Internet innovator works to make the artificial more intelligent
Six years ago, while flying from Seattle to Los Angeles for his brother’s wedding, Oren Etzioni began to wonder what his fellow passengers had spent on their tickets. So he asked a few, and was puzzled by their answers. No one had paid the same amount; those who had bought the latest had paid the least; and he had paid more than anyone. To the UW computer science professor, it simply didn’t compute.
So Etzioni started thinking about how to use historical data to predict future prices. The result was Farecast, a successful start-up company that helps prospective air travelers know the best time to buy their tickets.
Etzioni is also the wunderkind behind Netbot and MetaCrawler—seminal innovations from the Internet’s salad days. His current projects include the search tools KnowItAll and Opine. Although he’s only 43, Etzioni is a genuine pioneer in the world of artificial intelligence. For 20 years, he has been bringing his very real intelligence to bear on it.
“I think of the computer as a big pencil,” he says. “Some people find the object itself fascinating, but I just think of it as a tool. Computers are very rigid and literal-minded and dumb. So a huge motivation is to make computers our friends.”
With his unpretentious manner, unruly hair and deep purple T-shirt that says—what else?—“artificial intelligence,” Etzioni hardly has the look of a high-flyer in the world of computer science. In fact, he’s decidedly down-to-earth. He likes talking about the recondite research questions that interest him even with people for whom opening a Word document is a victory.
While many of his ideas lead to commercial ventures, Etzioni is primarily a researcher and teacher who wants to develop solutions and tools to help real people. Most of all, he wants to design a more useful Web that will eliminate barriers among people, foster communication across nationalities and cultures and be accessible to all.
“Computers are very rigid and literal-minded and dumb. So a huge motivation is to make computers our friends.”
Oren Etzioni
“There are researchers all over the world who excel at top-down ideas that are beautiful in their own right but don’t connect to useful applications,” says Andrew McCallum, an associate professor of computer science at University of Massachusetts, Amherst, and a researcher in artificial intelligence. “Many people do well at one or the other, but Oren’s tremendous strength is how he brings these two together.”
Etzioni’s interest in artificial intelligence has nothing to do with inventing Robocops to patrol the streets or anything remotely Hollywood. Rather, he wants to figure out how it can be used to make computers better at “knowing” what people want when they ask for information. Opine is a good example—a program that applies text-understanding techniques to product reviews. It can automatically identify the product attributes that people care about (e.g., hotel location), relevant opinions about the attributes (e.g., “perfect”), and more. Opine can distill hundreds of reviews into a concise summary of attribute and opinion, allowing people to understand, at a glance, the gist of the reviews without having to read each one.
Likewise, KnowItAll takes the massive body of information that a Google search would return and culls it for the results most likely to be relevant. Etzioni says it “transforms the problem of answering questions from finding a needle in a haystack to a process of being presented with a variety of needles and choosing the one you want.” He hopes a version of both products will be ready for the market in five years or so.
One product is already here. Farecast claims to predict with 70 to 75 percent accuracy whether airline ticket prices are likely to rise or fall. Etzioni’s quirky sense of humor led him originally to name the project Hamlet—a slight twist on the Danish prince’s famed query: “To buy or not to buy. That is the question.” “Farecast is helping people save money,” he says. Etzioni also sits on the board of Zillow, another Seattle-area company that uses predictive technology to calculate the market value of real estate.
And as the director of the UW’s Turing Center, a forum dedicated to exploring communication issues from both the human and machine perspective, Etzioni is taking steps that he hopes will lead to the creation of a semantic Web or Web 3.0. (The Turing Center was founded in 2005 with a multi-million-dollar gift from Jonathan Pool, an entrepreneuer and former UW professor.) Anyone who searches the Web using today’s engines knows that a search will generate hundreds of pages through which they’ll have to sift to find useful nuggets of information. Etzioni is laying the groundwork for a Web that could bring human-like intelligence to search questions so that they yield information in tidy pre-organized bundles. The conundrum is how to get the intelligence into the machine.
Etzioni becomes particularly animated when he talks about another project he is working on called panlingual translation—a computer program that aims to translate from any of the world’s languages to any other. It’s a particularly challenging problem because there are there are 49 million potential language pairs. But he’s taken a preliminary step toward panlingual translation, he says, by creating a “panimages” search engine. This improves search results by automatically translating terms that probably won’t yield many images (say, the Turkish word for “rhinoceros”) into ones that will (the English word “rhinoceros”—because the English Web is so much larger than the Turkish).
“When you create knowledge, it’s more interesting than following other people’s creations.”
Oren Etzioni
This quest is in sync with the Turing Center’s mission of promoting communications across nationalities and cultures. And, at a personal level, it’s an appropriate quest for Etzioni, who knows what it’s like to spend years in one culture and then live in the shock of a new one.
Etzioni was 14 when he moved with his father to New York City from Israel, where he had spent his childhood. He first became interested in computers because his father, Amitai Etzioni, was always giving him articles about artificial intelligence. “He was worried that I would become a pure mathematician,” Etzioni says. He felt lost in the new culture until he discovered what is now considered one of the dinosaurs of personal computing, the TRS-80. From this adolescent fascination Etzioni went on to study computer science at Harvard, earning that university’s very first degree in the discipline in 1986.
