Gemini Nano: Google's revolution in AI for mobile phones even without an internet connection

Google announced this year GeminiIts most advanced AI model to date. Android 14 And the sample said Artificial intelligence They are soon integrated to revolutionize work on mobile devices.

The new tool will allow advanced functions such as generation Clever answers and grammar correction in applications without requiring a network connection.

Gemini Nano It comes in three optimized variants: Ultra, Pro and Nano; The latter is ready to run in settings as a preview for developers Pixel 8 ProOffers high-quality auto-replies in apps like WhatsApp.

Implementation Gemini Nano It contains Pixel 8 Pro Explains its capabilities, allows application Recorder Even if there is no internet, keep the content short.

Android AIcoreA new system service in Android 14 makes it easier to access the Gemini Nano and handles model management, runtimes, and security features.

AiCor It encourages customization through adaptation Low Standard (Lora)It allows developers to create small LoRA adapters tailored to their own training data, thus enabling a large, customized language model for specific use cases in each application.

Application of new ML hardware like latest Google Tensor TPU and NPU from companies like Qualcomm, Samsung S.LSI and MediatechExtend the reach of AICore and Gemini Nano to more devices soon.

Developers who are interested in exploring the possibilities it offers Gemini Nano To learn more about the tools and APIs available for modeling on devices, you can sign up for the Early Access Program Android.

The goal is to merge Advanced AI research With accessible tools to foster innovation in mobile applications that enhance user experience through highly flexible and powerful artificial intelligence.

To test the tool, you need to enter this form.

The announcement everyone was waiting for Google DeepMind with GeminiA multimodal and flexible artificial intelligence (AI) model promises to revolutionize the way technology intertwines with everyday life and business development.

Google and Alphabet CEO Sundar Pichai said the new model, part of the company's “AI-first” initiative, has state-of-the-art performance on several academic standards and is capable of outperforming human experts in tests. .

Optimized to run on devices of varying capabilities (Ultra, Pro and Nano), Gemini has shown excellent performance in 30 out of 32 academic tests used for language models, including a score of 90.0% in language comprehension. Many opinions Massive (MMLU) covers the analytical ability of human experts in fields such as mathematics, physics, history, and medicine.

Google CEO He promised the continued acceleration of AI advancement in the company and that these advancements will be of global benefit.

The new system, the result of a broader collaborative effort with Google, is designed to efficiently understand and analyze text, code, audio, image and video.