Google Pixel 8 Pro runs generative AI models: Can compact LLMs change the way we use smartphones?

Tensor + on-device generative AI models

At the Made by Google event this week, Rick Osterloh, SVP of devices and services at Google, revealed that the Pixel 8 Pro’s custom-made Tensor G3 chip, which is optimized for AI tasks, can run “distilled” versions of Google’s state-of-the-art models for generating text and images. These models can power various applications on the phone, such as image editing.

Tensor G3 enables better audio and video quality. (Image: Google)

“We’ve worked closely with our research teams across Google to take advantage of their most advanced foundation models and distill them into a version efficient enough to run on our flagship Pixel,” Osterloh said.

Unfortunately, the Pixel 8 didn’t make the cut, even though it has the same SoC (system on chip) as its bigger brother.

Google’s main selling point for the Pixel phones is their artificial intelligence capabilities. These phones have the Tensor chip, which may not score very high on benchmarks, but excels at AI performance. It is Tensor that handles the Pixel’s famous computational photography and its smart speech recognition features. And since the phone already has a chip that can handle AI, it makes sense to run AI models on it.

Having AI models on the phone means that it doesn’t need to send data to the cloud for processing, which can improve speed, privacy, and battery life. Now let’s take a look at what AI models on the Pixel 8 Pro can do.

Features powered by on-device generative AI

Magic Eraser

Google Photos has had a Magic Eraser tool for some time now, but it’s getting an upgrade thanks to on-device large language models. Now you can erase bigger objects and people from your photos without leaving any traces behind. The new Magic Eraser fills in the gaps with generated pixels that match the surrounding area, giving you a cleaner and more realistic result.

Zoom Enhance

It isn’t perfect (notice the cables), but it’ll probably get better with time. (Image: Google)

But the most impressive feat of on-device AI is Zoom Enhance, a feature that can sharpen your zoomed-in photos and get rid of any pixelation. Google showed off this feature by zooming in on one of the towers of the Golden Gate Bridge and enhancing it with a proprietary imaging model. The Google Pixels have always had great zoom capabilities, but they still lag behind Samsung’s insane 100x zoom cameras. Now Google is using artificial intelligence to bridge the gap.

Recorder app’s summarisation tool

Google is also applying on-device generative AI to its Recorder app, which will soon be able to automatically summarise your conversations and give you some key bullet points. This feature will be available later this year on the Pixel 8 and Pixel 8 Pro.

Smart replies in Gboard

Another large language model running on the Pixel 8 Pro will power smart replies in Gboard, Google’s keyboard app. The improved Gboard will offer “higher-quality” reply suggestions that are more aware of the context and tone of the conversation.