As Apple’s new iPhone, the 4S, reaches buyers’ hands, reviewers are raving about Apple’s intelligent assistant, Siri. The virtual helper, which can recognize casual spoken questions and return answers from services like Wikipedia, Wolfram Alpha, and Yelp, is being labeled a paradigm shift in how we interact with computers.
But to truly deliver on that potential, Apple will have to make Siri more powerful than the beta version on the new iPhone, say mobile software developers and artificial intelligence experts. Today, Siri can answer a range of queries from weather forecasts to restaurant recommendations and help with tasks like calendar scheduling, but it’s far from a universal helper. For example, Siri can’t provide flight times or book movie tickets—which it could before Apple bought the technology last year from the eponymous startup.
“It’s clear that it would be technically possible to integrate any Web service into Siri; you can put a Siri front end in front of anything,” says Norman Winarsky, who helped transform DARPA-funded research at independent research lab SRI International into the Siri that was bought up by Apple. Winarsky says that combining Siri with other apps would make both the personal assistant and services it plugged into more useful. “People would be happy to optimize their services and apps for Siri,” he says. “All those very different user interfaces that have to be built—and we have to learn—can now rely on a very natural, polished user interface built by Apple using Siri.” For example, instead of manually tapping through several screens to watch a movie, a person could just say, “I’d like to watch Stanley Kubrick’s film AI, please.”
Apple’s success with the iPhone was in no small part due to its allowing outside developers to create novel apps that brought new technologies and ideas to the device. Winarsky says the same approach could give Siri a boost.
Allowing Siri to connect with other apps could result in even more significant jumps in its ability if those apps have artificial intelligence of their own, says Winarsky. Siri’s intelligence is in the way it can work out the user’s intent from the words they say, despite the use of slang or ambiguous phrasing, says Winarsky. “Today it just matches that intent up with a Web service that can help, but there’s a lot of opportunity to use artificial intelligence for that next step of acting on your intent, too,” he says. At SRI, Winarsky and colleagues are building AIs to assist with specific tasks such as travel or entertainment by drawing on knowledge of things like vacation destinations and what kinds of leisure activities fit together. Those systems are also able to personalize suggestions based on a person’s past selections and preferences, says Winarsky.
One company interested in being able to combine its own AI with Siri’s is Cleversense, says Babak Pahlavan, cofounder of the startup, which earlier this year released a personal assistant app called Alfred for Android and iOS devices. Alfred draws on data harvested from the Web, including review sites such as Yelp, to recommend restaurants, cafes, and bars based on the time of day, your location, places you’ve checked into on Facebook, and your previous feedback on the app’s recommendations.
“We are the next piece after Siri in terms of using AI to help people,” says Pahlavan, who agrees with Winarsky’s explanation that Siri’s abilities really revolve around understanding natural phrases. “Our focus is on learning so that over time we eliminate the need for you to specify what you want, and you can make high level declarations like ‘I want lunch.’ ” Siri won’t respond to a statement that general. It needs more specifics, such as “an Italian restaurant near San Francisco.” Nor can it learn from your past actions or expressed preferences. If Siri were connected to Cleversense’s technology, though, perhaps it could.
In less than three months, Alfred has dished out seven million recommendations, says Pahlavan, and more than nine in 10 users choose to give Alfred such general requests, relying on its AI to make a recommendation based on their past feedback about what they like, rather than specifying a particular type of place. “It shows that we can make good, personal recommendations and that people like that kind of intelligence that curates the noisy world around you,” says Pahlavan. His company’s AI will make its technology available for others to draw on in their own apps or Web services. “The Siri guys started a conversation about how to commercialize AI,” says Pahlavan. He hopes Apple will take the opportunity to encourage an ecosystem of AI apps, just as it did with mobile apps.
At time of writing, Apple hadn’t responded to queries asking whether more services will be added to Siri, or if outside developers will be allowed to boost Siri’s brainpower.