Kostas Eleftheriou has a problem with his iPhone. "If I’m writing and I just miss a single letter by a little bit, the keyboard doesn’t know what I mean," he says. "It can’t be very ambiguous. It surely could make a better guess about what I tried to type. I found myself having to stop on the sidewalk when I was typing something, and I don’t think it should be that way."
So he and business partner Ioannis Verdelis set out to build a better predictive-typing app for which they set themselves an impossibly high standard. "If our keyboard technology was good enough, you’d be able to type without looking," Verdelis says. The result is Fleksy, an iPhone app that uses artificial intelligence to figure out what your fat fingers tried to spell on that tiny touch-screen keyboard. And when Verdelis says "type without looking," he really means it: Fleksy’s first version is specifically for the blind.
This wasn’t their original intention. But after some quick research, the pair learned that the iPhone has become popular within the visually impaired community, thanks to Apple’s accessibility feature, in which a voice simulator "reads" whatever lies under a hovered finger. The process works, but it’s incredibly tedious. Verdelis and Eleftheriou realized they could change that system. "We thought we could have an immediate application for some people who really need it," says Verdelis. "It gave us the opportunity to challenge the technology, to prove it. If someone is blind, then you have no excuse. It has to work."
Fleksy is the first release from Syntellia, the startup Verdelis and Eleftheriou launched with the goal of using artificial intelligence to enhance everyday technology. Both natives of Greece, they met studying computer science at university in the U.K. "When I was introduced to computers and programming I realized that the only thing that limits what a computer can do is less about the CPU speed and more about the imagination of the person writing code," says Eleftheriou. He started his work on Fleksy by thinking of the computer as a sentient being. "With touch screens, we give a new sense to the computers," he explains. "Apart from the eyes, using a camera, or hearing, with a microphone, we give them the sense of touch. But it’s a very primitive sense, even for us. It seems that still it’s even more primitive the way the computers are currently using it." Fleksy trains the computer to not just register where a user has pressed on the keyboard, but to learn where that user tends to press, as well as what they usually talk about, what words they’re more apt to use, etc. "We’re trying to make computers understand what people are thinking, and then try to think in the same way," says Verdelis.
Fleksy launched for the visually impaired about two months ago, and a trip to a National Federation for the Blind conference got the word-of-mouth ball rolling. Despite rave reviews, Verdelis and Eleftheriou are constantly accepting suggestions and feedback on possible improvements. "Of course, because it’s a keyboard technology, the first thing someone does when they get it is go and write something," Verdelis laughs. They’re also getting some surprising fan mail. "The other day we got an email from a paralyzed girl who wrote to us using an eye tracker to type, which was incredible," he says. "She wanted our keyboard technology to be on her equipment. We suddenly see the impact of this technology on somebody who has a real need for it, and a real use for it. That’s very fulfilling."
The Fleksy site and App Store page are peppered with disclaimers that the technology is still by and large optimized for use by the visually impaired community, although the goal now is to customize it for everyone—despite the limitations of iOS. (Apple won’t allow any third-party keyboards to be integrated system-wide, although Fleksy can now message, email, and tweet in-platform; they’re also offering software development kits for anyone wishing to build their tech into other apps.) "Our premise is to be able to do the whole typing experience without requiring accuracy," says Eleftheriou. "And it takes a little faith sometimes for users to realize that they can trust the system, that they don’t have to concentrate on being accurate. The more you think about it, that’s the way it should be."
This piece is part of Change Generation, our series on young, change-making entrepreneurs. Read the rest here.