Why Android AICore occasionally takes more storage - Check Google Answer

Google now has provided the reason for occasional high AI core storage usage and this happens due to the use of models for various AI intensive tasks. In this post we are exploring how Google AI model usage for various tasks is responsible for high storage usage.

With the help of Android AICore you can “run generative AI features directly on your Android phone or tablet’s hardware.” The AI models such as Gemini Nano are large and take a lot of space. Google is now providing a more detailed explanation of the occasional spike of storage while using Android AICore.

Google Android AICore is available on Android 14 and higher operating systems. Here are features and app that uses local model which contributes to higher occasional storage spike:

Advanced proofreading: If you are running grammar correction and language understanding features of the app then these apps will use Android AICore.

Automatic Speech Recognition (ASR): This is one of areas where you use the Android AICore for conversion of streamed audio and audio files into text transcripts. These are CPU/GPU  intensive tasks and use AI Models to achieve this. During this process storage usage might spike.

Scam detection: Scam detection is used to find messages that can be considered spam or scams. If such services are in use in your mobile phone then there might be more usage of storage in your mobile phone.

Smart reply: If you use the smart reply option in the applications such as WhatsApp, Line, and KakaoTalk through Gboard, then the storage usage might spike due to use of AI Model

.

Summarization: Summarization of the audio is another task which uses AI models and there might be spikes in the storage.

Translation: If you use the translation feature of the apps, where you are translating text from one language to another then there might be a spike in storage use.

“Sensitive information remains on your device. It’s never sent to the cloud or to Google.”

“AI-powered features work even when you have no network connection or are in Airplane mode. This includes text summarization and smart replies.”

“Local processing removes the lag associated with cloud-based services providing a faster and more reliable experience.”

A Google support article says “Why storage use increases.” When a new version of Gemini Nano is available, AICore “temporarily keeps both the old and new versions of the model for up to 3 days.”