Kishore Chakravadhanula, Qualcomm’s product manager for robotics, believes 5G is the connective fabric that’s going to radically change the landscape for AI at the edge in the coming years. Addressing the audience at Transform 2019 this month, he said it will become as ubiquitous as electricity and Wi-Fi and will bring about massive change.
“5G is unique in that it’s been re-architected from a completely different perspective than regular technologies,” he said. “It’s built for massive data, ultra-reliability, and low latency. It will enable new use cases and applications that we’ve not thought about so far. This will enable new experiences for the end user.”
To understand how 5G will change things, it’s important to first consider the challenges in AI at the edge (including the edge cloud). As billions of devices send more and more data, processing all this data in a timely fashion is going to be a massive challenge on the cloud side. Secondly, everything must be done securely while maintaining privacy. The final challenge is achieving all of this in real time.
“That is the biggest hurdle that most of these devices have to clear,” said Chakravadhanula. “The latencies are sometimes not acceptable, going to the cloud and coming back. Here is where on-device processing, on-device sensing, on-device decision-making, and maintaining security is going to change how things will evolve in the coming years.”
But the unique requirements of edge devices present their own challenges. Because they operate in a very constrained environment, they have low memory or storage. “They don’t have too much processing power,” he said. “Battery life is always an issue. Finally […] heat dissipation, especially in small form factor devices, is a challenge.”
Alongside these issues are the requirements for AI. As Chakravadhanula explained, these are high-compute, multiple concurrency models that happen to run at the same time. They also need to run in real time and need to be always on.
“How do we marry or intersect these two diverging requirements at the edge?” Chakravadhanula asked. He went on to explain how 5G will be a game-changer in solving for these issues — particularly the challenges of massive data, low latency, and better experiences. In turn, this opens up massive opportunity for IoT — the ultra-reliability and low latency that enables machines, manufacturing, and more to benefit from AI without challenges.
“5G enables low-power wide area networks by switching down to narrowband IoT and LTE that is available today, but it’s derivative in the next generation,” said Chakravadhanula. “I’m from robotics, [and] nowhere else is this convergence of AI and 5G happening more, as we see all these kinds of use cases where this takes off.”