Despite advancements in compute power and data storage over the past couple of decades, artificial intelligence (AI) and natural speech remain too hardware-intensive to be deployed on consumer robots. Apple's implementation of Siri may have inspired a new generation of AI engineers to change their approach. It's over to the cloud.
"On-board computation reduces mobility and increases cost." said Dr Heico Sandee, Robo Earth program manager at the Dutch University of Technology in Eindhoven in a statement. As wireless data speeds increase more and more robotic thinking could be offloaded to the web, he said.
Rapyuta, a cloud-based AI for consumer robots just went online. Its development began in 2011 by a group of scientists working for the Robo Earth Project. Rapyuta works much like Siri. Sensory inputs (tactile, visual, voice) are sent from robots to Rapyuta's server clusters, which process it, and send the robot their appropriate responses to the stimuli. There could be a slight latency involved, but the net result is a more intelligent, inexpensive, and maneuverable robot. The approach is both similar and dissimilar to the Borg Hive Mind from Star Trek. It's similar in that there is little or no individuality left to the drones, but different in that the drones don't contribute to the AI, apart from raw data that can be used to improve it.
Pushing the AI completely to the cloud has another advantage. The AI is monolithic, and can be controlled at a single point. The drone relies on its owner's Internet connection. It remains to be seen how idiot- and bullet-proof they make Rapyuta, because that is what could stop it from becoming Skynet.