Category: Uncategorized

This morning in the Hilton, John Boiles of Yelp and Dan Fernandez of Microsoft held a panel on Kinect hacking. Their special guests were the popular boxing Kinect robots last seen on Saturday night at the official opening party.

As you might already know, the key components of a Kinect are the IR Emitter, Color Sensor, IR Depth Sensor, Tilt Motor and a microphone array. With these tools, a Kinect can track up to 20 joints on up to 2 people simultaneously at 30fps. The new version (Kinect for Windows) adds such features as speech recognition, directional microphone capabilities and the ability to track objects as close as 18″. Interestingly, one can hook up multiple Kinects to one PC and have the same program control them all. The caveat is that only one of the Kinects in an array can track skeletal data. But color, depth and audio sensing can be assigned to any of them.

In one demo, Mr. Boiles showed how they used a programmatic interface for Half-Life 2 to insert a Kinect version of themselves into a sandbox environment powered by the real game physics engine and interact with game objects. He was able to kick objects and swing a barrel around.

Mr. Fernandez demonstrated a nerf gun that is attached to a servo motor and then through to a computer and a Kinect. When the Kinect sensor picks up a person it locks on and tracks them. Kinect measures proximity down to 10cm of accuracy, and when the person gets too close, it fires the nerf ordinance directly at their crotch.

Another demo that is in the works is the ability for a Windows Phone client to pick up Kinect data over a network. This data can then be integrated into apps on this platform. This opens up a whole new field of interaction possibilities for out of home marketing.

Mr. Boiles demonstrated a Kinect hack that allowed him to control a toy Helicopter by waving his arms. This was the biggest hit of the session, until the boxing robots slugged it out, which is always a big crowd pleaser.

This afternoon at the Intercontinental Hotel, Glenn Platt and Peg Faimon of Miami University (Ohio) gave a presentation to discuss the idea of Brands creating APIs the same way online services create APIs.

Their definition of a Brand as an API was essentially this: Have consumers build new and useful thing with brand primatives. In this context, a brand primative is a single unique element of the brand that someone can tap into.

To set a contrast, Platt and Faimon called out the Charmin mobile app that helps you find public bathrooms, and the Dominoes pizza tracker. In both cases, the Brand made interesting data available but the prescribed how that information would be delivered.

In contrast, they want to urge brands to be more flexible and open. They cite the “maker movement” as an example of how people want to create things if given the raw building blocks. As they put it “get out of the way and enable doers.” They urged brands to connect with consumers on their terms, rather than their own terms, by creating an abstracted remixable layer alongside the traditional communication layer.

Their presentation is posted on their slideshare here.

At the Omni Hotel here in Austin, MIT Media Lab has an installation previously alluded to on this blog. It graphs Facebook data regarding users’ hometowns on a globe, and using a hacked Xbox Kinect a user can spin the globe around and zoom in and out using just hand gestures. Below is a video of the installation in action:

What caught my eye is that they are still using the Kinect for Xbox. Were they to implement this with Kinect for Windows, and all the new capabilities that have been added, I wonder how they would enhance this.

In the SXSW “startup village” in the Hilton, the Head of Business development for TaskRabbit spoke about the experiences of his startup, on a panel with other startups (e.g. Hipmunk travel site)

TaskRabbit is a service that pairs seasoned background-checked helpers with people who need help with a task. Examples include furniture assembly and dry cleaning pickup. The helped pay the helpers (the “rabbits”) and TaskRabbit takes a cut.

One interesting anecdote he told was that they did A/B testing with imagery for their monthly newsletter and discovered that if the header image featured a Rabbit (i.e. a person) rather than a featured task, response rates for the same newsletter in the same geography jumped 20%.

But what was perhaps an even more interesting piece of information was a data visualization they produced in-house plotting their sales data against geography. The example he shared in particular was for San Francisco. It showed which areas of town were much more ripe for potential customers and which we’re slower, and that gave the company direction as to where to send street teams.

What interested me was that this was a lean startup with not much money to spend, but they devoted resources (“if we had our last dollar what would we spend it on?”) to this analysis. It’s smart marketing with targeted reach, and it didn’t take a lot of money to do it. What I also thought about was how this data could be visualized in real time, or even collected over time and shown as a time lapse animation. If a startup can afford to prioritize these sorts of steps, then large established companies should be able to pull it off as well, for a practically negligibly small slice of their budgets.