By Davey Alba
Original link: http://www.wired.com/2015/05/nara-logics-ai/
I’m fairly new to San Francisco, so I’m still building my mental database of restaurants I like. But this weekend, I know exactly where I’m heading to for dinner: Nick’s Crispy Tacos. Then, when I get home, I’m kicking back to a documentary I’ve never heard of, a Mongolian drama called The Cave of the Yellow Dog.
An artificially intelligent algorithm told me I’d enjoy both these things. I’d like the restaurant, the machine told me, because I prefer Mexican food and wine bars “with a casual atmosphere,” and the movie because “drama movies are in my digital DNA.” Besides, the title shows up around the web next to Boyhood, another film I like.Nara Logics, the company behind this algorithm, is the brainchild (pun intended) of its CTO and cofounder, Nathan Wilson, a former research scientist at MIT who holds a doctorate in brain and cognitive science. Wilson spent his academic career and early professional life immersed in studying neural networks—software that mimics how a human mind thinks and makes connections. Nara Logics’ brain-like platform, under development for the past five years, is the product of all that thinking..
'Nara is AI for the people.' Nathan Wilson, Nara Logics
The Cambridge, Massachusetts-based company includes on its board such bigwig neuroscientists as Sebastian Seung from Princeton, Mriganka Sur from MIT, and Emily Hueske of Harvard’s Center for Brain and Science.
So what does all that neuroscience brain power have to offer the tech world, when so many Internet giants—from Google and Facebook to Microsoft and Baidu—already have specialized internal teams looking to push the boundaries of artificial intelligence? These behemoths use AI to bolster their online services, everything from on-the-fly translations to image recognition services. But to hear Wilson tell it, all that in-house work still leaves a large gap—namely, all the businesses and people who could benefit from access to an artificial brain but can’t build it themselves. “We’re building a pipeline, and taking insights out of the lab to intelligent, applied use cases,” Wilson tells WIRED. “Nara is AI for the people.”
Problems Worth Solving
Wilson is not alone in his populist ambition. The list of other companies that have been created to make high-level artificial intelligence more accessible to the broader universe is not short. But so many of these have also been acquired by the aforementioned giants of tech—in fact, such acquisitions often seem to be a key reason behind their existences in the first place.Nara, on the other hand, exhibits some actual consumer-facing friendliness. The site where I got my recommendations, Nara.me, is available for anyone to sign up to get intelligent recommendations for movies, restaurants, and hotels based on your expressed preferences. There are also some filters available, so you can tell the software what you’re in the mood for at any point in time—maybe Chinese food and a comedy—so it can adjust to your changing impulses. Once you’ve fed Nara a few of your likes, it can learn your tastes and work across the entire system. Say you’ve moved from New York to San Francisco. Tell the app a few of your favorite restaurants in Manhattan, and it’ll surface similar options in your new home city. There’s even a mobile app that can detect your location and serve up suggestions that are close by.
But Nara says it’s not ultimately focused on delivering a polished consumer product. Wilson and his team want to focus on enhancing the AI, he explains, not on acquiring users. There are businesses that already have a plethora of users. For Nara, it’s those established businesses whose existing problems are most worth solving.
To that end, Nara customizes specific solutions for companies across different industries. The startup says it can’t reveal which companies it works with, but CEO Jana Eggers, the one-time head of Intuit’s innovation lab, says Nara recently signed a global bank to use its tech for a variety of services: managing the institution’s reward points and recognizing customers’ preferences; assessing loan approvals based on detailed financial history; and analyzing user transactions for the bank in real-time to detect fraud. Nara also has a healthcare services company on its client roster that it assists with doctor-patient matching based on a patient’s medical history and a doctor’s academic credentials. For a major airline, Eggers says Nara is tweaking its seating system so that some empty spots can be allocated to passengers who have had an unsatisfactory experience in the past.
Thoughtful Tech
The key to Nara’s technology is personalization, Wilson says. Nara is essentially a matchmaking system that finds and understands entities in any data set, from people and places to businesses and abstract concepts, then builds a massive knowledge graph that shows weighted links between those entities. Wilson says Nara inserts users right into that knowledge graph to offer personalized recommendations. Knowing a bit about the user is what allows Nara to light up other things they might like. And the system can scrape public databases to enhance its knowledge.Richard Socher, a former natural language processing researcher at Stanford University and current CTO of artificial intelligence startup MetaMind, says research shows this approach to AI works well for recommendation engines. Traditional approaches typically take just probability and user history into account, he says. If three people like Product A, and two of those folks also like Product B, there’s a high chance the third person will also like Product B. Nara’s approach could create more powerful nodes and more easily take external information into account, Socher says. “This could be a useful service to help other companies that can’t build their own recommendations system,” he says.
In the meantime, despite the aggressive efforts within bigger, richer companies to forge smarter AIs, Wilson says he’s not worried about possible competition. “In the scheme of things, working on artificial intelligence is not a numbers game,” Wilson says. “It’s more about careful thought with a small group of people, and technology that works thoughtfully.”
