Intel's Web Standards tech lead, Anssi Kostiainen, has said the World Wide Web Consortium's (W3C) Machine Learning for the Web Community Group now has "all major browsers – Google, Microsoft, Apple, Mozilla – on board along with the broader AI & web ecosystem."
The purpose of the group is to develop web APIs for hardware-accelerated neural network inference, necessary for the intensive processing required for AI features like real-time image analysis to take place in the browser.
Existing APIs like WebGL are unsuitable because they lack access to hardware features such as CPU parallelism, general-purpose GPU programming or dedicated machine learning (ML) processors.
"With these [new] Web APIs, web developers can make interoperable content on all hardware platforms," says the group's charter.
Kostiainen said the project will "unlock advanced deep learning usages in computer vision, speech recognition and more in a performant manner, all in your browser" and showed off a proof-of-concept video for Web ML (below).
The group, which was started in October 2018, promises to coordinate with other relevant W3C groups such as GPU for the Web, WebAssembly, Web Real-Time Communications, and Devices and Sensors.
A proof-of-concept video shows how a browser application could interpret objects seen through a smartphone camera, something which would otherwise require a device-specific app.
Standards groups work slowly, and not all projects see wide adoption, so the usual note of caution applies.®