Two years after setting up an artificial intelligence research laboratory with the University of Amsterdam, Qualcomm Technology has acquired one of its a spinoffs – an outfit called “Scyfer”.
The acqui-hire brings the university’s Dr Max Welling into the Qualcomm fold. Dr Welling is a research chair in Machine Learning at the university, and also holds a secondary appointment at the Canadian Institute for Advanced Research (CIFAR).
The Qualcomm Technologies-University of Amsterdam joint research lab, QUVA, is co-directed by Dr Welling (along with Arnold Smeulders), with a focus on applying machine learning to computer vision.
The acquisition brings new software under Qualcomm’s umbrella – a move in line with the company’s strategy to use more general-purpose silicon for AI tasks.
Last week, we noted Qualcomm pivoting away from the original aim of its Zeroth project, an effort to build Neural Processing Unit (NPU) that would work alongside Snapdragon SoCs.
The change of heart was first signalled in July, when the company released decided to make its neural network software free to developers.
Qualcomm has previously identified industrial IoT, mobile applications (for example Facebook’s Caffe-2) and vehicles as targets for its AI work.
Scyfer seems to offer a broader approach to AI and already boasts of applications that include quality inspection in the steelmaking business, health, traffic, sound analysis and financial projection.
But it’s probably the software that matters most to Qualcomm. Anyone wandering into cloud-based AI sees an array of flags belonging to giants like Google and IBM.
None of that works without a connection, and coincidentally, an AI in the cloud doesn’t sell Snapdragon chips. So Qualcomm’s strategy is to try and apply enough smarts to the software that it can live on constrained devices and function even when the host is offline. ®