What could be one of the most significant advances in artificial intelligence in a decade is heading toward the iPhone App store this fall.
Siri, a San Jose company, announced Wednesday that it would offer an "intelligent agent" for Apple's iPhone that would, the company said, be able to find movie theaters, book restaurant reservations and airline flights, buy from online retail sites and even answer trivia questions like "How many calories are in a banana," all by understanding spoken commands.
Dag Kittlaus, CEO of Siri, which emerged from stealth mode to announce the product, said, "The future of search isn't search. It is a conversation with someone you trust."
Experts in artificial intelligence, or AI, say Siri will either be the first "intelligent agent" that responds to natural language -- or the most recent failure in a series of spectacularly unsuccessful attempts to write software code that replicates some basic functions of the human brain.
Precursors to Siri included Apple's "Knowledge Navigator," touted by then-CEO John Sculley in 1987, and a project Microsoft dubbed "Hailstorm," which got canceled before it was launched.
"I am skeptical of anything that uses the word 'intelligent' to describe itself," said Charles Petrie, a senior research scientist at Stanford University's Computer Science Department and a member of Stanford's renowned artificial intelligence lab. Petrie has not had a chance to try Siri, which is currently available only to a limited number of testers associated with the company.
But the company describes an intriguing vision. Kittlaus, in a demonstration, speaks into his iPhone: "Siri, I want to see 'Star Trek.'"
Within milliseconds, the phone displays a result that shows a nearby theater where the movie is playing. If Kittlaus wanted, he could click on the result and Siri would buy the ticket for him. Or he could ask Siri, a Scandinavian girl's name that means "beautiful victory," to find a theater closer to his home in South San Jose.
Tom Gruber, a computer scientist who began working on AI-related projects in the early 1980s, was also initially skeptical of Siri. After two decades of research, Gruber was all too aware of the limitations of machines.
But within five minutes Gruber realized that Kittlaus and his cofounder, Adam Cheyer, had cracked open "the grand opportunity." Gruber immediately signed on as chief technology officer. What impressed him so much was Siri's ability to offer a consumer product built on breakthroughs in machine learning, computer systems that learn from experience, as well as advances in natural language processing and in something geeks call the programmable Web -- software code that lets Web sites share their data.
Siri could use the code, also known as APIs or application programming interfaces, to search for the cheapest flight to Denver on a site like Kayak.com and then add it to a Yahoo calendar. Or it could scan AllMenus.com and Yelp to find if there was a popular sushi place within four blocks of a user's location.
Siri's abilities are well beyond those of Google and the other major search engines, which are still primarily content-indexing systems, Gruber said.
Siri is not the first company to realize that knitting APIs together might be useful. Rearden Commerce of Foster City already offers a Web-based personal assistant for business customers that can book travel, coordinate schedules, control costs and more.
But Norman Winarsky, a member of Siri's board of directors, said Siri is much more than just an integrator of Web services. Indeed, it's the culmination of one of the government's largest artificial-intelligence projects.
In 2003, the Pentagon's Defense Advanced Research Projects Agency, which is most famous for sponsoring the research that led to the development of the Internet, awarded SRI International the first of several grants to develop CALO, a Cognitive Agent that Learns and Observes.
Over the next five years, DARPA invested $150 million in the project, said Winarsky, who is a vice president at SRI International. Hundreds of computer scientists and nearly three dozen universities and corporate research centers worked on parts of the CALO problem.
"I think we are going to surprise a lot of people with what's possible now," Kittlaus said. Kittlaus was an entrepreneur-in-residence at SRI International when he cofounded Siri in December 2007 with Cheyer, the chief architect of CALO, and Gruber.
A former executive at Motorola and Telenor Mobile, the Scandinavian telecom giant, Kittlaus imagined putting Siri's intelligence on the iPhone from the start. "The iPhone single-handedly changed mobile," he said. People were ready to do all kinds of new things with their phone, provided it was easy, fun and free.
Siri's business model is simple. As a virtual agent, Siri will ask companies to give it a cut of the transactions it brokers. Regular people won't have to pay for the service.
"It is one of the first applications of AI that could really benefit consumers," said Nova Spivack, chief executive of Radar Networks, another spinout of the CALO project that is developing new Web technologies.
Spivack, who has seen a demo of Siri but not yet tried the software, said a digital assistant with artificial intelligence could perform well provided its roles are limited to certain defined areas. For example, Siri will not be able to help someone choose a pet or provide relationship advice. But it could excel at automating tedious tasks, like finding cheap airport parking or a public bathroom with a diaper-changing table.
There won't be much room for error. Siri will need to prove that it is both reliable and secure in order to win consumers' trust. "It better do a really good job about getting me the reservation I wanted," Spivack said. "If it sends me on a wild-goose chase, I will fire it, just as I would fire a real-life assistant."
(c) 2009, San Jose Mercury News (San Jose, Calif.).
Visit Mercury Center, the World Wide Web site of the Mercury News, at www.bayarea.com/mld/mercurynews
Distributed by McClatchy-Tribune Information Services.
Explore further: Research reveals we may need a new definition for privacy