Google is opening up about its new super chip and is looking to help move the AI race ahead.
Because Google is smarter and faster, they strongly believe that artificial intelligence (AI) is the future in the sense that it will affect everything from human health, cars to languages. Even with the breakneck race between Silicon Valley giants to develop faster and more complex AI capabilities, many of them still heavily rely on existing processing chips. However, Google has now built a new weapon to finally break clear of the pack by developing its own chip.
Named Tensor Processing Unit (TPU), Google notes that this is a futuristic chip that has enabled it to move its technology “more than seven years into the future.” The company admits that it’s been using this chip for over a year in trying to improve its deep neural network with features that can tag people in photos, offer “smart reply” services on emails and offer “search by voice” options.
The chip’s most recent use was on Google’s AlphaGo algorithm, even though the company didn’t reveal it at the time.
“We’ve been running TPUs on our data centers for some time now, and have confirmed that they can exquisitely deliver on better-optimized routine per watt for machine learning,” Google explained in a blog post.
While this news of Google creating its own chip is a positive development for the AI industry, it could spell doom for traditional chipmakers such as Nvidia and Intel who have underpinned their hopes on AI as their future growth segment. Nonetheless, we all hope that this chip will be a cool way of improving the quality and accuracy of Google maps and navigation, among many other things.