Imagine a “smart refrigerator” that lets you order milk or soda or coffee from your local grocery store whenever you’re running low. It’s an old idea, but Mounir Shita looks at it in a new way.
Shita believes such a contraption could generate about $14,000 in revenue over its lifespan. Most of this would go to the grocer, and in all likelihood, some sort of middle-man would take a cut (think Google or Amazon). But Shita wants to replace those middlemen and share the revenues with the company that made the fridge and the ISP that provides the internet connection.
With his new company, Kimera Systems, he wants to help create all sorts of internet-connected devices without help from the Googles and the Amazons, spreading the wealth to electronics makers and broadband providers instead. For consumers, who’s actually taking a cut of all these purchases might not be important. But Shita sees it as a way to fund Kimera’s real purpose: the creation of artificially intelligent virtual assistants that can help you with just about any aspect of your life, from handling grocery shopping to scheduling meetings.
What It Is
Kimera is fashioning a technical solution it calls Nigel, and it includes three things. There’s a Siri-like virtual assistant that sits on your smartphone, letting you control your smart refrigerator. There’s a “personal cloud” service that stores data on behalf of your frig. And there’s server software that lets the frig manufacturer and ISP keep track of who what groceries you ordered, collect payments, and disperse them among all the companies involved.
The hope is that manufacturers and ISPs will use this system to build not just smart refrigerators but, well, whatever other internet-connected devices they want to build. And because they can control where the money goes, they have ample incentive to build them.
You might argue that this could give ISPs a way of taxing all transactions that happen on their networks. But middle men—ranging from credit card payment processors to marketplace sites like eBay and Amazon—are already taking a cut of so much that happens online.
For Kimera, the trick lies in building a virtual assistant that’s better than what you can get from the Googles and the Amazons of the world. And that’s where Shita’s true passion lies.
The Original Nigel
Shita has always been fascinated by the future. As a kid in Norway, he spent his time, like so many other geeks, taking things apart and trying to build new inventions. “My room looked like something out of Star Trek,” he says. “If the time machine had been invented, I would already have gone to live in the future.”
In 2012, after earning degrees in electrical engineering and computer science and moving to the US to start a few startups, he met Nigel Deighton—the virtual assistant’s namesake—and the two started spitballing ideas for what the future should actually be like, and how to get there. They decided they wanted to be able to talk to computers in a much more conversational, human-like manner than tools like Siri enable today. And they set out to make it happen.
Instead of reviewing the literature, which they worried would limit their creativity, they decided to try coming up with a way to do artificial intelligence on their own first own. Shita says he fully expected to just end up “reinventing the wheel,” but soon they had a system that Shita says worked surprisingly well.
That’s when Shita and Deighton realized they had enough to start a real company. Unfortunately, Deighton passed away in 2013. The virtual assistant Nigel was named in his honor.
The Why and the What
The idea is to understand the “why” and not just “what” of a voice command. One of Shita’s favorite examples is the statement: “I’m out of Coke.” If you’re at home, you might mean that you want to add Coke to your shopping list. If you’re at a restaurant, you might want Nigel to text your waiter. If you’re a drug deal, well, use your imagination.
Shita says Nigel is smart enough to take context like location into account while determining what you want it to do. And it goes further still. If you’re at a restaurant and Nigel sees that you have a business meeting on your calendar, it will mute your phone. If you’re running late for the meeting—and everyone else a the meeting also happens to use Nigel—you can ask “is everyone there?” and it will give you an answer.
It’s hard to tell just how advanced Nigel really is. Unlike the academics at companies like Facebook and Google who pioneered the “deep learning” subfield of AI, Kimera hasn’t published the algorithm for peer review. Shita says that’s because of where the company is in the patent filing process—and because outside experts who have seen the system are under strict non-disclosure agreements from their respective companies.
‘I’m a Believer’
But Steve Taylor, an AI expert and former VP of business development for robotics company RoboKind who consulted Kimera on its intellectual property, says the company has developed some pretty sophisticated technology. “I’m a believer. It should work and it should learn.”
Shita was willing to say that like deep learning systems, Kimera is based on creating networks of artificial neurons that share information with each other. He says Kimera’s approach isn’t in competition with deep learning, but is rather something that could actually prepare a data set to be passed along to a deep learning system for further processing.
Of course, we can’t truly judge Nigel’s talents until it’s out in the wild. It must compete head-to-head with other virtual assistants from companies like Microsoft, Google, and a whole host of startups, such as Viv, a company founded by Siri’s creators that aspires to create a smarter artificial intelligence that, like Nigel, can weave together data from many sources to intuit the context and purpose of a request. But Kimera has something that no everyone else has: a business model that could appeal to some of the world’s largest companies.