"You must be nuts!"

Why I'm building Language Engine

posted by Darryl on 23 Jan 2015

Heh, yeah, maybe. You've got to be a little nuts to start a startup, and probably quite nuts to start one with such audacious goals. But they're important goals.

Perhaps a little history would put it in context. I've been interested in AI for a long time, since I was a kid, and as I learned more about linguistics, I decided to play around with the idea of building systems that could interact with a user. After playing with some fairly simple ideas in 2009, I put it aside and just continued my studies. The speech recognition side just wasn't where it needed to be yet.

Early last year, however, I found myself using Twitter a lot while driving. (DON'T Tweet and drive!) I'd wait till I got to a red light or a stop sign, read some tweets, reply with speech-to-text, etc. It was around this time that I had seen Her (go watch it if you haven't seen it). I thought to myself, "jeez, can't I just talk to my phone?"

  

Americans lose enormous amounts of time and money driving. $161 billion a year! If you've got a chauffeur, or at least someone sitting next to you, you can recover some of it, but most people don't. What could we be doing with that time, that money?

So I tried Siri. Maybe Apple was offering a solution and I had simply over looked it. That didn't go as expected, and even today it doesn't. I can tell Siri to send a tweet, but that's about it. I looked around at what else was available and while there were interesting APIs, they couldn't handle the complexities of natural language well enough to build a conversational AI.

I figured, since I know a bunch of linguistics, I know we can make that stuff work, and more. I knew we could make something completely new, that no one's seen before except in scifi. And here I am, almost a year later, trying to make computers think. Nuts. But is it really? We don't need to aim for the Samantha to do something amazing, just Jarvis.

Unlike Samantha, Jarvis isn't really a person with his own aspirations, opinions, and free will. Jarvis is just a really good UI. He knows what to do to help you achieve your goals, as you've conveyed them in talking to him. He'll help you discover a molecule in the Unisphere, he'll catch you when you're a little too drunk to stand. But he won't run off with the cyberghost of Robert Anton Wilson, because he's not a person, just a really great UI.

We could build drones that can fly around at your command. You could have a fleet of drones that monitor crops — "check the southern fields for blight, and if you find any, remove the plants". You could have drones that inspect bridges, with the operator safely on the deck — "ok, go to the next pylon. wait, hold on. what's that?" (camera zooms in). This super cool startup Skydio is aiming in that direction.

It goes beyond that, even. We could build robots that can manufacture, install, and maintain giant solar farms for pennies. The dirt cheap power could be used to purify water in huge quantities, ending drought, greening the deserts, and providing arable farmland to eliminate hunger, which would of course be farmed by drones and robots.

Steve Jurvetson, of Draper Fisher Jurvetson, wants to use new tech to make a world where all material goods are $1 per pound, and thereby eliminate hunger, poverty, lack of health insurance, homelessness. I think that's a great goal. I think we can achieve it if we really try, and AI is the key. I hope Language Engine can be a part of that.

If you have comments or questions, get it touch. I'm @psygnisfive on Twitter, augur on freenode (in #languagengine and #haskell).




Email address not valid.
Sign up!
Follow on Twitter