How to talk to computers - Meet Wink

With the rise of new technologies like Conversational Interfaces and Artificial Intelligence, new and more human ways of communicating with technology are becoming possible, and necessary. Wink is an experiment with giving a computer the ability for reciprocal perception; you can see that the computer can see you.

Asking someone in the same room to perform a certain task, like switching off a light, comes natural to us. When there are only 2 people in the room, the person knows intuitively that he is being addressed. When there are more people your gaze will show who you are asking for help. With voice interfaces like Amazon Alexa, Apple Siri or Google Assistant things are not so clear. These services are unlocked by a keyword or phrase. With more functionality being developed for these services daily, and the smart home becoming more capable, a better more human system should be developed.

Wink is one of the ways this problem could be addressed. Applying 'signs of life' to a small robot that looks like an eye, allows the user to have a point to direct himself to to talk to the smart home. Wink can show that he sees the user by looking back at the user, completing the communication loop. In the movie below you can see how Wink moves and shows the user that he is aware of what is happening around him:

Of course the above video is an exaggeration of the current state of technology. Current AI systems allow for conversational interfaces, but a real-time one-on-one conversation with a robot, including body language is not quite at this level yet. The closest we can get to this currently available on the market is Anki's Cozmo. Cozmo is a kids toy that uses Pixar level character design, with state of the art programming to interact with its user. It can recognise its owner, and follow their face around a room.

The ability for a robot to express itself through 'body language' allows it to communicate with people at their level. In the next video Wink used Computer Vision to track the users body language, and copy it. The translation of the users body language into that of the robot is almost imperceptible by the user, since the motion comes across as if the robot has intention and awareness. You can easily imagine giving a command to Wink, and you would not doubt that the system picked it up.

Special thanks go to Alan Nguyen who helped build the original Wink