Paul Allen Hires Oren Etzioni for New Artificial Intelligence Push
Paul Allen has tapped one of Seattle’s foremost professor-entrepreneurs to lead an ambitious new institute tasked with expanding the frontiers of artificial intelligence research.
Dr. Oren Etzioni, a professor in the University of Washington Computer Science Department for more than two decades with a half dozen successful startup companies to his credit, began this week as executive director of the Allen Institute for Artificial Intelligence, based in Seattle.
Etzioni says the new endeavor is a “once-in-a-lifetime opportunity” for him personally and represents a potential boon to Seattle’s research-driven innovation economy, though it also means he is done being a company founder.
Allen, the Microsoft co-founder and polymath investor and philanthropist, has backed research into artificial intelligence (AI) since at least 2001, when he brought together researchers to create a “Digital Aristotle,” conceived as “a computer embodiment of an insightful teacher.” That effort continues today under the auspices of Project Halo, which is currently working on programs that can acquire knowledge from science texts, crowdsourcing, and some manual input, and then successfully pass high-school-level biology tests.
The AI institute is modeled on the nonprofit Allen Institute for Brain Science, which Allen established in 2003 and has backed to the tune of $500 million. That effort’s mission is “to accelerate the understanding of how the human brain works in health and disease” and is itself in the tradition of “big science” projects such as the Hubble Telescope and Large Hadron Collider.
At the new institute, Etzioni will focus on the cutting edge of artificial intelligence research, which has proven to be a moving target.
“Over the years, more and more technical problems have succumbed to AI techniques,” he says. “An example is speech recognition or computer chess. In fact, there’s an old adage that if it works, it ain’t AI, which means that once we figure out how to solve a problem that used to be core AI, people turn around and say ‘Well, that’s not really AI.'”
While efforts such as the 2011 Jeopardy-champion Watson computer from IBM are impressive, Etzioni says people in the field—including Allen and himself—feel that significant areas of research remain to be addressed. One example is what is sometimes called general intelligence, as opposed to savant-like machines that are good in a narrow arena, he says.
It’s “the ability for a program to really know something,” Etzioni says. As he points out, people quip that Watson won Jeopardy—but it didn’t know it won.
“To some extent that kind of refers to consciousness, which is another dimension to this, but the point is, to what extent does the program have basic knowledge and reasoning capability, and how intelligent can it be if it doesn’t?” he says.
Etzioni says there are potentially life-saving practical applications of AI in fields such as medicine, where a “tireless medical assistant who keeps up with all the literature on your behalf” could help physicians avoid errors or extend better care to underserved parts of the world. (He does not suggest, however, that a computer could replace a doctor’s bedside manner or emotional intelligence.)
Such an assistant would need the ability to acquire “high-quality, encyclopedic knowledge that you can draw upon to make diagnoses, identify side effects of drugs, et cetera,” he says, and to do it as fields like medicine are faced with ever-expanding amounts of new data and research.
“It’s not enough that we can codify some small number of things. There’s just so much knowledge being produced that you need to figure out how to put that into the computer in a scalable way, and that’s one of our big challenges with AI, too,” Etzioni says.
Etzioni hopes the institute will benefit from the existing AI expertise already present in the region at Project Halo, as well as Microsoft Research and the University of Washington, among others. “I’m both going to be gently competitive with these places—because I want to hire the best and the brightest and so do they—but at the same time highly collaborative because we can’t do everything, and we’re part of this really exciting ecosystem,” he says.