Section 3.8 Special Topic: Intelligent Agents
Wouldn't it be nice if we had a computer program that could schedule appointments for us, remind us of meetings and commitments, find information for us on the WWW, and manage our e-mail messages for us? Wouldn't it be nice to have a computerized personal assistant?
Actually, such programs are called intelligent agents, which are programs that are capable of acting autonomously to carry out certain tasks. Intelligent agent technology is becoming an important research area in computer science. Most agent programs incorporate some kind of machine learning capability, so that their performance improves over time.
As a typical agent activity, suppose I was able to tell my intelligent agent to buy me a copy of a certain book that I just heard about. Given a command like “buy me a copy of X,” the agent would perform a search of online book sellers and come up with the best deal. Once it had found the best buy, the agent would communicate with a computer-based agent representing the book seller. My agent would make the order and pay for it (assuming I gave it authority to do so), and the book seller's agent would process the order.
As far-fetched as the capability may now seem, this is the direction that research in this area is headed. Researchers are developing agent languages and describing protocols that agents can use to exchange information in a reliable and trustworthy environment. Obviously, you wouldn't want your agent to give your money to a fraudulent book seller, so there are significant problems to solve in this area that go well beyond the problem of simply exchanging information between two agents.
The best way to learn more about this research area is to do a Web search using the search string “Intelligent Agent.” There are numerous research groups and companies that provide online descriptions and demos of their products.