Wall Street Bankrolls New Artificial Intelligence Software
NEW YORK — Mathematician Ben Goertzel is producing a computer program he says will take the Internet out of its "infancy" and spearhead a new era in intelligence-based software. His plans are part science fiction and part actual possibility, say critics and champions of his project.
If he succeeds, Dr. Goertzel will follow the path of ground-breaking inventors across the centuries whose radical ideas were first greeted with profound skepticism. Otherwise, his computer, called WebMind, will become yet another casualty in the field of artificial intelligence, replete with talented enthusiasts promising astounding feats and failing to deliver them. Artificial intelligence, or AI, is the attempt to get computers to simulate aspects of human intelligence like speech recognition, deduction, and the ability to learn from experience.
"We are on the verge of a transition equal in magnitude to the advent of intelligence, or the emergence of language," says Goertzel, who has taught mathematics, computer science, and psychology at universities in America, Australia, and New Zealand.
Although his vision can startle the AI novice, Goertzel's view that today's supercomputers will merge into a collective "global brain," fast outclassing human intelligence in many areas, is a well-known, and controversial, theory. Now, this imaginative academic is about to test his ideas in the deeply pragmatic world of Wall Street.
WebMind is based on software designed along structures of the human brain in all its aspects - a first in artificial intelligence, if it works. But what has stirred the interest of Wall Street executives who are funding Goertzel's company, Intelligenesis, is that in the competitive high-stakes environment of programmed trading, even the slightest superiority in intelligent Internet software yields high returns.
Goertzel has developed WebMind for financial markets and the general business community, with two software products scheduled for release Jan. 31. Some aspects of the technology will be unveiled to the public at the Financial Technology Expo in New York Sept. 15 to 17.
But Goertzel's vision reaches far beyond the realm of Wall Street. He wants to see a piece of WebMind installed in every device we use, even extending to outer space (one of the senior engineers at Intelligenesis used to work for NASA). In short, he is embarking on a largely unexplored path toward what he terms "the worldwide brain ... a self-organizing, emergent intelligence vastly exceeding that of any single human, encompassing humanity without denying us our freedom."
Not surprisingly, the mathematician's ideas have been categorically dismissed by experts from rival AI fields, while drawing the admiration of others working in similar areas. Mark Watson, an Arizona-based AI expert and author, says the young scientist "both inspires and sets the imagination wild."
"Scientists who can step outside of the boundaries of current methodologies are the creators of great new inventions and technologies," he adds. Goertzel "at least has a chance of success," Dr. Watson says, and even if his AI fails, it can help move the field forward.
"Of course there are skeptics," Goertzel notes. "If all that we were doing was accepted by everyone, then many larger and wealthier companies would be doing it." In principle, physical science shows that a brain can be simulated by a computer to any desired degree of detail, he says. And he believes 20 to 100 gigabytes of a computer's random access memory is sufficient to allow him to do this. Random access is the computer's ability to find and directly go to a particular storage location, the same way a person immediately finds a phone number without searching through the entire telephone book first.
In 1996, Goertzel found a supporter in a Wall Street global market analyst, Lisa Pazer, who inadvertently stumbled across his Web site when he was a research fellow at the University of Western Australia in Perth. "Ben's ideas were extremely novel to me, yet they seemed so obvious - or perfectly intuitive - that I was surprised that they were not more widespread," she says. Together, they decided to raise the capital to create Intelligenesis.
Ms. Pazer says the "holy grail" for Wall Street is WebMind's ability to automatically integrate the analysis of text and numerical information. Currently, she adds, financial markets either rely on basic "search and retrieval" systems for text information, or on complex data-mining number-crunching products.
Wall Street analysts are still cautious to assess the fledgling product. But "it does sound promising and innovative," says Ivy Schmerken, editor in chief of Wall Street & Technology, a publication covering information technology. She notes that the program is "not totally new" because highly competitive financial giants probably have spent millions of dollars building similar systems for their exclusive use. By offering the program commercially, Goertzel's company could benefit those who lack large information technology budgets, Ms. Schmerken says.
For Goertzel, Wall Street is just "a foot in the door," a bankroll to enter the world of ideas - or "getting to the heart of thought," he says - that interests him. When he started college in his mid-teens, he tried to answer his childhood questions about life and the universe through mathematical equations. He decided that intelligence wasn't limited to the physical structure of the brain. He saw it as a universal form of mathematical processes that interact with one another to create new processes, or ideas. This model could be duplicated on another platform, like a computer.
But it wasn't until the Internet emerged that he found a computer model powerful enough to allow him to test his theory. Eventually, he says, this mathematical model can be placed in the most mundane items, so that household and electronic devices will be infused with a kind of "understanding intelligence."
This is actually a standard vision within the AI industry that is already taking place at some levels: Computers will be able to drive your car, whip up the cake of your choice, and cook your steak the exact way you like it. But Goertzel also predicts that tomorrow's computers will check people's documents for factual errors and assess whether their ideas are similar to other publications.
In five years, "we'll have a computer operating system where you sit down and just tell the computer what you want to do," he explains. "You can talk or type, and it understands the context of what you're saying."
In his opinion, computers are already more intelligent than humans in narrow fields, but he says this will broaden to other areas over the next 10 years. There is no commercial benefit in programming a computer to experience a sunset, empathize with a novel's protagonist, or intuitively respond to a painting, he adds.
"We're doing science fiction right now, but it's real. That's what's so cool about it," he says. "It's just like people were with flight at first, you know. They were always talking about it, and then suddenly, it becomes possible."