It may sound like something out of a 1960s sci-fi movie, but Qualcomm’s ‘brain chips’ are due to go commercial as early as next year, says the firm.
It has been working on the Zeroth neuromorphic chip platform, which claims to mimic the way the brain works, and could be a significant new source of revenues and IPR power if its neural networking techniques really can solve some of the challenges facing robotics.
The aim is to create a widely applicable ‘neural processing unit’ (NPU) to perform the core functions to support activities such as deep learning, in the same way that a GPU supports key underlying functions to drive graphics. It feels like a long term bet ‘ and an area where Intel is also active ‘ but the pay-off could be huge if neural networking, after so many false dawns, becomes mainstream in the internet of things, and if Qualcomm establishes the leading platform to deliver it.
Those are two big ‘ifs’, but Qualcomm has created the base platform and is now signing up research institutes to test the chips ahead of commercial implementation. Zeroth is part of a broad neural networking R&D activity at the San Diego giant, and is named after writer Isaac Asimov’s 4th zeroth law of robotics ‘ that ‘a robot may not harm humanity, or, by inaction, allow humanity to come to harm’.
Its first public demonstration, reported by MIT Technology Review, features a small robot called Pioneer, which rolls up to a Captain America action figure, picks it up with a scoop and pushes it towards three pillars. The robot ‘sees’ a Qualcomm engineer pointing to one of the pillars, and duly sets the action figure down in that place. It then takes a second figure and puts it in the same place, having ‘learned’ the right location. It is also able to distinguish between the dolls and other objects, which it ignores.
The difference between this and object recognition technologies is that the robot also takes action based on what they see. That is a key feature of many neural networking and neuromorphic R&D programs targeting computing devices such as smartphones, as well as robotics. IBM Research and HRL Labs are both separately working on neuromorphic chips for DARPA, the US Defense Advanced Research Projects Agency, according to the MIT publication, while the European Union is backing a 10-year Human Brain Project, which aims to make core contributions to neuroscience, medicine and future computing.
Google has been the most high profile investor in robotics for mobile consumer purposes, with the purchase of Boston Dynamics and others. Its main interest is to drive entirely new methods of web-based search and interaction, based on deep learning, and so to enable new services which will deliver the user data that is Google’s lifeblood. Qualcomm’s objective will be different ‘ to extend its chip technology and patents into areas where it can shape future standards.
While there are many futuristic applications based around robots, in the nearer term, Zeroth could be used in an area where the company has already been highly active, medical and fitness monitoring ‘ not only tracking a person’s vital signs through a wearable gadget, but taking action in response, such as adjusting medication.
Perhaps the most important short term use will be to contribute to the ongoing battle to make the search experience ever smarter. Apple Siri and Google Now rely on voice recognition and artificial intelligence to provide new levels of smart response to user queries, and will be further enhanced by gesture recognition, rich machine learning platforms and, perhaps, by Qualcomm’s NPUs, embedded in handsets. “If you and your device can perceive the environment in the same way, your device will be better able to understand your intentions and anticipate your needs,” Samir Kumar, a business development director at Qualcomm’s lab, told MIT Technology Review.
But these technologies have a long way to go to become mainstream ‘ they need to be affordable, have compelling and easy-to-use applications, and to be easily programmable. CTO Matthew Grob commented: “We’re blurring the boundary between silicon and biological systems.’ This was a claim being made by neural computing developers 25 years ago, and their impact, outside a few esoteric academic or defense projects, was minimal. Blurring those lines is clever, and even a bit scary, but it needs a real business case just like any other technology.