How much intelligence do computers show? Actually, quite a lot. Modern computers carry out complex computations at blinding speeds, often performing millions of computations per second—a capability far beyond that of mere mortals. It is therefore not surprising that computers are more proficient than people at doing repetitive tasks requiring speed and accuracy. This is where the concept of AI or Artificial Intelligence comes in; it is the capacity of computers to demonstrate performance that, if it were produced by human beings, would be described as showing intelligence. But is it possible to construct robots capable of interacting with humans in more meaningful ways? Research carried out at MIT in 2003, termed the Cog Project, suggests that the answer may be yes.
One product of the research was Kismet, an autonomous robot designed to engage in social interactions with humans. The development of Kismet was inspired by the way infants learn to communicate with adults. According to researcher Cynthia Breazeal, basic motivational factors play an important role in establishing meaningful social interactions between human infants and their caretakers. In an attempt to understand how these interactions develop, the MIT team had endowed Kismet with a motivational system that mirrors our own. So it could maintain homeostasis, learn behaviour to satisfy its needs as well as make a wide range of facial expressions which would tell potential caregivers what needs must be tended to.
An all too human feature built into Kismet was its capacity to become overwhelmed by too much stimulation. When overstimulated, Kismet was designed to terminate the interaction so that it could restore homeostatic balance. To do so, Kismet would shut its eyes and “go to sleep”. When balance would be achieved, Kismet would wake up to resume interaction. Through a series of such interactions, the robot learned how its actions could influence the behaviour of the caretaker, thus ultimately learning how to have its needs met.
Efforts to demonstrate computer intelligence with regard to language—clearly an important human ability—have had somewhat mixed results. On one hand, the language abilities demonstrated by computers are remarkable. For example, banks, credit unions, and credit card companies now regularly use computerized voice recognition systems to handle certain business transactions, such as customer calls to check account balances. Even more impressive are computers that can converse with their own owners and carry out a variety of tasks, including booking airline reservations. These complex machines possess large vocabularies, grasp syntax well enough to be able to understand normal sentences, and know when to ask relevant questions if they do not understand or do not have enough information to act.
On the other hand, it has proved frustratingly difficult to teach computers to comprehend many of the subtleties of human speech. Many ordinary activities that most people take for granted, such as understanding an everyday conversation, exceed the capabilities of even the most powerful of today’s computers. In response to this, researchers have designed computers that imitate the way the human brain, perhaps the most powerful computer in the universe, operates. Computer processes information in a sequential fashion, one step at a time; in contrast, the brain processes the input from all our senses simultaneously through a complex network of highly connected neurons. The new computer systems, called neural networks, are structures consisting of highly interconnected elementary computational networks that work together in parallel. Neural networks have the capacity to learn from experience by adjusting the strength of the output from individual units based on new information.
Where does all this leave us with respect to artificial intelligence? Researchers who specialize in this field would readily admit those early capacities of computers to show such characteristics as intention, understanding, and consciousness were greatly overstated. However, these specialists note that computers, in certain contexts, demonstrate performance that closely resembles that of intelligent human beings.
– Contributed by Ankita
Picture Credits: thegreatcourses.com