简体中文
EnglishWhich Web Specifications Does Web Machine Learning Work On?
Updated
Web Machine Learning (webmachinelearning.github.io) is the home of the W3C Web Machine Learning Working Group and its sister Community Group. The site names WebNN and WebMCP as examples of specifications you can contribute to. The Working Group standardizes Web APIs for in-device machine learning inference, while the Community Group incubates new proposals and explores ideas before they enter formal standardization.
How the work is split between the two groups
The site describes a two-stage structure, and knowing which stage a proposal is in tells you where to look or contribute.
| Group | Role | What happens there |
|---|---|---|
| W3C Web Machine Learning Working Group | Standardization | Develops Web APIs for in-device ML inference, working within the W3C ecosystem |
| Web Machine Learning Community Group | Incubation | New proposals are discussed and explored before formal standardization; well-received incubations serve as seeds for the Working Group |
The Working Group's scope is specifically in-device inference — running models locally in the browser rather than calling a remote service. The site frames this as combining performant numerical computation with the reach of the web.
The specifications named on the site
WebNN
WebNN is the specification the site leads with. It fits the Working Group's stated goal of enabling efficient ML inference in the browser, and the site lists it first among the specs open to contribution.
WebMCP
WebMCP is the second specification the site names as an example of work to contribute to. The site gives it as a parallel example alongside WebNN rather than describing its scope in detail, so treat the site itself as the entry point for its current status.
The phrasing "such as WebNN WebMCP and more" indicates these are examples, not a complete list — the site does not enumerate every specification in progress.
Why in-browser inference is the organizing goal
The specifications exist to deliver four properties the site attributes to in-browser inference:
- Low latency — inference runs in the browser, enabling use cases built on local media sources.
- Privacy preserving — user data stays on the device.
- High availability — no network reliance after initial asset caching, so offline use cases work.
- Low cost — computation happens on client devices, so no server farms are needed.
These are the stated motivations behind the API work, not separate specifications.
Where to get involved
The site's call to action is a repeated "Join us" invitation to contribute to web specifications such as WebNN and WebMCP. If you want to shape a proposal before it is standardized, the Community Group is the described venue for discussing and exploring new ideas; if you want to work on the standardized APIs themselves, the Working Group is the relevant body. The site does not publish membership steps, meeting schedules, or contribution mechanics on the page, so use the join links there as the starting point rather than expecting a process guide.