The announcement was part of CEO Brian Krzanich’s opening keynote at the Intel Developer Forum (IDF) in San Francisco today. Specifically, the kit taps the pattern matching engine that’s inside Curie, an IoT module that Intel has been promoting for about a year and a half.
It’s an indication not only of Intel’s growing role in IoT but of the way machine learning is creating opportunities for the world’s biggest microprocessor vendor.
The company has already indicated its next step in artificial intelligence, acquiring Nervana to gain expertise in the discipline of deep learning. For now, though, the state of the art in the field is machine learning — and machine-learning systems need lots of examples to learn from.
The idea is to let developers do all this work locally. Previously, information had to be loaded into a computer or, if there was a large amount of data involved, the cloud. That’s where the pattern recognition algorithms would be run.
With the Knowledge Builder toolkit, Intel is letting developers create applications that use Curie’s sensors and pattern-matching engine. It could cut development time “from weeks down to literally minutes,” Krzanich said.
Krzanich explained this using motion capture as an example. Curie is a small module, smaller than a fingernail, so it’s suitable for wearable devices or athletic equipment; it can be used to trace the motions of a karate kick or tennis swing. The developer could capture examples of typical motions, then feed them into Curie so it could recognize them real-time.
This kind of off-cloud processing matters because of the massive amount of data churning that IoT is likely to require. The average person generates 600 to 700 MB of data per day; that’s going to increase to 1.5 GB by 2020, he said. But that’s paltry next to the expected numbers out of IoT: 40 TB per day from airplane engines, or 1 million GB per day for a smart factory, Krzanich said.
So, it would mean a lot to not have to upload all that data to the cloud.
Krzanich had a couple of other IoT demos to show off at the keynote. BMW executive Elmar Frickenstein demonstrated an autonomous i3 car by having it drive him onto the stage. And GE showed off a smart-city system whereby street lights could measure traffic by counting pedestrians and taking note of their speed.
Intel also showed off a set of smart safety glasses that could alert factory workers, real-time, whether a component is misplaced. That example was powered by Joule, a development board that Intel announced today. It’s the successor to other kits known as Edison and Galileo, and it’s being used on a trial basis by Airbus.