Feeling sad? Soon your dolls will be able to tell. To demonstrate the power of a new chip that can run artificially intelligent algorithms, researchers have put it in a doll and programmed it to recognise emotions in facial images captured by a small camera.
The doll can recognise eight emotions in total, including surprise and happiness, all while running on a small battery and without doing any processing in the cloud. The total cost of putting the new chip together is just 鈧�115 鈥� an indicator of how easy it is becoming to give devices basic AI abilities.
鈥淚n the near future, we will see a myriad of eyes everywhere that will not just be watching us, but trying to help us,鈥� says project leader at the University of Castilla-La Mancha in Ciudad Real, Spain.
Advertisement
Recent advances in AI mean we already have algorithms that can recognise objects, lip-read, make basic decisions and more. It鈥檚 only a matter of time before these abilities make their way on to little cheap chips like this one, and then put into consumer devices.
鈥淲e will have wearable devices, toys, drones, small robots, and things we can鈥檛 even imagine yet that will all have basic artificial intelligence,鈥� says Deniz.
One of the advantages of Deniz鈥檚 chip not needing the internet to function is privacy. Last year, a smart doll called My Friend Cayla attracted a lot of controversy because it couldn鈥檛 do its processing locally. To recognise what children were saying, it sent audio clips to the cloud and then worked out an appropriate response. 鈥淐an I tell you a secret,鈥� a child might ask. 鈥淪ure go ahead,鈥� the doll would reply.
Children could share intimate details about their lives with their new friend only for that conversation to be recorded and sent to a data centre. Clearly, that鈥檚 not how many parents want smart toys. Privacy advocates also raised concerns over the Hello Barbie doll, which used speech recognition to interact with children鈥檚 requests 鈥� but also passed the data to third-party servers for storage and processing.
At the moment, Deniz鈥檚 project is focused on processing data from a camera rather than a microphone, but many of the issues are the same.
Running computer vision algorithms locally is also incredibly important in many situations, says at Boston University. A self-driving car moving at 120 kilometres per hour doesn鈥檛 have time to do all of its decision-making in the cloud because of a lack of bandwidth. So those decisions need to be made locally.
鈥淭oday we live in a world where devices are dumb. Tomorrow we will live in a world where devices can think,鈥� says Versace. Moving away from dependence on the cloud is a vital step for AI, he says.
Sensors
