SAN JOSE, Calif. — The wait for Siri, a much-hyped virtual digital assistant application for the iPhone, may soon be over.
“Soon, really soon,” said Tom Gruber, co-founder and CTO of Siri, when asked when the free service will be available. Gruber’s appearance here at the Semantic Web conference was one of a few previews the company has given ahead of what’s expected to be a public beta release this year.
The release is highly anticipated because Siri promises capabilities long-featured in science fiction movies that have proved difficult to implement.
Gruber said Siri offers a breakthrough conversational interface that lets users have a dialogue with an intelligent software agent that understands the context of the conversation. For example, users can ask, “Find Italian restaurant for tonight” and the system will search based on your current location using links to the OpenTable site database. You could further refine the search based on the results, specifying a particular city or time, for example.
“We now have this enormous structured data trove that’s available in APIs that gives us the conditions for the first virtual personal assistant,” Gruber said. He added that the system is still being built with many features still to come after the beta release. “Now, it’s a small matter of programming,” he joked.
The idea of a voice-responsive digital assistant recalls science fiction movies, like Stanley Kubrick’s 1968 classic 2001, which featured the computer HAL. In Kubrick’s scary vision of the future, the computer is interactive — but not always cooperative.
But Gruber said that a key feature of Siri, or any system of this type, will be trust.
That’s because the more the system knows about your personal life, the more effective it is. In a brief demo, he asked Siri “I want to see Star Trek” and was presented with show times for local theaters based on his location. Theoretically, Siri could manage the whole transaction if it had access to personal credit card information.
“This is not AI, but it’s a virtual assistant that’s task-oriented,” Gruber said. “It does things for you and understands you in your terms and language.”
Siri also uses other technologies in the iPhone, such as the GPS for location-awareness, to personalize a user’s experience.
Gruber drew a distinction between what search engines do and Siri’s approach in trying to understand and anticipate user’s needs. For example, if you ask for a restaurant, Siri will ask whether you one one that’s “near where you are, or home?”
“The intent is to understand the conversation,” Gruber said. “It’s not the same for everyone — it gets to know you. Search engines are mostly optimized to be the same for everyone.”
He reeled off several other task-oriented examples that Siri is designed to handle. Ask “Giants tickets for tomorrow,” and Siri knows the reference is to the San Francisco Giants baseball team and would display the game schedule. Or ask, “What is the next flight to Minneapolis?” and Siri can show a flight schedule that assumes the starting point is the nearest airport to the user’s current location.
And by tapping into other Web resources via APIs, Siri can also serve as a kind of information butler, answering questions like “How many calories are there in a banana?”, “What time is it in Beijing?” or “What’s the weather?”
Tapping the “Big, fat head”
Gruber took pains to explain that this is far from the second coming of HAL in terms of understanding meaning and intent. The initial version will be closer to “Google-ease” understanding very specific words.
“I hope people will understand you don’t chat with it and ask ‘should I dump the boyfriend?'” he joked. “AI tried to solve universal intelligence. At Siri, we said we need to solve really practical problems on small domains.”
But Siri will still address issues or requests that have broad appeal. Gruber said it’s not about the long tail of trying to figure out every possible request. “This is more the big fat head, getting at the things everyone does.”
Gruber predicted that Siri will soon have plenty of competition, however, and that people will debate which solutions are better. He said Semantic Web technology will help companies like his come out on top by more easily figuring out inferences about tasks and a user’s preferences.
“As users learn to trust agents and store their credit card information, the payoff will be great,” he said.