The more you read about AI advances, achievements that bring this “unnatural” intelligence closer to humans, the more you want to close your eyes and escape this scary world. With Stephen Hawking-like warnings and jobs in peril, your fears are normal… but are they grounded? We met Dennis Laudick, the VP of Marketing at ARM, leader in mobile processors, to talk about their take on AI and if machine learning can do more harm than good #mwc18 #machinemagic
ARM is renowned for their CPUs and GPUs, the vast majority being enabled on mobile devices now. They’re leading in this part of the market but as AI is shaping up to be the future of smart devices, the company was bound to look into it as well. When it comes to a company of this level, looking into it means deploying a three-component project, Trillium.
We talked about the project here – what does it do, how it will be implemented, what will developers and OEMs be able to do with it. Catching Laudick at MWC 2018 was a prime opportunity to pick his brain and find out the thinking process behind it:
“[The] machine learning processor, this one we’ve designed from the ground-up specifically to support [this] type of processing and make sure it can push forward the capabilities of the devices in a power-efficient way. […] In addition to that, we’ve seen that […] when you’ve got a camera looking in space and people are in the scene, a lot of times that’s when the rest of the system wakes up. We’ve developed a processor specifically designed to achieve that in the most optimal way possible.”
A smartphone with machine learning can decode the world for you
So, what does the third piece do? “Software is very, very important. So, as part of Project Trillium, we’ve announced and released a layer of software which makes it very easy for people to get from the common APIs and frameworks they’re used to down to the ARM hardware in the most efficient way possible”.
It’s a complete suite, as you see, a technology impressive enough to entitle us to ask Laudick exactly how will ARM’s efforts help the end user. How helpful is machine learning for us? Could it be more of an assistant than a threat?
Laudick definitely thinks so: “It can tell you more about what’s happening. One analogy is about a colleague, who is a diver. He says that one of the challenges he has is that when he’s underwater he sees some amazing things, some scary things and some he can’t understand. With machine learning, […] something small as a smartphone could decode all of that for him”
We got the sense that while sci-fi movies depict AI as a superior intelligence that is constantly trying to subjugate humans, in reality, in the tech world, AI is an auxiliary brain, trying to satisfy your needs before you pronounce them. Predictive learning is key in this equation, an idea Marcel Campos from Asus also talked about here.
“Predictive learning is trying to understand what a pattern is, trying to recognize what you do – the way you walk, you move around, how you spend your money”, added Dennis Laudick. He strongly believes neural networks are ways to approach problems today and get reliable experiences.
The trend (and challenge for many companies, still) is to do all of that in real time, by moving machine learning to the edge and out of data centers. We’ll see how will that evolve and at what pace in the years to come!