The new Google Pixel 6 is is all about machine learning

Google goes all in on Machine Learning with its own bespoke SoC for the new Pixel

by Justin Choo

Watch out Apple, Google has joined the processor game too. Google’s latest Pixel phones – Pixel 6 and Pixel 6 Pro – feature Tensor, the company’s first processor.

But first, let’s look at the phones. Google’s latest Android 12 phones have distinct designs: the Pixel 6 is vibrant and comes in versatile colour options while the Pixel 6 Pro’s polished metal unibody and curved glass aesthetic were designed with a luxurious look in mind. The distinct camera bar draws attention to the camera and gives the phone a very modern, chic feel. Google has also put in the work to tweak the overall look and feel and even the UI changes to match the chosen wallpaper.

New Pixel, new camera

Both phones now use a new 1/1.3 inch sensor for the rear camera. Google says this new sensor captures up to 150 per cent more light than the main sensor on the Pixel 5. Both phones will also have new ultrawide lenses with larger sensors.

The Pro version will get a 4x optical zoom telephoto lens with an improved version of Super Res Zoom, which can reach 20x zoom. The Pixel 6 Pro will also have an ultrawide front camera that can record video in 4K resolution (1080p on Pixel 6). That wider front camera on the Pro also supports Snapchat’s new ultrawide selfie feature. For that matter, a new Quick Tap to Snap feature will be available exclusively on Pixel 6 and Pixel 6 Pro later this year.

But the attention mostly will be focused on Google’s take on their own processor. Google Tensor is designed around machine learning (ML) and is the natural evolution for what Google has been doing with Google Assistant so far.

Mobile ML computer

Effectively, Google wants to build a ‘mobile ML’ computer into your phone. The platform will be used for a variety of features, which falls under what Google describes as core experience areas – speech, language, imaging and video.

Tensor SoC is designed around running Google’s advanced ML models but with a more efficient way of requesting resources, which in turn, lowers power consumption – making the ML model viable to be used as a core feature on the mobile device. Google cites the example of the company’s most accurate Automatic Speech Recognition (ASR), which is by Google Assistant on Tensor-equipped devices. Tensor makes it possible to run ASR on apps like Recorder or tools like Live Caption without draining the battery quickly.

ML Features

The power savings alone opens the door for many useful applications that have to be active for longer periods like the new Live Translate feature. Apps like Messages and Whatsapp will let you translate natively to save time, and yet it uses half the power. It happens on-device through the Private Compute Core so no data ever leaves the phone; it doesn’t need the Internet to work.

Speech recognition and language understanding models have been improved and you can use Assistant and simply speak to quickly type, edit, and send messages in Messages, Gmail and so on. Mundane stuff like punctuation and corrections will be handled by Google Assistant who can also help with inserting emojis and sending your messages.

Other features and benefits include:

  • Tap unwanted objects and people in your photos away with Magic Eraser .
  • Motion Mode: Add stylistic motion to your shots with modes like Action Pan and Long Exposure. Parts of HDRNET, a dedicated feature to implement the signature Pixel look more efficiently, is now embedded directly onto the chip, so it now works on all video modes for the first time and even at 4K and 60 fps.
  • Real Tone: Updated camera enhancement now handles darker skin better because cameras have been designed to photograph lighter-toned skin for decades.
  • More accurate face detection: quicker location and focus but at half the power compared to its predecessor.


Google claims that its new phones are remarkably secure, and they use a variety of security tools for this. At its heart is the Google Tensor security core, which utilises Titan M2 for security keys and is accompanied by TrustZone with Trusty OS to provide a Trusted Execution Environment. Tensor security core is the CPU-based subsystem that works with the new Titan M2 to protect sensitive user data. Google says that independent security lab tests showed that Titan M2 can withstand attacks like electromagnetic analysis, voltage glitching and laser fault injection (I suppose that’s more for dramatic effect than anything else).

Perhaps the most important thing to know is that Google has extended its support window to at least five years of security updates so that we can all make it till next year when we upgrade to the next big thing (just kidding).


The Pixel 6 is priced at US$599 and the Pixel Pro at US$899. The phones are slated for an October 28 release in the US, though there are no details for a Singapore release at the moment.