BY STEPHEN STRAUSS Science Reporter, The Toronto Globe and Mail
"Sensitive Robots", 17 may 1997
Is it an ant? A cockroach? Or Simply a Squiggle?
ALTHOUGH no one is quite to ready to assume the mantle of God, the line between the purely biological and the purely mechanical has begun to waver. Suddenly, the questions, "Is it alive?" and "How alive is it?" are becoming scientifically respectable.
At-least among robot experts. Two separate and quite distinct developments are fueling intense speculation about whether the day is nearing when humans will be able to make robots that are truly alive.
On the one hand are attempts to construct computer models of life processes. Baptised "artificial life" by their proponents, they consist of simulations of existing species which are allowed to "evolve" in a computer milieu, according to survival principles and reproductive capacities embedded in the software. One such program involved a two-dimensional representation of ants. Following mathematical sequences no human could recreate, a race of better-adapted ants was seen to take over the simulated world. With a kind of evolution taking place before their eyes, researchers at several institutions-who call themselves A-Lifers-began to ask: Is this computer program in any sense alive?
Critics say no. They say that where survival is concerned, there is a tremendous difference between the highly structured and artificial universe of a computer program and the physical chaos of Earth.
"They may be systematically interpretable as creatures in a world", says Princeton University psychobiologist Stevan Harnad. "But all they really are are squiggles and squoggles on a computer screen."
To overcome the obvious artificial quality of the simulations, he and others have pointed out the need for something we call "alive" to interact with the physical world. That leads to the second development. Robot engineers are turning to insect or other animal models to make existing robots more intelligent.
"Even the simplest animals are better at changing their behaviour to cope with the real world than the most sophisticated robot today", says Randall Beer, a computer scientist at Case Western Reserve University in Cleveland.
To many researchers, this interest in animal models has a dear evolutionary rationale.
"Nature never evolved a brain and then built the first body around it That is nonsense" says Universe of Waterloo engineer Mark Tillen.
"What Mother Nature must have done is first evolve the spine" Mr. Tillen says he is less interested in computer simulations of brains than in actual workings of the primitive nervous systems he builds into the insect robots he makes out of cast-off components.
When life is defined not as manlike or dog-like, but cockroach-like or ant- like, living machines suddenly seem closer. "We are abandoning the normal humanocentric ways of building robots," Mr. Tilden says. "We are not trying to build Commander Datas [the humanoid robot of television's Star Trek, The Next Generation]. We are trying to evolve whole new life forms."
These developments have been encouraging futuristic speculation. Last year Omni Magazine proposed no fewer than eight possible varieties of animal-bot. But the growing interaction between biology and mechanics has also produced concrete advances-more than anyone dared hope a decade ago when work in this area first began.
It has spawned a new field called computational neuroethology, in which engineers and computer scientists try to create programs that mimic animal behaviours.
Robert Full, a comparative Physiologist at the University of California at Berkeley, has been advising roboticists at the Massachusetts Institute of Technology in Boston and at Case Western about how insects move around.
Traditional robot design, he says, emphasized stability while in motion, as exhibited by R2D2, the rolling garbage can in the Star Wars movies. This precluded rapid acceleration and deceleration, as well as leg positions less stable than a threelegged stool.
However, Dr. Full's studies showed that the walking and running gaits of many living creatures seemed more like pogo sticks and pendulums than slow-moving mechanical stools. "At any given moment they would probably fall over if you stopped them", he says.
He has videotaped cockroaches zooming along on only two legs after reaching speeds that propel them 50 body lengths in one second.
Engineers have used this sort of information to make robots more lifelike. MIT researcher Mark Raibert has applied some of Dr. Full's principles of animal movement and produced running, hopping robots - one of which can do flips while on a treadmill. So what we have shown is that there are general concepts which are not restricted to animals," Dr. Full says.
Prof. Beer and his colleagues produced a computer simulation or a cockroach's walking abilities based on what neurobiologists knew about the neural connections that produce the scurrying behaviour of real cockroaches. When given a command to move, the simulated cockroach showed a variety of gaits which real insects exhibit-even though they hadn't been programmed in.
Similarly, when the scientists simulated lesions in their computerised nervous system by cutting neural connections, they were able to produce a version of that well-known phenomenon: the cockroach that continues to stumble around after its head is cut off.
Their success on the computer screen led them to build a six-legged, one- kilogram mechanical cockroach in which the nerve circuitry which made the computer simulation move was electronically reproduced. After a bit of tinkering the robot cockroach also scurried about like a real one.
Insights have now begun to flow back from engineering to biology. Engineers recently told Dr. Full their simulation work indicated that a pan of the cockroach's leg must play an important role in preventing slippage. "When we look at it their way we find it is important", he says.
In the future, researchers hope to connect the robot's Nervous system to a three-dimensional computer representation of a cockroach's leg to learn why muscles are arrayed as they are."It is like having a moving blueprint [with which] we can ask. questions of form and function," Pro Full says.
Prof Raibert believes he may be able to create a robot which can accurately mimic the movements of dinosaurs. Others talk of building crablike robots that can withstand the crash of surf.
The blending of robotics and biology also has an immediate practice application. Rodney Brooks and some of his former graduate students at MIT have taken the insect approach and spun off IS Robotics, a firm that makes $66,000 insect robots for research. Company president Colin Angle says it hopes in the near future to make "not a save-the-world robot" but legions of "vacuum-your- floor, cut-your-grass, bring-you-your-beer robots."
He predicts that within a year the company will be ready to ship by mail order a $3,000 insect-like vacuum cleaner that cleans while it scurries. - But none of this directly answers the question: Are any of these robots alive, or will any be alive in the future? 'You would want something that a spider would think was a spider.' The truest answer is that nobody knows. "No robot I know fits that description," Prof. Raibert says. "... But I wouldn't say this in the future. "
Prof. Full adds: "I was skeptical about how much biological information can be transferred, but not now. I see incredible things out there but they are just not put all together yet." Prof. Beer tries to distance himself from something that is traditionally considered God's territory. "I am not setting out to build robot cockroaches. So mostly the issue of life or not life do not come up" he says.
However, he admits there is a barrier of definition which more and more life-like robots might breach. "A lot will come down to this: If people start to act toward robots like they act towards insects, then for all practical purposes they are alive... and if you start treating them like they are alive, what difference does it make whether it is alive?"
Several years ago, Prof. Harnad, the Princeton psychobiologist, was having dinner with some of the world's leading roboticists at a conferences in Belgium. Prof. Harnad, who is a vegetarian, was asked why he didn't eat meat.
"I told them I wouldn't eat anything that has a mental state - and by the way, I would eat any of their robots, any time. After that the unofficial theme of the conference became: 'How can we built robots Harnad won't eat?'"
Today, Prof. Harnad suggests that appetite or any other evaluation carried out by humans is not the essential determinant of what is alive. "If people build a robot spider, they should not be satisfied if it convinces you and me. You would want something that a spider would think was a spider".
Mr Tilden, the Waterloo engineer, has formulated his own unique definition of life to apply to the "junkbots" he assembles. "Life", he says, beaming at a table full of his mechanical creatures - some of which are unaccountably twitching or lurching forward - "is that which moves for its own purposes".
Reproduction, on of the classic traditional definitions of a living organism, is subsumed into the robot maker. "A robot cannot reproduce itself, but a human being can be seen as a robot's way of making another robot".
Then Mr Tilden brags about how he personally is getting better at creation. His first robot took him two weeks to build. His most recent, just 40 hours.