advertising campaign in October. The television ads, which were geared primarily toward children and fans of science fiction shows like "Star Trek: The Next Generation," began attracting angry letters from viewers who said the landscape that the bugs were crawling over looked like the ruins of the World Trade Center.
Then came the anthrax attacks. "We had the worst name you could come up with for selling toys during an anthrax scare," Mr. Tilden said.
Whatever the reasons, Hasbro decided that expanding the line this year was too risky. B.I.O.-Bugs shipped last year will remain on the shelves in this country, and B.I.O.-Bugs will be introduced in overseas markets that did not get them last year. But Mr. Tilden was told late last year to put aside plans for new B.I.O.-Bugs and focus instead on enhancing dragons, hovercraft and several other toys that WowWee introduced last week at the Toy Fair.
"They would have been like Ferraris compared to Model T's," Mr. Tilden said, sighing over the B.I.O.-Bug enhancements he was told to shelve.
If the B.I.O.-Bug experience has done less than Mr. Tilden had hoped to highlight the commercial value of his robotics concepts, it certainly has not shaken his faith in them. He still believes that large numbers of such simple devices are more likely to be able to execute many tasks without human supervision than the brainy robots most researchers have been trying to build. As evidence, he often points to the tiny, slow-moving devices he has built to clean the floors and windows in his condominium apartment.
Meanwhile, he is still having fun working for Hasbro and is constantly on the prowl for chances to demonstrate his concepts, both inside the toy business and beyond. On the whole, he said, the experience with B.I.O.-Bugs has been good. That has not always been the case with his inventions, he said.
Mr. Tilden recalled a woman who fled their first date after being approached on his couch by a television remote control to which he had grafted a snakelike robotic tail. "I designed it to move when someone sat down because I kept losing the remote in the cushions," he said.
But life -- robotic as well as human -- goes on. Some of the same technology is embedded in a fantasy snake that Mr. Tilden recently designed for Hasbro.
Thanks,
--
Matt
To avoid the annoying "registration required", use the username : slashcode0 and password slashcode0 Have a nice day
Of course, using accountname "Password" with password "Password" does the trick.
Writers imply. Readers infer.
I think he was talking about the stimulus-response behavior that many creatures exhibit. Even my cat does this quite often--"I'm hungry; get food.", "I'm tired; go to sleep.", "Big, scary dog; run away!" This sort of behavior can be easily modeled using analog parts.
My cat would probably be a little tough to build without a microprocessor, but a beatle would be (relatively) easy.
Of course, Robot Wars was adapted from robot combat sports that had already been going on in the US without TV broatcasts, but everything comes from something.
For example, there is no such thing as a "strong" vs. a "weak" pulse in a nervous system - it's an on-or-off thing, a 1 or a zero. A "stronger" message is sent by firing along the nerve more frequently.
That's an analog signal that happens to be encoded using time. Put it his way: if there were discrete levels that the pulse frequency could run at, it would be digital (digit-al). Since there isn't, and it's a continuous function, it's analog.
Sometimes it's best to just let stupid people be stupid.
Ummmm yeah. Sort of....
The important thing is not what a nerve fiber is doing, in the case of the neuron the action potential is quite stereotyped in the axon, but what the effect on the _target_ is. Again for brain (what I know best) a transmitter release event is an analog event as far as the post-synaptic cell is concerned. The fundamental event from the viewpoint of the post-synaptic cell is rate of ion flow through a gated channel. You can't get much more analog than that.
For the pre-synaptic cell the stereotyped AP is only good until it hits the terminal region. The calcium entry is an analog event, vesicle docking and transmitter release is pretty much a binary event, but the content of the vesicle and transmitter/receptor kinetics are analog all the way.
--everytime you learn something a piece of your brain is replaced by something that someone else said
Solarbotics!
:]
couldn't really have an article about Tilden without mentioning another REALLY DAMN COOL provider of nifty little robotics, that are based on the same concept (B.E.A.M) that Mark Tilden is famous for.
not for the lazy, though. most products are in kit form, which means you have to build it yourself. On the plus side, you get to build it yourself!
Of course, using accountname "Password" with password "Password" does the trick.
It did not work for me.
Nerves are not as straight forward as you would like to think. Different kinds of pain have different kinds of rates of firing on a neuron, and they aren't necessarily linear. For example, you would report feeling pain from a pin at 1/5th the firing rate compared to your finger being squeezed (I could have gotten this backwards, but the idea is valid). There is more going on than an 0 to 5 level here.
The brain itself is even more bizarre, it is inherently "noisy" and that "noise" appears to be necessary, as the brain SEEMS to be chaotic (in the mathematical sense) with competing subsystems (basins of attraction?) that keep the others on their toes. It becomes more ordered to pass information. This strikes a balance of being responsive (always in motion and adaptable), and being stable (well, for some of us at least). ; )
I haven't followed this for awhile (you could find info in any Neural Net group that is interested in mimicking biology probably), but for an interesting (older) article on this hypothesis:
Freeman W (1991) The Physiology of Perception, Scientific American, SA Inc, 264, 2, 34-41
(sorry, I don't know if it is online)
All this makes the idea of "analog computing" interesting if you ask me.
chaos:
sensitive dependence on initial conditions
periodic points are dense
topologically transitive
... Mandlebrot Sets, Julia Sets, Strange Attractors, oh my!