People tend to be a little creeped out by a machine pretending to be human, as evidenced by the almost universal uncanny valley response to humanoid robots. However, that doesn’t mean it is a bad idea for technology to be a little more human in the way it responds to us. The EmoSpark is an attempt to make artificial intelligence more friendly by packing the AI into a small cube that can recognize and react to your emotional state.
The EmoSpark console is only 3.5-inches on a side and fits in the palm of your hand. It is being designed by inventor Patrick Rosenthal who wants to use the EmoSpark to recognize not only people, but their emotions in real time. The cube does this with traditional face-tracking technology and a content analysis engine developed by Rosenthal. Theoretically, by glancing at you, the EmoSpark will be capable of differentiating between basic human emotions.
Our brains are hardwired to recognize emotions in others — the tiniest cues can tip us off that someone is angry or upset. For a machine, though, it’s a herculean effort to extract a signal from the noise of human expressions. Rosenthal says the key to allowing the EmoSpark to recognise emotion is a dedicated chip called the Emotion Processing Unit (EPU). Users interact with the cube through voice or text via a smartphone, tablet, or computer. This allows the EmoSpark to build a so-called Emotional Profile Graph (EPG) of itself — each EmoSpark becomes unique. Combine this with face tracking, and the device can get a sense of a person’s mood.
Assuming the EmoSpark is able to figure out your mood, what can it do with that information? At first, not very much. As it is starting to get to know a person, the AI is able to recommend various songs, videos, and other content on sites like YouTube and Facebook. As the cube’s EPG gets more advanced, it will be better able to respond verbally to people it recognizes. This is essentially a machine that learns empathy. The EmoSpark will have an API that allows app developers to plug into the EPG. This will graft new abilities onto the EmoSpark, giving it new ways to interact with its human friends.
The EPU that powers the facial tracking and empathy business is a custom-built 20MHz chip, but that’s not the only hardware at work. The EmoSpark also contains a 1.8GHz CPU, 2GB of RAM, and an internal WiFi antenna. The interface shown on your screen while the cube is learning about you is running on top of Android, which should make integrating EmoSpark easier for developers.
EmoSpark is currently on Indiegogo with almost two months to go. The creator is seeking $ 100,000 to build the device, but it’s a flexible funding campaign, meaning everything pledged will be handed over even if the goal isn’t reached. The $ 224 early bird deal gets you one EmoSpark to be delivered in May 2014.