Generative AI Power Right in Your Browser: Local Model Execution with WebNN and Prompt API

The surge of AI is unmistakable, with tools like Generative Fill in Adobe Photoshop and Microsoft Copilot seamlessly integrated into Office and Windows. Developers are increasingly exploring on-device AI execution, challenging the dominance of cloud-based approaches.

This paradigm shift offers advantages such as offline accessibility, enhanced data privacy, and cost-free execution, albeit with considerations for device capabilities and initial download times. Large Language Models and Stable Diffusion models are already running in web browsers, enabled by the WebGPU API. The Web Neural Network API (WebNN) by W3C is set to further enhance the performance of AI model execution by providing web apps with access to OS-level machine learning APIs. Additionally, Chrome’s Prompt API allows web app developers to share commonly used models, simplifying implementation.

In this session with Christian Liebel, Thinktecture's W3C representative, you’ll explore diverse approaches, practical use cases, and the future of AI integration in web development.