Our phones are increasingly capable of doing amazing things, from controlling the appliances in our homes to bringing a museum’s animal specimens to life. A crop of startups out of NYU Tandon’s artificial-intelligence incubator seek to add even more functions to that list: say, figuring out our diet plan simply by viewing a photo of our plate.
Last Tuesday, the school held the second edition of its Future Labs AI Summit. The event served as both a forum on the future of AI and a showcase for the school’s AI Nexus Lab, the artificial-intelligence-focused incubator it launched last year in partnership with ff Venture Capital.
We caught each of the four presentations from the AI Nexus Lab cohort. Here’s what we saw.
A receptionist on the clock 24/7
Ron Fisher, the CEO of Bowtie, laid out a simple yet significant problem for many businesses. They, according to him, miss some 25 percent of calls from potential customers. Why? They don’t have a receptionist on duty at that moment. Those missed calls, Fisher said, represent a lost opportunity: most of those customers won’t call back.
That’s where Bowtie, developed by a group of Cornell Tech alumni, including Fisher, intends to come in. The company has developed a bot that sends text messages to customers whose calls go unanswered. It can set up appointments, offer customer support and sell upgraded services. Right now, Bowtie is focusing on beauty, fitness and wellness businesses, especially those with one to four locations, Fisher told Technical.ly.
Bowtie currently has some 35 beta customers. The service costs $129 a month, or $99 a month with a yearlong subscription. Strikingly, the company does not offer a free trial. That’s by design, Fisher said, so that its customers are invested in using the service. Their feedback is critical to Bowtie as it refines its product, which uses a proprietary natural language processing model.
“We could not use out-of-the-box models because they did not fit our needs,” Fisher said.
So far, Bowtie has yielded some impressive results for its customers. Ten percent of new customers to Deify, a laser hair removal salon, have come through the service. And Wah London Nails, its sole international customer, now generates 30 percent of its bookings sales through Bowtie, Michael Wang, the company’s chief technology officer, told Technical.ly.
A voice-activated search engine for business
“No matter what industry we’re in, there’s always a knowledge gap,” Kul Singh, the CEO of SecondMind, told the audience.
Indeed, online searches are routine for most office workers. According to Singh’s presentation, those searches eat up two-and-a-half hours —a whopping 30 percent — of the workday. SecondMind is aimed at giving workers back much of that time, namely during calls with coworkers and customers.
SecondMind’s platform uses voice recognition and natural language processing to bring up relevant information in real time during phone calls. For instance, it can instantly bring up data related to a particular client during sales calls. It integrates with other common software services, such as Salesforce and Google Apps, in order to call up relevant information stored inside companies’ computer systems. In that sense, it’s similar to chat platforms designed for business collaboration. Singh, in fact, dubbed SecondMind the “Slack for voice conversations.”
In its testing so far, SecondMind has reduced companies’ average call length by 80 to 90 percent. That efficiency, Singh said, illustrates what voice-activation services can bring to the business world. Indeed, as he pointed out, consumer voice-activated devices such as Amazon’s Echo have opened up a wide array of new product applications. In the future, Singh added, he aims to apply SecondMind to in-person conversations by integrating another hot technology: augmented reality.
A quiz generator that makes teachers’ lives easier
As any student knows, most classes don’t simply follow one textbook. There’s usually plenty of supplemental reading involved, much of it drawn from online sources. Gathering that extra material can be an expensive proposition: altogether, U.S. teachers spend $1.6 billion in out-of-pocket costs on such content. Plus, that online content often requires teachers to develop new assessments for that material. Even if it comes with built-in quizzes and tests, they usually have to be graded manually.
It’s an oddly analog process in our digital world, said James Villarrubia, the CEO of Mt. Cleverest.
“The classroom experience has not changed,” he said.
Through its platform, Mt. Cleverest seeks to streamline teachers’ out-of-pocket costs and their workloads. The company has developed a search engine that enables teachers to find high-quality content for free. It also has the capability to generate quizzes from any text source, whether found on its engine or uploaded by the teacher. Artificial intelligence also kicks in as students take the quiz: the platform continually refines the quiz based upon students’ answers and ranks questions based upon their effectiveness.
Right now, Mt. Cleverest is offering its platform to teachers for free, in order to gather training data. Villarrubia, who lectures on cybersecurity at the University of Virginia, is using the platform in his own class, he told Technical.ly. In the future, teachers can pay a subscription fee to access analytics related to their content and quizzes.
A visual food-logging app
Keeping a diary of each meal you eat is a tested strategy for losing weight, and there are several apps devoted to that purpose. But entering all that information can be a pain. What if you could simply snap a photo of your plate and capture the nutritional information related to that meal?
It may sound farfetched, but that’s the app that Vinay Anantharaman and Michal Wolski, the cofounders of Bite.ai, have developed. Anantharaman and Wolski previously worked at Clarifai, an alum of the NYU Data Future Lab, which runs an image recognition platform. They’ve applied that experience to their app, Bitesnap, which is available for iOS and Android.
Here’s how the app works: take a photo of your meal, and Bitesnap will identify its components. Enter the portion size for each component, and the app will automatically return its nutritional information. Anantharaman gamely demonstrated the app during the lunch break at the AI Summit; his company’s booth included plates of cantaloupe and blueberries for attendees to snap. Technical.ly tested out the app, and it worked quite well — though we’d guess discrete pieces of fruit are relatively easy to identify.
According to Anantharaman, however, the app is designed to recognize photos whose contents aren’t so clear. As he pointed out, few people are cooking Instagram-friendly meals on a regular basis. The app is also designed to hone its predictions based on a user’s profile. If you’re vegetarian, for instance, it will make the assumption that your burger is more likely to be tempeh than beef.
“We’re actually combining who you are with what you’re eating,” Anantharaman said.
Beyond the obvious consumer application, the founders of Bite.ai are also hoping their service can be of use to health professionals. The company is launching an API for this purpose, and it’s also involved in two programs: a pilot with 500 nutritionists, and a clinical study with dialysis and kidney transplant patients.-30-