Siri, an ambitious new free iPhone application, is now available in the App Store–and it’s not Just Another iPhone Application. Based on $150 million of research by the Stanford Research Institute and DARPA (the Defense Advanced Research Products Agency), Siri aims to be a “virtual personal assistant” that understands your spoken requests–”best sushi in san francisco,” “remind me to order flowers,” “order tickets to a show at the Castro Theater”–and takes action on your behalf.
Retrieving information by voice on the iPhone is nothing new–Google’s Mobile App is just one of several that let you search the Web by speaking. But Siri isn’t Web search. It’s all about actions you want to take, and it returns information and opportunities to do things, not search results. And it uses the iPhone’s GPS to refine its responses to your local area.
(Right now, Siri is designed for the iPhone 3GS; versions for the iPhone 3G and iPod Touch, as well as other mobile platforms, are in the works.)
It all happens via a slick interface that looks a little bit like an instant-messaging session, with embedded info and actions (such as the ability to make a phone call).
For Siri to work well, it needs to do three challenging things really well:
- Its voice recognition needs to understand what you said;
- Siri needs to correctly parse the meaning;
- It needs to return relevant information.
In most of my tests, Siri handled all three tasks adeptly. The voice recognition (which is powered by Nuance, creator of Dragon Dictation for the iPhone) was perfect about ninety percent of the time. Even when it wasn’t, Siri was almost always able to figure out the gist of my request. And when it got the gist, it usually provided useful results. (If you’d rather just type requests, you can.)
But not always. For instance, here Siri knew I was asking for information on animated movies–pretty impressive–but while it told me it was giving me information on ones playing in San Francisco, it mostly listed ones that won’t be out for months.
The goal of Siri’s creators is to build in lots of functionality for transactions such as booking a taxi, making a restaurant reservation, and buying movie tickets. (That’ll both make the application as useful as possible and let Siri make money by collecting a finder’s fee on transactions you complete.) Some of those features are already there, but not all of them, and when Siri can’t take action, the experience is sometimes a little clunky:
One other criticism: Siri tries to not only tell you about local businesses, but about the best local businesses, based on user reviews at sites such as Yelp and CitySearch. When there’s a critical mass of such reviews, this technique works well. But in my tests, Siri sometimes claimed a restaurant or other business was recommended based on too little information to mean much. It was the only time that its responses to my requests seemed crude and unhelpful rather than surprisingly clever.
Siri is neat, useful, and well-designed. But I’m at least as excited by its potential as by this initial version. If it adds all the data and partners it needs to respond effectively to just about anything you ask it for, it could accomplish something that remarkably few applications or services ever have: beating Google hands-down at retrieving practical information from the Web and helping you make use of it.