The W3C officially launched the Web Machine Learning Working Group and plans to publish the first public working draft of the Web Neural Network API in the first half of 2021.

THE following is a translation of THE W3C Notice THE WEB MACHINE LEARNING WORKING GROUP.

W3C launches Web Machine Learning Working Group

This article was co-authored by Anssi Kostiainen (working Group Chair), Ningxin Hu and Chai Chaoweeraprasit (Network Neural Network API Editors) and Ping Yu (TensorFlow.js core team).

Machine learning (ML) is a branch of artificial intelligence. A sub-area of ML is called deep learning, and its various neural network architectures can bring new and compelling user experiences to web applications.Use casesThe range ranges from improved videoconferencing to improved accessibility features and potentially better privacy than cloud-based solutions. Enabling these use cases and many more is newWeb Machine Learning Working GroupThe focus of the.

While some of these use cases can be implemented within the device in a limited way through existing network apis such as WebGL graphics apis or future WebGpus, the lack of access to platform capabilities such as dedicated ML hardware accelerators and native instructions limits the scope of the experience and results in inefficient implementations on modern hardware.

With these design goals in mind, a W3C community group began work on incubating possible Web neural network apis in 2018, in response to encouraging feedback at the TPAC breakout sessions. Starting in October 2018, the community group identified key use cases, working with a variety of participants, including major browser vendors, key ML JS frameworks, interested hardware vendors, and web developers. After identifying the key use cases, the team decomposed the key use cases into requirements and began drafting the NNN API specification in mid-2019. The purpose of this use-case-driven design process is to put the user’s needs first.

“With access to native ML accelerators, machine learning frameworks such as TensorFlow.js can greatly improve the efficiency of model execution, truly democratizing ML for web developers.” – Ping Yu, Google’s Tensorflow.js TLM.

“Early empirical results from network neural network API implementations show tremendous power and performance improvements for network AI workloads. “By accessing the full native AI capabilities of modern heterogeneous hardware, the NNN API enables a new and transformative intelligent user experience across a variety of hardware, software, and device types on an open network platform.” – Ningxin Hu, chief engineer of Intel Network Platform Engineering

The W3C is organizing a workshop on the Web and machine learning in August and September 2020. One of the conclusions of the workshop, which brought together web platform and machine learning practitioners to investigate the broader intersections of web technology and machine learning, was to recommend the creation of a new W3C working group to standardize and graduate from the incubation phase of the Network neural network API. Starting in 2021, the Community Group continues its incubation capabilities, working in parallel with the working group, similar to W3C’s Work on WebAssembly and WebGPU.

“The NETWORK Neural Network API is a very important step towards the intelligent network of the future, where ARTIFICIAL intelligence is injected into users’ daily web experience. With the current pace of progress and innovation in AI hardware, it will help connect these experiences from the cloud and make them personal to the user through seamless native hardware performance on edge devices across the network. That’s the future to dream about!” – Chai Chaoweeraprasit, Principal engineering Leader for Machine Learning and Computing Platforms at Microsoft

The Web Machine Learning Working Group plans to publish the first public working draft of the Web Neural network API in the first half of 2021 and welcomes new participants from different W3C communities to help identify new use cases, document moral hazard and its mitigation, and contribute to the technical work. Conduct extensive reviews in privacy, security, accessibility and other important areas to ensure that the views of different online communities are heard. – which is particularly critical for group work, given some of the ethical implications of machine learning algorithms. Join us!

We would like to thank all participants in the W3C community group and THE W3C workshop, whose contributions helped shape this work, and the W3C for providing a venue to advance this cross-industry work towards widespread adoption.