COMPUTERS OF THE FUTURE: Intelligent Machines and Virtual Reality - 6. Robots and Silicon Brains
Figure 1. 'Eddie', a wall-climbing robot, lacks the brain of his human creator. But will computers and robots always be less intelligent than humans?.
Figure 2. An industrial robot uses electric welding gear to seal the internal pipework of a car exhaust.
Figure 3. 'Gnat', an insectlike robot, is so small it can be held in the palm of a hand.
Figure 4. 'Richard the First', a humanoid robot, looks into the face of one of the researchers who built it.
Will computers ever become as smart as, or even smarter than, human beings? (see Figure 1) This is one of the great unsolved problems of ARTIFICIAL INTELLIGENCE, or the study of how to make computers hat can think and learn in ways that we would consider intelligent. Some researchers argue that the human brain is simply a kind of thinking and learning machine. Therefore, in theory it ought to be possible to design an artificial brain – a computer that is as clever as a person. The difficulty in doing this is that the brain is extraordinarily complex. It contains up to 100 billion nerve cells, each one of which is joined to about 10,000 of its neighbors. Building an exact copy of a brain by using silicon chips, for example, is obviously out of the question. So how then might a machine be given some form of intelligence?
Early success in programming computers to do seemingly clever things – for instance, to play chess – led some researchers to fore cast that intelligent machines might be built by the end of the twentieth century. But machines can beat humans at chess simply by looking ahead at thousands of different bard positions and using a set of rules to choose the strongest move. This is called a brute-force strategy, and all it proves is that the computer can process information a lot more quickly than human beings can.
People may not be as fast at reviewing chess moves as a computer, but they can do other things that are far more remarkable. They can see and can understand what they see. They can learn and adapt to new situations. They can appreciate beauty, show friendship, imagine, plan ahead, and be creative.
Computers can calculate at lightning speed, but in many ways they are more stupid than an insect. In performing most tasks, computers just work their way blindly through whatever instructions are contained in the program given to them. Some scientists now doubt whether computers, operating in this slavish way, can ever be programmed to be intelligent. Instead, these researchers believe that, to become smart, computers need to be organized more like human brains and then be allowed to learn from experience.
Most computers are built around a single central processor that can only carry out tasks one at a time. Your brain, on the other hand, consists of billions of nerve cells, or neurons, each of which acts like a tiny processing element. Because of the way it is organized, the brain can work on lots of pieces of information at the same time. This makes it extremely effective at handling problems that involve, for instance, vision, language, and the association of ideas.
The vast web of interconnected neurons in the brain is called a NEURAL NETWORK. In recent years, some researchers have sought to build artificial neural networks from three-dimensional arrangements of electronic chips. Other researchers have written programs that make ordinary computers behave like neural networks. The results of both approaches have been so promising that artificial neural networks are now starting to be used in preference to ordinary computers for certain special applications. Such applications include understanding human speech and recognizing faces in pictures taken by video cameras.
Rather than being programmed in advance, artificial neural networks are "trained" to become progressively better at whatever tasks they are given to perform. In this sense, they are like children who learn by trial and error as they get older.
In the case of a neural network being trained to read a person's handwriting, various samples of that person's writing are presented to the network. The words in each sample are scanned with a video camera and turned into data for the neural network to process. Every time the neural network correctly identifies a word, it is told effectively, "good – you're doing fine." It will then respond by continuing to process other words in the same way. However, if it makes a mistake, it is told effectively, "not so good – you slipped up there." Consequently, it will adjust the state of its processing elements a little so as to improve its chance of success next time.
Neural networks are already being used by some finance companies to score the credit ratings of applicants for loans, by the military to extract information from radar signals, and by chip manufacturers to help design and inspect the circuit patterns on complex microprocessors. One of the most important future uses for neural networks will be to give robots and other machines a sense of sight.
The Robot Invasion
Most robots look nothing like the mechanical people portrayed in the movies. The most common variety has a single large arm, at the end of which is a gripper that can hold various tools, such as a paint sprayer, welding torch, drill, or screwdriver. Rows of these industrial robot arms are used on assembly lines in many kinds of factories (see Figure 2).
Each robot is taught how to do its task by a human operator who, with a set of controls, guides the machine through all of the actions it has to perform. This sequence of actions is stored as a program on a computer linked to the robot. Thereafter, the root is instructed by the computer to repeat the set of movements over and over again.
More sophisticated robots, including ones that can sense and react to their surroundings, will eventually be produced through developments in artificial intelligence. Much effort is now being expended to create animats – small, artificial animals that move about, sense their environment, and gradually learn how best to perform certain tasks (see Figure 3). Some of these animats exist only on computer screens. They may consist of a few different kinds of simple virtual organisms that are allowed to interact with one another and evolve into more complex organisms as part of a computer simulation. Other animats are three-dimensional machines, generally in the form of insectlike robots that move on jointed legs and gain information about their surroundings with the help of SENSORS. These sensors may respond to warmth, moisture, light, or other conditions in the environment. Equipped with miniature television cameras and a neural network that can recognize shapes and surface textures, an animat possesses a crude sense of sight.
Researchers working in the field of robotics foresee animats that will be used as toy pets, servants in homes, or workers in places that are dangerous or inaccessible to human beings. As time goes on, animats will be able to take on more and more complex tasks and may eventually be treated as a new form of life.
A Glimpse of the Future
Researchers at the Massachusetts Institute of Technology are developing a human-looking robot called COG (see Figure 4). One of the goals of the researchers is to try to give COG the mental ability of a two-year-old child. At Waseda University in Tokyo, scientists are working with another robot, called WABOT-2. This device can recognize music – either by its sounds or by the written note – and then can perform the music on a piano at a professional level.
Elsewhere, people are developing chess-playing computers that will consistently be able to beat the best human players in the world. Other researchers are building computer systems to translate speech from one language to another, discover new scientific and mathematical theories, and create exciting forms of music and art.
Research is underway to develop revolutionary chips and computers based on the same complicated chemicals that are found inside living things. It has even been suggested that, before too long, it will be possible to link a computer directly to a human brain. As a matter of fact, sound-sensing devices that use microchips have already been implanted in the skulls of thousands of profoundly deaf people, giving the recipients a limited amount of hearing. In the future, it will be possible to attach to the brain a tiny chip that will provide instant knowledge and expertise on any subject. Some scientists have even suggested that it may be possible to copy and store in a computer one's memory and personality so that when the body dies, the mind will exist indefinitely in a machine. Is this something we should want to happen? Recent developments in computer technology make such questions relevant and important.
Because of the rapid progress in information technology, virtual reality, and artificial intelligence, we face an exciting but also an uncertain future. We are all about to embark on a remarkable journey. And no one knows for sure where it may lead.