Home ยป Google unveils AICore: the avant-garde Android 14 service empowering mobile devices with model execution capabilities, commencing with the groundbreaking Gemini Nano.

Google unveils AICore: the avant-garde Android 14 service empowering mobile devices with model execution capabilities, commencing with the groundbreaking Gemini Nano.

Google has revealed the details of AICore, a new service for Android 14 that utilizes AI models within the device. The journey begins with the new model called Gemini Nano, which is the first of its kind.

AICore serves as a service for the operating system, similar to ARCore, which handles AR functions. Its purpose is to act as a middleman that allows apps to access AI features, while the actual processing is done on the Google Tensor TPU chip (for Google phones) or NPU (for Qualcomm/Samsung chips).

Within AICore, in addition to the model itself, there is a component called Low Rank Adaptation (LoRA), which has nothing to do with IoT protocols. LoRA focuses on fine-tuning and incorporates security features for the models.

Several Google apps now utilize AICore, such as Recorder or Gboard, which offer new features for the Pixel 8 Pro. For other developers who wish to use it, Google has opened up Early Access Preview.

TLDR: AICore is a new service for Android 14 that runs AI models internally. It acts as a middleman for apps to access AI features while processing is done on specialized chips. Google apps like Recorder and Gboard already use AICore, and other developers can access it through Early Access Preview.

More Reading

Post navigation

Leave a Comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Python/JavaScript: Unleashing Artificial Intelligence Software Ollama on PC

The Emergence of Kolmogorov-Arnold Networks: A Novel Architecture with Potential to Revolutionize the World of Deep Learning Through Reduced Parameters for Simpler Understanding

A Myriad of AI Models on the Hugging Face Platform Surpassing One Million Models