What Is Google Tensor And Why Is It A Game Changer?
October 25, 2021Google’s Tensor is the company’s first attempt at creating its own SoC that can be utilized to create dedicated software. This is a similar approach Apple has taken with its iPhones and recently launched MacBook Pros (Apple Silicon). Building a dedicated chipset will mean Google can build a smartphone the way it would like, instead having to make compromises.
That’s why, a few years ago, its researchers came together to create the best mobile machine learning (ML) computer. This was to gauge what could be possible with Pixel smartphones. Co-designing Tensor with the Google Research team allowed the Search giant to build an AI/ML platform that could keep abreast with the software being developed at the company.
Tensor has features that require state-of-the-art machine learning technology. This includes Motion Mode, Face Unblur, Speech enhancement mode for videos and applying HDRnet to videos. It can adapt to suit a user’s needs instead limiting them due to the constraints placed on it by traditional technology.
Google Tensor in the Pixel 6 and Pixel 6 Pro use ML models but with lower power consumption than their predecessors. It also uses accurate Automatic Speech Recognition (ASR) tech the company has ever made. ASR can be kept running for applications like Recorder or Live Caption for extended periods of time without detriment to battery life.
Credit: Google
One of the best things about the new Pixel phones is the ability to converse in real-time with people of other languages. Features like Live Translate are possible because of Tensor, which works on many apps including Messages and WhatsApp. The translations happen within the apps, instead of having to shuffle between said apps and Google Translate.
Computational photography is better than ever. Tensor lets the new Pixel phones take impressive pictures. This includes Motion Mode, which can add motion to a picture’s background. You can also remove people or objects from pictures. Take a look at what the new Pixel 6 and Pixel 6 Pro can do here.
Tensor’s heterogeneous architecture means it can use the entire chip to complete tasks instead of a portion of it. This allows it to do more in the same amount of time. Since parts of HDRNet have been embedded, video recordings can be captured at 4K at 60 frames a second more accurately and with vivid colours.
There’s also more accurate face detection. Tensor helps the new phones locate and focus on a subject more quickly, but consume only half the power of the used by the Pixel 5. These are just of the advantages that Google’s home-baked SoC brings to the table. But this is only the tip of the iceberg because the new chip could be the genesis of AI capable smartphones.
Working together with Titan M2 (security chip), Google Tensor Security core and Trust Zone, they give the Pixel 6 and 6 Pro the most layers of hardware security of any phone. Tensor Security core is a CPU-based subsystem that works with Titan M2 to protect sensitive data.
According to the Search giant, Titan M2 can withstand attacks like electromagnetic analysis, voltage glitching and laser fault injection.
