Image credit: https://flic.kr/p/di9TVx
This week, two major tech events took place just miles from our office. Apple announced the iPhone 5 at the Yerba Buena Center, and Tech Crunch Disrupt SF announced its winner. Scoble had a great blog post about how, despite the upgrades to Siri in iOS 6, its context awareness was still a major flaw.
Siri is quite stupid about context, location, and a bunch of other things.
He’s right, but context as simply more data is better data doesn’t solve the underlying problem. Siri’s bugs don’t come from a lack of data, they come from an inability to truly “get” its owner. When we use Siri, it’s with a subtle belief at the back of our head that we have our own personal assistant.
Essentially, Scoble’s issue boils down to “she has the information, she just isn’t aware of how to use it.”
Yet as anyone who’s ever worked as a personal assistant knows, just having more data doesn’t solve the problem. What still makes people better personal assistants is not access to information but an ability to predict necessary information before it is asked for.
Scoble mentions Siri’s inability to understand the true meaning of the query, “where is my next meeting?” A bad personal assistant would easily be able to answer that question, a good one would remind you of the salient details needed for the meeting, checking Google Maps to determine the best route.
But checking traffic patterns is something even the boss would probably remember to do on his own. A truly great personal assistant picks up on the little details. Using contextual data, a truly great digital assistant would see problems before they became problems.
That assistant would know that his boss to get a jolt of caffeine on his way back, so he’d have a Starbucks picked out. He’d look at upcoming events and, based on traffic and the length of previous meetings, decide if the calendar needed to be changed. He’d be know if the car needed more gas, and figure that into both the route and the schedule. Finally, he’d remember that his boss liked to bring home a movie for the kids after a long day’s work, so he’d have one picked out based on a combination of what was popular at the time and what the kids enjoyed.
A truly great personal assistant picks up on the little details. Using contextual data, a truly great digital assistant would see problems before they became problems.
Through apps and APIs, Siri can get access to all the raw data it needs to perform these functions. What Siri lacks is the ability to reason; to fill in the gaps rather than just search. Siri can understand that a “meeting” requires a check for a route, but it can’t predict if the car will need gas, even if it has the pricing data. It can find coffee shops, but it doesn’t know whether to add in a stop for coffee, despite having access to purchase information.
Siri can’t determine if future meetings should be moved because it can’t predict how long the meeting will last. Finally, it has no ability to pick the right movie because it doesn’t know anything about how previous movie choices relate to currently available options.
TC disrupt offered a potential solution that’s near and dear to our hearts. Prior Knowledge offers a public API that can analyze existing data and “fill in” those missing rows. It discerns patterns by understanding context. The result is that if I give it data on a daily schedule and previous gas purchase history, it can determine if i’ll need to fill up on a particular trip.
Siri can understand that a “meeting” requires a check for a route, but it can’t predict if the car will need gas, even if it has the pricing data.
By piping in data on existing movies and previous personal choices, it can select the right movie for tonight. The examples are numerous but the point is simple: using predictive analysis, digital assistants can begin to reason.
As an API marketplace, we’re excited whenever someone adds an API to their product. Yet what’s most awesome about Prior Knowledge is that their product is their API. These guys get that predictive analysis is information independent; as long as you can express the data in a format machines can parse, the specifics are irrelevant.
There’s no difference between picking movies and outfits (string), between being late and needing to get gas (boolean), between how long a meeting will last and how long the perfect movie is for a family evening (integer).
Many startups lock this technology up. They work with partners to build and test their algorithms, but hold off opening up their service to the general public. These guys have taken the diametrically opposite approach. They built v1, launched at Disrupt and said “here, play with this, show us what you can do.” At Mashape, that’s what we’re all about.
We’re enthused that an API made it into the finalists at Disrupt. As developers experiment with the power of APIs and the capacity of solid documentation and turnkey implementation, we’re excited to discovery just what the next generation of API-enabled software can do.