Intel ditches the cloud in speech recognition

Cloud this, cloud that… These days you cannot read a day’s worth of tech news without the word “cloud.” The cloud is streaming movies. The cloud is securing your passwords. The cloud is storing your documents. And the cloud is processing your voice commands.

Enough of that, says Intel.

One of the reasons to utilize a cloud infrastructure is processing power. This is something that chip giant Intel is an expert at. It ruled the desktop era, and now it wants to rule the post-computing era with its mobile-friendly chips. But being mobile-friendly means packing enough processing power while using minimum electric energy.

Deciphering human speech into something a computer could “understand” requires a lot of processing and a lot of power. That’s why Apple and Google elected to process voice commands in their massive data centers. This lowers energy consumption so your smartphone doesn’t run out of juice after listening and processing your voice commands. The downside is that this requires a reliable and constant data connection in order to reach the cloud. If that connectivity is lost, you might as well just talk to a rock.

Intel seems to realize this user experience opportunity, and according to Quartz the company is working with an “unnamed third party” (probably Nuance – who else is there?) to pack all that speech recognition goodness into a small Intel package, creating a prototype called “Jarvis“:

Intel partnered with an unnamed third party to put that company’s voice recognition software on Intel mobile processors powerful enough to parse the human voice but small enough to fit in the device that’s listening, no round trip to the cloud required. (Update: That third party is almost certainly Nuance, which is busy licensing its voice recognition software to all comers.) The result is a prototype wireless headset called “Jarvis” that sits in the wearer’s ears and connects to his or her smartphone. (Perhaps coincidentally, Jarvis is also the name of the voice recognition and artificial intelligence software in the Iron Manfranchise.) Jarvis can both listen to commands and respond in its own voice, acting as both a voice control and a personal assistant.

Not only is Intel’s voice recognition solution more responsive than those offered by its cloud-obsessed competitors, but it also leads to what Bell calls “graceful degradation,” which means that it works even when the phone it’s connected to is not online.

Obviously Intel is in the business to sell more chips, and this development shouldn’t come as a surprise. It’s expected that advancements in chip architecture will allow more processing power and become more energy efficient. However, the cloud isn’t going away easily, because whatever one chip can do, the cloud does it n times better.

There’s already a lot of hybrid clouds out there, and that combination is gaining lots of traction among the enterprise. I believe in the area of speech recognition (and mobile computing), there will be a trend of hybridization as well which the device will handle most of the processing and pass it to the cloud as it needs reinforcements.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s