Google's Stealth Release of Localised AI App Steps Up Privacy and On-the-go Processing

Stepping into a new era of the edge computing and privacy-focused AI deployment, Google released a revolutionary app, called AI Edge Gallery. The app enables Android users to run complex AI models directly on their phones without the need for a typical cloud-based connection. This notable development introduces the potential for a host of services such as image analysis, coding assistance, text generation, and conversational AI for multi-turn dialogue scenarios.

Users can download AI Edge Gallery directly from GitHub. Once installed, they can access a range of openly available AI models from the popular AI dev platform, Hugging Face. This means users can find, download, and run the models even if they don’t have access to the internet, enabling a homogenous set of capabilities across various processors and devices.

Dubbed as an ’experimental alpha release’, the application’s home screen offers shortcuts to AI tasks and capabilities like ‘Ask Image’ and ‘AI Chat.’ At the tap of a feature, users can pull up a list of models suited to complete the task. The app also comes loaded with a ‘Prompt Lab,’ which lets users initiate ‘single-turn’ tasks, like text summarisation and rewriting, through task templates and adjustable settings.

However, there’s no one-size-fits-all in terms of performance. Modern devices with more robust hardware may go through these models faster than their older counterparts, and the model size also plays a significant role. Larger models can take more time to answer a question about an image, for example, than smaller models would.

Being under an Apache 2.0 license, the app’s wide use-case contexts give it an advantage, allowing almost any commercial or non-commercial use of the software without restrictions. As the AI world awaits its iOS version, the developer community is eagerly hopping on board to provide feedback and improve the Google AI Edge Gallery experience.