This post is also available in: עברית (Hebrew)
Machine learning could soon be coming to mobile devices, if recently announced partnership between Google and chip manufacturer Movidius bears fruit. Machine learning is a branch of artificial intelligence that focuses on giving computers the capability to recognise patterns in data based on rules learnt from previous sets of data. Think IBM’s Watson – the computer that won American game show Jeopardy. It can now diagnose cancer patients and has even created some not-bad-at-all recipes. This is also the technique that allows Facebook to recognise your face in photos.
Most of these applications require a constant internet connection to send queries to server farms that perform the processing. This can become a restricting factor for people with limited data plans or in areas where there is no connectivity at all. Movidius’ chips, however, need very little power to run and can perform complex computations right in the device – no server needed. This could help bring independent AI to smartphones, drones, and even robots.
“Instead of us adapting to computers and having to learn their language, computers are becoming more and more intelligent in the sense that they adapt to us,” said Blaise Aguera y Arcas, head of machine intelligence research at Google, in a video jointly released by the two companies.
Google is quite secretive about some of its plans, and what it intends to do with Movidius’ chips is no exception. But judging from recent developments, we could be seeing some form of machine vision in Google’s Android devices in the very near future. Phones in the next few years could not only see the world, but make sense of it as well.
Additionally, obstacle avoidance is one of the biggest challenges for commercial autonomous drones (like Amazon’s drone delivery scheme). If Google’s partnership with Movidius succeeds, it would help autonomous drones finally gain traction in the real world.