Add Generative AI to your Android app with the Gemini API

Introduction to Gemini models, including Gemini Nano, Gemini Pro, and Gemini Ultra models that were launched in December. These models are designed to understand various types of information such as text, code, audio, image, and video.

Introduction of AI Core, a new service for the Android operating system that handles Model Management, runtime safety features, and more. It enables running generative AI directly on the device and is currently in private documentation phase.

Detailed overview of Gemini models including Gemini Nano, Gemini Pro, and Gemini Ultra. Gemini Ultra is highlighted as the most advanced model that outperforms human experts on various topics evaluated by the MLU Benchmark.

Introducing Chemma, a family of open generative AI models using technologies from Gemini models. ML researchers and engineers interested in experimenting with language models on Android can use the experimental MediaPipe LLM Inference API.

Exploration of use cases for generative AI on Android including using the device's camera, improving text input, summarization of text, and examples of features powered by generative AI implemented in Google apps such as Pixel Recorder and Messages for Android.

Explanation of prompting techniques for generative AI including being specific in instructions, defining the expected response style, using fusion prompting, and experimenting with different prompts. Clear instructions and constraints are key for satisfactory results.

Step-by-step guide for integrating Gemini Pro into an Android application using the Google AI client SDK or the Vertex AI SDK for Firebase. Process includes prototyping prompts in Google AI Studio, generating API keys, and integrating the SDK into the project.