| AI + T |
Up until the late 80s and early 90s hardware was King: IBM, Intel, Texas Instruments were market leaders. Devices from Japanese corporations like Sony, Toshiba, and Panasonic were everywhere. But the use cases of most of their devices didn’t evolve significantly as the tastes of consumers around the world shifted to software and mobile devices. The shift to software, when it came, favored companies like Microsoft, Oracle, and SAP. The latter then missed the shift to mobile, opening the door for companies like Apple, Google, and Samsung. But none of the above-mentioned companies is planning on missing the coming, and increasingly obvious shift to “edge” computing, where virtual assistants and other AI applications permeate all physical devices that can hold them. This all means that hardware is making a comeback, riding on the tailcoats of AI. Software may still be eating the world, but as Nvidia’s chief Jensen Huang said in a 2017 interview, AI will eat software.
And AI is starving.
Artificial Intelligence has been undergoing an awakening, no longer an esoteric or emerging trend, but a growing reality. Many of the actions being carried out by humans today; like driving, packaging, accounting, and maybe even cooking will be automated sooner or later. Actions we carry out by typing into smartphones or TV remote controls will give way to speaking to virtual assistants like Siri, Alexa, and Google, who will live in devices everywhere, a concept called edge or “ambient computing”.
AI is not a new concept, nor are the math fundamentals underpinning it. But for AI to “eat software” and become truly useful it needs hardware: processors and chips fast and strong enough to gather, crunch, and act on the vast amounts of data our world is creating. Those hardware components are finally here, catching up to the software that’s been growing in things like big datasets and neural networks.
The growing need for dedicated hardware to fuel the services and products companies want to offer users has created a resurgence of hardware worldwide. There is now an understanding that the gaps between the desire to offer advanced services and the hardware limitations needed to offer those services must be bridged.
When referring to specialized hardware for AI, this usually means processors with incredibly strong computational abilities and sometimes even special architecture that makes it possible to run AI algorithms within a reasonable period of time. Nvidia, considered a leader in chips for AI, has grown exponentially over the past year. The growing interest in AI has sparked competition between Nvidia and other hardware companies like Intel, and even start-ups such as Graphcore, as well as tech giants that develop hardware capabilities for themselves, like Google, IBM, Apple, Baidu and Facebook.
The reality is that technology companies are developing the hardware they need by themselves. This is particularly apparent among the giants: Amazon, Google, Microsoft and Facebook, who several years ago concluded that existing products in communications, data processing and storage did not meet their requirements. A great example of this was AWS’s 2015 acquisition of Israeli chip company Annapurna Labs, which developed a platform on chip technology for data communications. A large part of the advanced services offered by AWS today are based on this technology.
Tags: Artificial Intelligence | Robots
Comments