It is somewhat surprising that robots are not already part of our everyday lives. They have long been stock-in-trade items of science fiction, making us all familiar with the concept. The heart, or more accurately the brain, of a robot is a computer.
In recent times the personal computer has been intimately absorbed into the culture and technological infrastructure of the developed world. The computing power of the PC is increasing rapidly. So why are robots not among us? The answer, according to Hans Moravec (Scientific American, December 1999), is largely that PCs are not yet nearly powerful enough.
But, are robots not routinely used nowadays in car assembly plants? Yes, but these robots are fixed in position, doing repetitive assembly line tasks. When I speak of robots in this article I mean mobile units capable of tasks such as vacuuming the house and mowing the lawn.
A computer's power is quantified by the number of instructions per second it can execute. An instruction represents a basic task such as adding together two 10-digit numbers. Until the 1990s, computers commonly available for robotics were incapable of executing more than one million instructions per second (MIPS).
Computer power took off in the 1990s and has now reached 1,000 MIPS in the top end of desktop machines. Such power may soon allow the commercial development of robots with functions well beyond today's capabilities. Obviously the closer a robot can approach (or exceed) human capacities the more useful it will be. A really useful robot could navigate its way around, recognise a wide variety of objects (including human faces), read text, respond to voice commands, and be capable of interacting with the environment in a wide variety of ways. A human brain can control all of these functions well, but computers currently fall far short of such sophistication.
The human brain is commonly pictured as a computer, although much more powerful than any man-made device. This analogy is true in one sense, but misleading in another. Almost all computers nowadays are "universal machines", i.e. true programmable general purpose machines. On the other hand, the human brain is a sophisticated special purpose computer that developed through biological evolution.
Our earliest ancestors had to be good at food location, escaping from predators, mating, and protecting offspring to survive. These tasks mainly depended on the brain's ability to recognise and to navigate. A facility for mathematics was irrelevant for survival.
Later, with the development of language, a part of our brain evolved into a sort of universal machine, but it processes numbers very inefficiently compared to a modern computer. Computers can carry out numerical calculations millions of times faster than any human. However, they are no match for humans in functions such as recognition and navigation.
Moravec believes that it will be possible to programme general purpose computers to match the human brain within 50 years. He envisages the construction of an electronic system capable of the same level of perception, cognition and thought as a human, and programmed to do the same things as the human nervous system, including the brain.
Of course, this will be possible only if biological structure and behaviour are entirely caused by physical law, and if physical law can be fully simulated by computer. This is not a foregone conclusion, but there is compelling evidence that both conditions hold.
Today's computers can only control a robot with the capacity of an ant. A typical PC would have to be one million times more powerful to rival the performance of the human brain. Computer power now doubles every year. At that rate it would take only 30-40 years to improve by one million-fold. Moravec envisages what he terms first-generation robots with 5,000 MIPS lizard-like minds available by 2010. These would be of human size and capable of carrying out a wide variety of useful tasks such as mowing lawns or vacuuming floors. They would be unable to learn from experience.
The second-generation robot with a mouse-like brain of 100,000 MIPS will succeed the first generation. This will be much more capable and flexible and will be able to improve its performance by learning from experience.
The third generation of robots will have a monkey-like five million MIPS brain and will be able to learn quickly from mental rehearsals in simulations that model physical, cultural and psychological factors. It will have a simple inner mental life concerned only with concrete situations and people in its work area.
Fourth-generation computers, available by 2040, will have a human-like 100 million MIPS "brain" and will be capable of abstraction and generalisation. Moravec believes that these robots will be immensely formidable, capable of outperforming humans in any area of endeavour.
Inevitably such a development would lead to a fundamental restructuring of society. Whole corporations would exist without any human employees, although humans would formulate the laws governing corporate behaviour.
Ultimately humans would cease to work in the conventional sense, probably spending most of the time in social, recreational and artistic pursuits. Moravec even envisages most scientific research being conducted by mass-produced robot scientists.
Moravec's predictions raise a raft of interesting issues and I will deal with some in future articles. I suspect that his projections are very optimistic and I would predict that developments will not exceed second generation robots by 2040.
William Reville is a Senior Lecturer in Biochemistry and Director of Microscopy at UCC.