Google Has Built A Custom Chip For Machine Learning
Google’s Tensor Processing Unit Artificial intelligence and machine learning have become central to Google’s business. The company has amassed a massive data centers to crunch vast troves of data and run its intensive machine learning algorithms, but little is known about how everything works on the hardware level. At Google’s annual developer conference on […]
Google CEO Sundar Pachai said it’s building what it calls the Tensor Processing Unit (TPU) dedicated to running machine learning algorithms.
TPU is its custom chip optimized for Google’s machine learning algorithm, TensorFlow. Google said TPU can run these algorithms with a better optimized performance on a per watt basis than existing third-party silicon. Google said TPU is the equivalent of moving the tech performance about seven years into the future, or about three generations of Moore’s Law.
Google said it’s been working on this chip in secret for the past several years and it’s been running inside its data centers for more than a year.