News

Qualcomm announces AI research division, wants to make AI pervasive

Qualcomm has announced it’s starting an artificial intelligence research division.

Dubbed AI Research, the division will be cooperating with AI research occurring across the company.

The company began exploring AI research over ten years ago. It started with investigating spiking neuron networks for computer vision and motion control applications.

Currently AI Research works on power efficient AI, personalization and data-efficient learning.

The goal is to advance AI to make is core capabilities more pervasive.

“We are constantly pushing the boundaries of what’s possible with AI,” said Jim Thompson, Qualcomm’s executive vice president and chief technology officer. “At Qualcomm AI Research, we are dedicated to bringing together some of the best minds in the field to shape the roadmap of AI hardware and software platforms. Our goal is to make on-device AI ubiquitous.”

Considering Qualcomm’s role in the communications business and in mobile devices, it definitely has the power to make a significant impact by integrating AI into its products.

For example, many AI features available on smartphones are done remotely. Take Google, which has developed tensor processing units (TPUs), a type of application-specific integrated circuit (ASIC) specifically for machine learning. By leveraging cloud computing and its TPUs, Google was able to launch features like Google Lens

Efficiency and power are some of the main reasons these machine learning applications are handled remotely. While we can’t expect TPU level performance on smartphones anytime soon, more efficient mobile chips developed with AI in mind could be a game changer. On-device machine learning could be an area that would benefit from this.

AI could also be massively beneficial in network communications as well. Using AI to intelligently reroute internet traffic around service outages could be one application. Something like that could be a major part of Facebook’s Terragraph project that Qualcomm has partnered on.

Source: Qualcomm

Comments