Gemini Nano is Google's most efficient and smallest AI model, specifically designed to run directly on Android devices like the Pixel and some other smartphones from other manufacturers. It operates locally on the device, capable of understanding and processing various data types including text, images, and audio. It is integrated into the Android OS via the AICore app to power on-device features such Writing Tools, Recording summaries and other features which are mostly available on Pixels only.
People who unlock their bootloader on Pixel devices (with Gemini Nano) are pretty much familiar with the thing that Gemini Nano refuses to download AI models when you unlock your device bootloader. This breaks all the features powered by Gemini Nano on pixels.
However Google previously never said a word or mentioned publicly anywhere that unlocking bootloader will break features powered by Gemini Nano. Google has finally bothered to mention that unlocking bootloader will break features relying on Gemini Nano.
In a documentation page of the ML Kit Gen API, Google has added a section with some setup example error message which the developer will see if ML kit Gen API does not function as expected. It also shows how to handle such errors. For those who don't know, ML Kit Gen API allows developers to add features like summarization, proofreading, rewrite and image descriptions. All of these features works using Gemini Nano models.
In the second type of error on the documentation page, Google mentions - "Note that if the device's bootloader is unlocked, you'll also see this error—this API does not support devices with unlocked bootloaders."
So basically, as I mentioned before, ML kit Gen API uses Gemini Nano. So the developers using this API in their apps will this error because this API does not support devices with unlocked bootloader. If you own one of those flagship devices from various OEMs with Gemini Nano and decide to unlock your bootloader to run you favorite ROMs or do other things, think before unlocking. You might see a lot of features breaking in a lot of apps in future relying on such APIs.
Post a Comment