What if computers could tell the difference between a smile and a smirk?
Well, it's not a "what if...?" question anymore. Some computers can. At least at Affectiva, a company based in Downtown Crossing that focuses on Emotion AI.
Founded in 2009 as a spinoff of MIT Media Lab, Affectiva raised $14 million in May, 2016, bringing its total funding to $33.72 million, according to Crunchbase. Investors are Fenox Venture Capital, Horizons Ventures, Kleiner Perkins Caufield & Byers and Myrian Capital, among others.
Affectiva has built a database with more than 5.3 million of videos of people showing different emotions in the real world. Videos were collected around the globe with the consent of all the people involved, the company said. The subsequent database has been used to train computer algorithms to recognize human expressions and matching them with a specific emotion. This is what Affectiva scientists call Emotion AI, which is short for "Artificial Emotion Intelligence".
Watch the video below to see us demo Affectiva's emotion recognizing technology.
Affectiva let us know that the easiest expression for computers to detect is the smile. However, things can become tricky if we ask computers to detect what kind of smile we're dealing with. Is it a genuine smile? Or a polite smile? Or a "I-can't-wait-to-get-out-of-here" smile? Algorithms still struggle in telling the difference, as many people do.
As for the hardest emotion for Affectiva’s tech to detect, the company’s Director of Applied AI, Jay Turcot, said that it's fear, because it happens rarely in the kinds of data that’s been collected to date.
Analysis of the Affectiva database led to some interesting discoveries. For example, women are usually more expressive, as they smile more often than men. On the contrary, men tend to show more negative emotions. Surprisingly, people over 50 are 25% more expressive, as they show emotions more intensely and more frequently than younger people.
As of today, Affectiva has been awarded 10 patents.
This is the third story in our AI in Boston series, which is running several stories on artificial intelligence the week of May 15.
