 |
The days prophesized by science fiction writers Isaac Asimov and Philip K. Dick are on their way: the race to create robots able to see, hear, and think like human beings has begun. The intense competition in hi-technology motivates companies and governments to invest increasingly high budgets on neuroinformatics, the science to design computers that mimic the animal nervous system. |
DARPA, the advanced research agency of the American military, has announced last November the award of a multi-million dollar grant to three companies for the realization of a chip able to mimic neurons and synapses of a nervous system: the SyNAPSE project.
Massimiliano Versace is a "monfalconese" [raised in Monfalcone, Italy, NT] who got his degree in Psychology at the University of Trieste and works at the Department of Cognitive and Neural Systems at Boston University: "DARPA has announced in November that the program, initially estimated to be around $50M, will involve HP, IBM and HRL as industrial partners, and my department as an academic partner. HP and HRL will work in collaboration with my department over the next seven years: the two companies will manufacture the chip whereas my department will provide models of the nervous system that will constitute the software, like an 'operating system', that will run on the chips."
"This is a 'visionary' project: - explains Versace – DARPA is interested in the military applications of this chip, but its projects always have impact in the private sector: the internet is an example".
Massimiliano Versace obtained his degree in Psychology in Trieste in 1997, and arrived in Boston thanks to a fellowship, where he works with scientists of the caliber of Gail Carpenter and Stephen Grossberg: "an interesting aspect is the difference between the Italian and American academic system that I have experienced so far: - he says – in Italy only walls are in between companies and academia."
In the past few years Asian manufacturers have been able to reproduce even the most recent computer processors pushing the computer industry and the American military to seek new routes: "DARPA and its industrial partners are betting high resources in developing new paradigms that will allow the US to maintain the supremacy in the fields of hi-tech and computer science. - says Versace – The path chosen is to build chips inspired by the animal nervous system, able to perform computations still impossible for traditional CPUs or supercomputers. One of the objectives of the program is to build a low-cost, low-power neural co-processor that will supplement the CPU of future PCs, enabling common personal computers to have functions such as visual and auditory perception and reasoning typical of animal intelligence. In summary, DARPA wants to finance the introduction of a new family of neural processors that will be paired up with, and one day may substitute, conventional CPUs".
The objective in the military domain is to substitute soldiers in the battlefield, deploying instead robots in place of soldiers able to fight autonomously: "Autonomous robots equipped with machine guns and other weapons are already being tested. – adds Versace – and we can say that the 'law of robotics' of Asimov, which forbid robots from hurting humans, is already being broken. This opens new and interesting issues for the future, for example the liability of the actions of the robots". Commercializable applications are numerous: Already in circulation is a cell phone equipped with a chip that mimics the functioning of the human ear and is able to filter the background noise of phone conversations, enabling the listener to hear only the voice of the speaker. "We can foresee several applications in the near future – says Versace – it will be possible to introduce cars equipped with sensing, smart surveillance systems, climate monitoring systems, etc."
Another application field of neuroinformatics is space exploration: the growing distances that can be reached makes it increasingly difficult to efficiently remote-control machines deployed in space. "A signal traveling at the speed of light takes approximately ten minutes to reach Mars from Earth: - explains Versace – a reaction time too slow to allow robots to efficiently interact with its environment. Neural engineering will allow for building robots that are able to compute autonomous plans and adapt to environmental changes to achieve generic goals, such as the exploration of a section of a planet".
by Giovanni Tomasin |