Why Google Has Been Hiring Chip Designers Secretly
Google's Deep Mind research team has developed an artificial intelligence program backed with hardware called AlphaGo. The program, which defeated Go champion Lee Sedol in 2016, runs on custom accelerator chips built by Google's hardware engineers. Google calls its new chips tensor processing units (TPUs) for its machine learning software Tensorflow. Here are more details about why Google has been secretly hiring chip engineers.
Tensor Processing Units (TPUs) The new faster chip is part of a group of processors called application-specific integrated circuits (ASICs). Separate ASICs can be programmed for specific tasks. Google engineer Norm Jouppi says that TPU allows for fewer transistors per operation. It means that more operations per second can be embedded in silicon.
The new accelerator is the first indicator of the work that Google's chip designers have been working on for years. TPU chips are soldered onto circuit boards and slide into disk-drive slots in servers. The accelerators are customized for machine learning and can do exhaustive calculations.
Rival Technology While Google has been experimenting with ASICs, IBM has built a processor called True North on a similar concept. The chip is able to run new AI software on limited power without a cloud connection. TrueNorth comprises about 5.4 billion transistors populated in an area the size of a postage stamp. Google's ASICs, however, are much faster than field-programmable gate arrays (FPGAs). Princeton University electrical engineering professor David Wentzlaff confirms that ASICs outpace FPGAs and microprocessors by three times.
Artificial Intelligence and Machine Learning Google has increased its focus on designing programs for machine learning, with over a hundred available programs. Google engineers believe their chip is seven years ahead of competitors. At the recent Google I/O summit, Google CEO Sundar Pichai made it clear that the company is on the cutting edge of machine learning. He predicted that machine learning and AI technology will affect a wide range of fields.