Sit back and let a robot clean your house and an autopilot drive your car. Sound like science fiction? We could see it in the next few years, together with reliable human identification, better indexing of the World Wide Web, and interactive programming--and even brain chips and spinal implants, although these may take a little longer. They?re all potential spin-offs of biomimetic robotics, a field that brings together the expertise of physicists, computer scientists, engineers, and biologists, particularly those with a leaning toward neuroscience.
Research into biological neural networks (BNNs) is used to design artificial neural networks (ANNs), which mimic BNNs in the hope of capturing some of their performance. The benefits are reciprocal. "Networks in nature currently outperform even the most sophisticated artificial systems," says Josef Schmitz of the University of Bielefeld, Germany, "but by taking biological principles into account, we can greatly speed up the technological progress." Biological investigations will help build robots with improved efficiency, flexibility, and intelligence. Meanwhile, "ANNs allow a structural description that can capture the connectivity and interactions of a physiological neural network whilst remaining mathematically comprehensive and computationally tractable at the same time," explains Volker Dürr of the University of Bielefeld's biological cybernetics department. So modelling neural circuits by building robots can reveal errors in our understanding of the underlying biology, when the robots don?t behave like the living animals on which they are based.
Invertebrates are the focus of much of the neurobiological research in this field because they are simpler, and so far easier, to model, than you or I would be. Tarry, for example, is a six-legged robot that moves and reacts to obstacles like real insects do. The programming that allows Tarry to do this was designed by Schmitz and his colleague Holk Cruse, based on studies of walking in stick insects. Tarry behaves like a complex system but actually employs basic principles of insect locomotion. Someday robots like Tarry could have real uses in space exploration.
Some scientists are taking the fusion of biology and physical sciences a step further, by connecting real nerve tissue directly to electronics. Steve Potter, a neuroscientist at the Georgia Institute of Technology and Emory University in Atlanta, has connected thousands of live rat neurons to a computer through some 60 electrodes. Interested in how collections of neurons solve complicated problems, Potter is hoping that some of the emergent properties will be applicable to ANNs and perhaps even to the hardware of actual computers and machines. Potter has linked the electrical output of the nerves to a virtual rat on his computer, the first ?neurally controlled animat.? The animat moves around according to the electrical impulses that the computer receives from the nerves. Potter hopes that eventually the virtual rat will be able to learn its way around its virtual surroundings in a way that reflects the movement of its carbon-based counterparts.
The ambitious idea behind this quirky research is the desire to build ?neurocomputers,? which will use living neurons instead of silicon chips. The ultimate fusion of biology and computers, such neurocomputers, should be able to learn in the same way as a living animal. It is thought that one day, it may be possible to upgrade living brains using virtual-reality training software, like that which Keanu Reaves uses to learn martial arts in the film The Matrix.
Biomimetic robotics may be largely an academic exercise at present, but its potential spin-offs mean that it could become big business. "Billions of dollars [will be pumped into the field] in the next 10 years or so," predicts Konrad Körding of the Institute of Neuroinformatics in Zürich. Potter anticipates that eventually biomimetic technology will be "bigger than fire or wheels."
But other researchers are more sceptical. "Just because there is an increase in robots that implement neuroscientific findings, doesn?t mean that they can be easily applied to a product ready for marketing," warns Dürr. Most researchers, however, agree that neurocomputers will be in production by the end of the decade.
So how can you get a slice of the action? The overwhelming conclusion seems to be that because the research is so interdisciplinary, the field is open to almost any scientist. "It is fairly irrelevant what academic background one has," says Potter. "All one needs is curiosity, smarts, and enthusiasm." People already working in the field, who sport a variety of credentials, are testimony to this fact. Potter studied undergraduate chemistry followed by a PhD in neurobiology, whereas Dürr studied animal physiology and mathematics, followed by a PhD in visual sciences before branching into biological cybernetics.
A varied background is clearly advantageous. And although such interdisciplinary training is rare, it is becoming more popular. The University of Bielefeld, for example, offers undergraduate courses in computer sciences with biology or robotics, the only degree of its kind in Germany--although similar courses are available in the United States and elsewhere. Those with a more conventional background who would like to move into the field through a PhD or postdoc should consider the experience they bring when choosing a research area. In areas dominated by computer modelling, experience in computer sciences and the use of mathematical modelling to solve physical problems are clearly a must. However, in the more biological fields you would do equally well as a neurobiologist with a flair for mathematics, or as a physicist with a flair for biological concepts. "It is generally easier to start with physics and maths, and learn the biology afterwards," says Kording. "But," he warns, "if people do not enter biology before the age of 25, they are very likely to approach the problems with an overt bias towards the mathematical perspective." And in experimental studies, manual skills can be as important as one?s grasp of theoretical modelling.
People considering a move into a new field would do well to ponder their position with respect to the ethical implications of the research. Is robotics a suitable cause to justify testing on animals, for instance? And if we could eventually tap into brains with computers, to what extent should we be allowed to control them? Such questions are of particular interest to Root Wolpe of the Center for Bioethics at the University of Pennsylvania in Philadelphia. NASA?s chief bioethicist, Wolpe believes that in creating an organism, which by its very definition could not exist in nature, we are raising novel moral implications that must be dealt with. Nevertheless, opinion is deeply divided, and Körding, for one, is unperturbed: "Ethics about robotics is not really an issue now and will not be for many years. It is more of a playground for philosophers instead."
Better make up your own mind, before someone makes it up for you.