-
Notifications
You must be signed in to change notification settings - Fork 53
Description
Spun off from TAG review feedback:
We would appreciate if the WG would evaluate the likely impacts on sustainability from introducing this API, perhaps in collaboration with the Sustainable Web IG. There are several competing likely effects, including the comparative energy efficiency of personal devices vs datacenters, the greater efficiency of WebNN over WebGPU for the same workload, increased use of neural nets as they get easier to access, faster device obsolescence if older devices can't effectively run the workloads this API encourages, and likely other considerations. Any sustainability impacts might be balanced by increased utility and privacy, but it would be good to know what we're signing up for.
The group discussed this at its 2025-06-05 meeting and noted the following (rephrased slightly for clarity):
- purpose-built ML accelerators aka NPUs are generally known to be more power-efficient than GPUs, especially well-suited for long-running tasks
- WebNN API is good for sustainability because it (is the only API that) can use NPUs and also use NPU together with CPU and GPU allowing the system to optimize for sustainability
- WebNN API leverages Native ML APIs and as such does not require shader programs to be downloaded; this reduces download-related energy consumption wrt graphics APIs
- sustainability aspects could be incorporated in the device selection explainer, WebNN explainer and/or the WebNN API spec
Contributions wanted:
This issue is to solicit further input, suggestions, corrections and clarifications to inform related explainer and/or specification updates in response to this TAG feedback. Also validation of any of the above claims with data is helpful.
Furthermore, the TAG proposed a collaboration with the Sustainable Web IG. As the first step to that direction, the group participants are encouraged consult Web Sustainability Guidelines (WSG) and see if some of the guidance maps to the WebNN API specification. And suggest improvement to that doc to better consider sustainability aspects of compute APIs such as the WebNN API.