Artificial Intelligence is continuously leading the way among the span of technologies in bringing out new software designs and capabilities. Google’s new AI chip is the latest one to get added to this.
At company’s annual developer conference held recently, CEO Sundar Pichai announced a new AI chip that can be a technological break-through in the world of machine learning and automation.
With this, Google signaled the increasing role of AI in software and hardware development.
The chips known as Cloud Tensor Processing unit (TPU) are designed to speed-up the machine learning operations and are also an upgrade from the first generation of chips Google announced at its last year’s I/O meet.
The second-generation chips can deliver nearly 180 teraflops of performance which is much higher as compared to its first-generation which could only handle inferences. The new chips can be used to train machine learning models. Training is an important part as it would help the machine learning program to identify and differentiate between images, data or other things.
Pichai also went forward announcing the creation of machine-learning supercomputers which will be based on Cloud TPUs to be used along with high-speed data connections. He said, “We are building what we think of as AI-first data centers Cloud TPUs are optimized for both training and inference. This lays the foundation for significant progress (in AI).”
Google emphasized on the application of machine learning and AI models to bring better work and performance oriented results. Google.ai is also the result of collective efforts towards building an AI first world.
However, Google did not clearly say anything about introducing the chip in the market, rather said that it will let companies rent access to the chip via its (Google’s) cloud computing service.
For the first came across cloud TPUs. Very helpful article. Best move by Google to build AI chips.