EmoSPARK: An "artificial intelligence console" that wants to make you happy
By Nick Lavars
January 2, 2014
For as long as we’ve been imagining emotionally intelligent machines, we have pictured something at least mildly resembling the human form. From George Lucas’ C-3PO to the recently-developed Robokind Zeno R25, our vision for robotic companionship has typically involved two arms and two legs. Taking a different approach is inventor of the EmoSpark console Patrick Levy Rosenthal, who aims to bring artificial intelligence to consumers in the form of a cube small enough to fit in the palm of your hand.
The EmoSpark console is a 90 x 90 x 90 mm (3.5 x 3.5 x 3.5 in) Wi-Fi and Bluetooth enabled cube that interacts with a user’s emotions using a combination of content analysis and face-tracking software. In addition to distinguishing between each member of the household, the device uses custom developed technology that Rosenthal says enables it to differentiate between basic human feelings and create emotion profiles of not just everybody it interacts with, but also itself.
“While the technology behind face-tracking is well established, what we've done differently is use it to track and process different emotions," Rosenthal tells Gizmag. "The EmoSpark Cube contains a unique chip invented by myself called the Emotional Processing Unit. This allows the cube to build up its own Emotional Profile Graph (EPG) as it interacts with its users. The cube saves all this information and, just like a fingerprint, will over time will keep an emotional print of each family member with which it interacts.”
Users communicate with the cube by either typing or talking to it through their television, or remotely via a smartphone, tablet or computer. By analyzing this data and using its face-tracking technology, the cube is designed to acquaint itself with the user over time by gauging their likes, dislikes and different moods based on eight primary human emotions: joy, sadness, trust, disgust, fear, anger, surprise and anticipation.
Initially, the cube works to improve your mood and overall happiness by connecting to and recommending particular songs and videos or content on sites such as Facebook and YouTube. As the relationship between the cube and user develops, the device becomes more skilled in the art of conversation and nuanced in its offers of comfort – something Rosenthal considers a significant mark of progress in artificial intelligence and integral to the technology.
“The major breakthrough was in developing a credible model to synthesize emotions in a machine and creating a machine that can reply to a question not based on a script, but on a system compatible with the human emotional spectrum,” says Rosenthal. “A system that will be able to reply to a free association test, not only based on logic, but also based on its emotional status at the time you ask it a question.”
This means that over time the cube will develop a personality of its own, the rate of which is largely determined by how often the user engages with it. “The emotional learning will never end, the cube will always learn and its EPG will change over time but it’s logarithmic,” said Rosenthal. “It will learn much more when it is young and developing, I would say it depends more on the frequency of use than time.”
While confident he has created a foundation for the assimilation of artificially intelligent machines into the consumer space, Rosenthal hopes to harness a keen general interest in artificial intelligence by handing control over to developers. “The cube will have open API (Application Programming Interface) to allow developers to create new blocks of technologies in the form of apps in Google Play store,” said Rosenthal. “So the conversational engine, voice and speech recognition are all modules that will be upgraded or will be replaced, so the user can make their own cube.”
The EmoSpark cube also doubles as an e-learning tool. It comes connected to Freebase, a collection of online knowledge owned by Google, which Rosenthal says enables it to answer questions on over 39 million topics. It can also be used to control robotic devices, bringing emotional feedback capabilities to a NAO robot or turning a Sphero ball into a virtual pet with its own emotions, for example.
Android powered, the cube contains 1.8 GHz CPU along with 2 GB of DDR3 memory and Rosenthal’s custom-built 20 MHz EPU (Emotion Processing Unit). It has an internal antenna, built-in Wi-Fi 802.11b/g/n capability and features USB 2.0, MicroUSB and HDMI 1.4 ports.
Citing an enduring trepidation held towards machines that replicate our shape and movements, Rosenthal said steering the design of the EmoSpark away from the traditional humanoid robot blueprint was a conscious and strategic decision. “I don’t want a fake human,” he says. “When we look at robots like that we see a bunch of silicon. The danger is machines that lie and tell us that they are human. That is why we made a cube, the cube has laws that ensure it respects human desires and happiness and knows that it is a machine.”
Aesthetics aside, Rosenthal believes that central to the success of consumer artificial intelligence is an ability to communicate with the machines in way that involves emotion, as it may serve to alleviate a wide-held apprehension toward the potential capabilities of intelligent machines.
“As the role of machines in society and the workforce continues to grow, it is important that humans and machines can connect on an emotional level,” he said. “It is about trying to show how this technology can be used to improve lives and trying to change the future while we can.”
The crowdfunding campaign for the EmoSpark cube launches on Indiegogo today. Early adopter pledges are available for US$224, with shipping estimated for May 2014 if all goes to plan. You can see the EmoSpark cube in use and hear from Rosenthal in the video below.