Etzioni explains away this groundbreaking achievement with characteristic modesty: “Until that time, Harvard’s position was that computer science might be a fad.” He’s similiarly understated when describing his father, “a sociology professor.” Born in Germany in 1929, Amitai Etzioni fled the Nazis, dropped out of high school to fight as an Israeli commando in the Israeli War of Independence, earned his doctorate at the University of California, Berkeley, taught at Columbia University for 20 years and founded the Communitarian Social Movement. He is one of the U.S.’s most senior and well-respected intellectuals.
Communitarianism seeks to shift the focus from the individual to civil societies. The elder Etzioni has a vision of a society whose members care profoundly about one another, assume responsibilities and attend to the common good. Although Oren Etzioni’s field of research is radically different from his father’s, it’s not hard to see the intersection where their underlying values meet. The father wants a society in which the common good is paramount, while the son wants a Web that will be more democratic in its accessibility to all people despite their linguistic and cultural differences.
For all Etzioni’s commercial successes, he says he’s more comfortable in his role at the University. In 1999 he took a year’s leave and worked in downtown Seattle as the chief technology officer for Go2Net, a consolidator and operator of high-traffic Web sites. “It wasn’t for me,” he says. “I find that with people under the pressure of time and money, there are more complex interpersonal dynamics than I enjoy. It’s like the difference between poker and bridge. I play both, but I wouldn’t play poker full time.”
By contrast, he treasures his relationships with graduate students, whose quality, he says, just keeps increasing. He’s proud to be contributing more than just products to the marketplace. “A lot of our students go to Amazon and Google and Microsoft,” he says. “Several of our graduates are working for start-ups around town.”
Although Etzioni knew from an early age that he wanted to be a professor, he imagined he’d be using his spare time to take classes in other disciplines.
“But when you create knowledge, it’s more interesting than following other people’s creations,” he says. “I’ve often thought, ‘I really should take a class in X,’ but I would be learning elementary things. Once you’ve experienced the creative edge, stood on that ledge and peered beyond into the murky future, that for me is a lot more exciting than standing in the 1940s or the 1600s.” Etzioni reflects for a moment and smiles. “But there’s room for all different tastes.”
Tech transfer: from prototype to product
Once upon a time, researchers at a top-flight university like the UW kept their ideas confined to learned journals. Not anymore. UW Tech Transfer serves as a bridge between the faculty workbench and the commercial world where good ideas benefit people.
Established in 1982, Tech Transfer continues to make its mark. In fiscal 2006 there were 10 start-up companies created from the “germs” of research at the UW, compared to just three in 2005. Most of those companies are located in the Puget Sound region. There were also 153 commercialization agreements completed in 2006, versus 109 the previous year.
One of the newest UW inventions arrives in selected retail stores this month—the Ultreo toothbrush. It is the first power toothbrush to successfully use ultrasound technology. In one minute it can remove up to 95 percent of the plaque in hard-to-reach areas. The toothbrush is also a symbol of the UW’s interdisciplinary research. Born in the UW Applied Physics Lab, it was fine-tuned through two UW departments, neurological surgery and pediatric dentistry, and brought to the marketplace through Tech Transfer.
UW Tech Transfer also licensed the technology for researchers at the UW’s Department of Civil and Environmental Engineering who created a startup company to deliver online, rich-media content and software to train the nearly one million employees in the construction paving industry. The company, Pavia Systems, Inc., is now selling modular, customizable training systems to construction paving companies around the world. Professors Joe Mahoney and Steve Muench, and Senior Research Engineer George White, created the content and delivery system.
The length of time required for laboratory research to yield a start-up varies from project to project. Oren Etzioni’s tool to predict whether airfares will rise or fall turned into the start-up Farecast in a matter of months.
“Many times the kinds of technologies we have are very early stage, and there can be a very long time between an idea and making and selling a product,” says Jim Severson, vice provost for intellectual property and head of Tech Transfer. “People also think that all we do is engineering, School of Medicine and computer science projects, but we’re much broader than that.”
Severson cites LegSim, a product that emerged from the UW’s political science department. In LegSim, students participate in a virtual Congress and learn through role-playing how new laws are made. With the help of the George Lucas Learning Foundation, LegSim will be the core of a new Advanced Placement Government curriculum at Bellevue High School.
One of the UW’s more recent efforts to support projects that pipeline from the University to the world of commerce is the Technology Gap Innovation Fund, started in 2004. So far more than $1.8 million has been awarded to 37 projects. The fund is a resource for UW inventions that are commercially promising but need a way to make the monetary leap from academic research to full-fledged commercial product or service. The money is used to test or refine innovations or create prototypes in anticipation of licensing.
The commercial world can expect to see UW Tech Transfer refine its processes to make working with the University on patents, licensing agreements and copyrights even more effective than it is today. “Technology transfer is part of the UW being a tier-one research institution,” says Severson. “It’s a service to the faculty and supports the research mission of the university for the public good. We are the vehicle to get these ideas into a stream of commerce.”