- From: Dominique Hazael-Massieux <dom@w3.org>
- Date: Wed, 7 Apr 2021 17:22:40 +0200
- To: "public-webmachinelearning@w3.org" <public-webmachinelearning@w3.org>
Dear Web & Machine Learning Community Group, As Anssi pointed out in February, the W3C Advisory Committee has been reviewing a proposed charter for a Working Group who would bring the WebNN API and possibly the Model Loader API on the W3C standardization track: https://lists.w3.org/Archives/Public/public-webmachinelearning/2021Feb/0004.html Following the input received in that review [1], we're proposing to incorporate changes in the charter before its final approval by the W3C Director. To match the request for a better defined scope in the charter to reflect the known scope of the WebNN API, the scope of the charter has been refined with inspiration from the charter used by the Web & Machine Learning Community Group. The tentative work on a Model Loader API has been clarified as depending on the availability of a well-defined standard format for Machine Learning models. The list of groups with which to coordinate has been extended to include the WebRTC Working Group (due to the desire to have WebNN operate on media streams) and ECMA TC39 (due to the desire to coordinate on possible extensions to the JavaScript language syntax). All the changes above can be seen in https://github.com/w3c/machine-learning-charter/pull/15/ In addition, given multiple input from the Advisory Committee review noting ethical risks with the use of Machine Learning, we proposed the following additional changes: * specific coordination with the TAG to integrate lessons from their Ethical Web Principles https://github.com/w3c/machine-learning-charter/pull/16/ * a commitment from the group to develop a Note documenting ethical issues associated with using Machine Learning on the Web, and documenting ethical considerations in the normative documents developed by the group https://github.com/w3c/machine-learning-charter/pull/17 Several reviewers expressed concern that the WebNN API was proposed at the wrong layer (either too low or too high level). We are not proposing any change to the charter on this - following up with the reviewers, we pointed out previous research done by the Community Group in this space showing that the proposed level has emerged as the right trade-off between optimal performance and widest interoperability. You can see the updated charter here: https://w3c.github.io/machine-learning-charter/charter.html and this link shows the changes highlighted in HTML: https://services.w3.org/htmldiff?doc1=https%3A%2F%2Fwww.w3.org%2F2021%2F02%2Fproposed-machine-learning-charter.html&doc2=https%3A%2F%2Fw3c.github.io%2Fmachine-learning-charter%2Fcharter.html Please contact me if any of these changes feel problematic, before April 16. Best regards, Dom 1. A subset of that feedback is available publicly: https://lists.w3.org/Archives/Public/public-new-work/2021Mar/0000.html https://lists.w3.org/Archives/Public/public-new-work/2021Mar/0004.html https://lists.w3.org/Archives/Public/public-new-work/2021Mar/0015.html
Received on Wednesday, 7 April 2021 15:23:02 UTC