Last year saw AI chips land inside premium smartphones, as Apple and Huawei sought to beef up tasks like system performance, camera quality and, well… augmented reality emoji.
But could this hardware ever come to budget phones? And when could we expect this to happen? We asked Jem Davies, fellow and general manager of the machine learning group at UK chip kingpin Arm, during our trip to MWC.
No ad to show here.
“Yes [when asked whether there was a market for AI silicon in cheaper phones – ed], so traditionally you would expect a new technology like machine learning or whatever, to appear in super smartphones, premium smartphones, first. And then slowly trickle down over four or five years. That’s actually not what we’re seeing here. What we’re seeing is that China, for example, is looking at rolling out machine learning capabilities, even in entry level phones,” Davies answers.
The Arm representative then points to the digital TV sector as evidence of the massive interest in the tech.
Jem Davies of mobile chip kingpin Arm reckons that today’s machine learning “has the hallmarks of a serious inflection”
“Digital television, traditionally… stayed some years behind the pace — they want ML (machine learning) now. DTV is quite an interesting case, really always a bit behind the pace, but this time they’re going ‘give it to me now. I’ve got a use case now, I need this much processing capability, just give it to me instantly’,” Davies continues.
“And so this has the hallmarks of a serious inflection. I know everyone says ‘the new thing is going to change computing forever’, and I’m careful not to say things like that. But I think it is an inflection, the likes of which you only see quite rarely, perhaps once in a generation.”
Challenges for budget AI hardware
The Arm executive elaborated on challenges faced by manufacturers in bringing AI hardware to cheaper phones, saying silicon space is at a premium on lower end chips. Nevertheless, Davies pointed to the company’s new ML processor tech as being very scaleable “compared to silicon partners who are producing a particular neural network processor”, ostensibly referring to Apple and Huawei.
“Premium guys are saying ‘what can you do in two watts of electrical power?’. The entry level guys are going ‘hm, I can only afford to spend so much silicon’ and we can fit both.”
Does that mean we can expect US$100 smartphones with tiny and/or low-end AI chips then? “That’s what we think,” Davies answers.
Disclosure: Hadlee Simons was a guest of Huawei in Barcelona at Mobile World Congress. Read all our MWC 2018 coverage over here.
