It is great to see the “personal digital assistant” hype go into full gear this past week with Apple’s inclusion of SIRI at the OS level.
SIRI is a smart technology. After all, it’s $150 million of taxpayer funded DARPA research. Even if it’s a bit dated.
Apple was very smart to integrate SIRI within the REALLY IMPORTANT embedded communication apps on the iPhone 4S — The previous “app version” of SIRI could barely handle integration with six APIs and could draw upon very limited data pools to get you a Taxi or a Restaurant reservation (from within it’s limited list). The illusion of intelligence continues with it’s access to Wikipedia headlines (an old chatbot trick), but it feels intuitive, which is important.
The real issue comes with “cross talk,” where an AI engine cannot parse between similar phrases or words that have different meanings. i.e. “I need to find Sam Adams.” Are we talking about the beer, the mayor of Portland, an historical reference or a friend with that name?
What’s the solution?
AI-based virtual agents that are tailored for a specific set of data; brand information and narrow-focused engagement; i.e. one agent per application. The dream of having a single agent who can access data from dozens of apps simply isn’t in the cards, not in the near future.
So, create a personal agent for each application. For each mobile brand engagement. It’s quick. It’s affordable. It’s what we do.
Of course, we do really enjoy seeing the enthusiasm for virtual agents that SIRI has brought to the forefront. For that alone, SIRI is a wonderful thing.
ASIDE: Let’s not forget, the voice recognition feature is from NUANCE, not SIRI. We can integrate with cloud based voice recognition platforms very easily. But, we’re wary of the issues these systems still confound us with, where a 10% failure rate can create incomprehensible user input. People type on their phones all day long. It’s the #1 form of written communication in the World. It’s private. It’s quiet. We still like typed conversations.
