|
|
     |
|
|
|
by Anna Rusch
Thirty-two years ago in Stanley Kubrick's 2001: A Space Odyssey the world met HAL, a sophisticated, slightly neurotic computer, capable of human reason and experience.
Today, in 2001, HAL is no longer a fantasy, though not yet a reality. Some view the inability to develop a robot like HAL as an indication that artificial intelligence (AI) technologies have failed to produce credible results. In many ways, current AI technologies have actually surpassed those predicted in Kubrick's movie. Tracing the history of AI and placing the release of 2001: A Space Odyssey within that history provides clues as to why there is still no HAL.
AI first took off following the 1949 invention of the stored-program computer. Earlier computers required the separate configuration of thousands of wires to perform a single operation. Stored-program computers eased the complex task of entering a program, allowing for developments in computer science and theory, and eventually AI.
In 1956, John McCarthy, often referred to as the founding father of AI, organized a month-long conference titled, "The Dartmouth Summer Research Project on Artificial Intelligence." Researchers and experts gathered in Vermont to discuss achievements and outline the future of AI research. The conference also designated "artificial intelligence" as the term to describe the notion of machines that could "think."
In the years following the Dartmouth conference, AI gained a foothold as a legitimate academic discipline. Carnegie Mellon University and the Massachusetts Institute of Technology (MIT) established research centers designed to examine the many new technological challenges associated with AI. Research focused on the invention of computer systems which could not only solve problems efficiently but also "learn."
By the 1960s, computer scientists predicted that machines capable of almost any human behavior would emerge within the next generation. Encouraged by the advance of technology, AI researchers felt that a human-like robot was not only possible but also probable. The AI community envisioned an age in which computers would engage in human conversation, politics, and humor, as well as perform various tasks and chores.
Yet today, the technology needed to retrieve HAL from the realm of fiction is not yet available.
In the 1960s, Marvin Minsky, an MIT computer scientist whose research greatly influenced the producers of 2001: A Space Odyssey, subscribed to the belief that intelligent machines were destined to appear in the near future. Minsky, who is still working in AI, now anticipates HAL's debut sometime "in between four and 400 years."
Why the sudden shift in expectations? AI researchers have been very successful in understanding the mathematical connections occurring within the human brain. However, even computers capable of solving the most advanced computations do not possess the technology necessary for natural language communication.
Current machines also lack sensitivity to emotions. HAL is a conscious being, vulnerable to feelings such as pride and self-doubt. AI researchers have not yet been able to develop a computational approach to inducing consciousness. In fact, many are doubtful that machine consciousness is a possibility at all.
Machines like Deep Blue, the computer that challenged and defeated chess champion Garry Kasparov, appear intelligent to some, but that victory was possible only because of the mathematical advantage computers hold over humans. Deep Blue was able to consider millions of possible outcomes resulting from a single move in a matter of seconds, a task the human brain could never perform. Its mathematical prowess alone does not suggest intelligence.
Computers are programmed with intelligent characteristics; they do not acquire information through cognitive experiences. Today, researchers are working toward a machine that could develop its own mental capacities, like HAL did. But the timeline for the arrival of an independent, "thinking" machine is uncertain.
Still, Arthur C. Clarke, the author of the book that inspired Kubrick's film, maintained in an interview in 2000 that, "When someone says that something is possible, he is almost certainly right. When he states that it is impossible, he is very probably wrong." Though present day computers may not be as "smart" as the fictional HAL, the future holds limitless possibilities.
Microsoft's "Paperclip" provides a familiar example of the state of AI technology today. The animated icon was created to interact on a human level with users of the company's Office productivity suite. By tracking and analyzing user actions, the "Paperclip" attempts to provide pertinent information relating to the current task. However, many users find the "Paperclip" irritating, intrusive, or difficult to interact with effectively. Although simplistic, such programs are an important step toward realizing more complex levels of AI.
AI may soon be used to help computer users search the Internet more effectively. One smart search technology currently under development is openCOLA's Smart Folders program. A user sends search terms to a segment of computer code called a "bot." This bot then searches the web to find digital files that match the search parameters. These potential matches are then returned to a folder on the user's computer, and the person rates the resulting files depending on how well they match what the person wants.
Over time, the Smart Folders program will learn the individual's likes and dislikes. During future searches, the bot will only retrieve files that match the user's tastes. Through the feedback, the bots become increasingly more effective at searching.
Recent advances also allow computers to recognize and generate speech. However, unlike HAL, present day computers do not truly understand natural language.
One measure of a computer's language skill is the Turing test. According to this test, a computer can be described as "intelligent" if a human judge is unable to distinguish between a conversation with another human being and a conversation with the machine.
To date, no computer has been labeled "intelligent" by the test. Today's machines are capable only of translating sounds into texts and vice versa, but cannot comprehend what is actually being said.
Researchers continue to improve computers' natural language and speech recognition capabilities. According to Maria Gini, a professor of computer science and engineering, "Computers will become more capable of understanding what people want and more capable of providing basic services, such as buying plane tickets, ordering books, finding information, on their own."
Gini emphasizes the need for improved natural language communication if this type of progress is to occur.
"If we think that language is the major vehicle of transmission of knowledge, [then] enabling computers to understand language will allow computers to access lots of knowledge and to learn from it," she says.
At this point, natural language seems to be one of the most prominent challenges facing AI technology. Will we one day be able to communicate with computers in much the same way we do with humans? That remains to be seen. As for HAL, Kubrick and Clarke's technological vision, the AI technology of 2001 seems to claim, "I'm sorry, Dave, I'm afraid I can't do that" ... yet.
