Artificial Intelligence Google’s new learning chip
Google’s father company Alphabet Inc. has created a first chip for machine learning.
Last week Google I/O developers have ended their conference, and the peoples can’t stop talking about the new services and products by the tech giant. One more thing they have announced there that the new artificial intelligence machine learning chip is currently being developed by Google itself.
This new chip, TPU or Google Tensor Processing Unit, is currently being developed to work with best Google Tensor Flow Technology. Google is using this chip for over past one year and claims that the Alpha Go supercomputer, which defeated the World champion go player, was also powered by a Tensor Processing Unit.
If we can see our last report it shows that earlier Google had hired a top chip engineer for their work to work with the design part of the new chip because of its complex structure and design. For this new chip, Google Inc. has designed a custom ASIC that was best and efficient on both performance and power while being used for artificial intelligence machine learning.
This new Tensor Processing Unit, can able to solve more complex algorithms quicker as compared to GPU because GPUs requires a long time for calculation due to its extra circuitry.
However, Alphabet Inc. latest chipset is very much expensive to build and will not likely make its way to consumer market. Building a custom ASIC for a Tensor processing unit can range anywhere from $40 million to $100 million, depending on the size of the team, complexity and timeline. Due to its high cost, the company stated that this chip will only be used for company and within the company and will not available in market for sale.
It is also said that the new Tensor processing units will not end its partnership with NVidia Corporation as a tech giant will continue to buy Graphic processing units from NVidia Corporation.
