Google allows developers to build on-device AI-powered apps for Android | Tech News | Laptop News

Google has shared details on how it plans to integrate artificial intelligence into its Android operating system with on-device processing, leveraging the small scaled Gemini Nano model. In “Android on-device AI under the hood” session at I/O 2024, Google shared insights on how it is already powering select apps such as Messages, Recorder and more with Gemini Nano and how it plans to open it up for third-party developers.

In the video session, Google explained that on-device AI is beneficial as it offers local processing of data, maintaining users privacy. Additionally, it also allows the AI model to process data without an active internet connection and offers reduced latency. However, there are some limitations to it as well. Due to hardware limitations, the on-device AI models have smaller parameter size compared to large language models (LLMs) and also a smaller context window.

Google said that it has chosen “Gemini Nano” as the foundation model for its Android platform and is looking at three major use cases for it:

• • Content Consumption (example: summarisation and overview of text)

• • Content Creation (example: suggestions and assistance in writing messages)

• • Content Classification (example: understanding the sentiment or emotion of the text)

Google said that it has integrated Gemini Nano into the Android system from Android 14 onwards through a service called AICore. The company said that it is already working with select third-party developers for “compelling on-device Gemini use cases” and will enable more developers to access and use Gemini Nano to bring apps powered by on-device AI to Android smartphones.

The company said that the AI model is already powering AI…

read more www.business-standard.com