TECH

How artificial intelligence dethroned manufacturers of processors
Do not think of processing chips in the traditional way, where computer technicians tried to find the best settings among Intel, AMD or NVidia proposals. Artificial intelligence and machine learning systems have forced technology majors to rethink processing, removing responsibility for the hardware introduced by computers for the next generation of cloud computing.In this sense, new companies, some improbable, appear in the market with technologies focused on data processing, as listed in Ars Technica. Intel will enter the market through the acquisition of startup Nervana Systems and Movidius dedicated to the processing of image AI. Microsoft continues to refine its HoloLens "mixed reality" system supported by a proprietary artificial intelligence processor that can be introduced to other devices.On Google's side, the company continues to improve its cloud computing platform to power its apps supported by AI. In that sense it has a chip called the Tensor Processing Unit, capable of running at 45 teraflops that can be grouped for greater processing power. Also Amazon plans to work on an AI chip to support its aide Alexa.To the sides of Cupertino, Apple recently hired engineers to Google to develop an AI processor called the Neural Engine to increase the potential of its Siri assistant and FaceID facial recognition system.But the list of technology to develop processors continues with ARM Holdings, which has developed two chips dedicated to image recognition, the ARM Machine Learning processor and the ARM Object Detection. IBM is also developing its AI processor, the Power9, even because its Watson system has been increasingly required, and as such, has also developed a partnership with NVidia for the licensing of NVLink, a processor-powered supercomputer (GPU) capable of managing 100 petaflops of computing.
If large technology companies are committed to the "war" of AI processing, The New York Times reports that at least 45 companies are developing new chips.The advancement and complexity of applications for mobile devices, especially those related to IoT functionalities and artificial intelligence, require a high level of processing, which should not be dependent on the limited infrastructure of the equipment. This may be one of the reasons for the "phenomenon", as it seeks to explain to Ars Technica.Artificial intelligence and machine learning systems are "monsters" that need to be fed, especially facial recognition, and even AI systems for autonomous vehicles need constant processing to determine the path and the combination of everything that is around you. Another detail is that not all processors perform the same tasks: some are dedicated to visual intelligence, others to graphic processing, Big Data, machine learning, and other essential purposes to their áreas.
The need for specialization of an area is possibly the greatest justification for having so many manufacturers developing their own processing technology dedicated to artificial intelligence. There is also a lot of software that stays in standby mode waiting for the processors to reach the required maturity to run with the speed and efficiency required.


Sapo