What happens when you put AI enthusiasts and AI skeptics into the same room? Well, at W3C you get conversations flowing and perspectives challenged. In this case, it has led to the W3C Advisory Board (AB)’s first ever public position statement: “Use of Large Language Models in Standards Work”.
The position statement reflects the AB’s current thinking on using Large Language Models (LLMs) in the standards process. It lists both opportunities, like brainstorming names and gap analysis, and risks, like subtle incorrectness and second-hand burdening as well as suggesting some guardrails for when LLMs are used.
The AB’s role is to provide guidance to the Team on issues of strategy, process, etc., and to gather and reflect W3C Member feedback. LLMs and their use are having a great impact on the world and technology, including the web. The balanced perspective, consideration of the thinking of other groups at W3C and proactive efforts AB members brought to this effort is an example of how the W3C community works at their best – bringing together different perspectives, backgrounds, and ideas to suggest a way forward for W3C and the web.
Hidde de Vries, co-author, noted: ”As an AI skeptic with a philosophy background, I’ve often been annoyingly pedantic about various aspects of LLMs and the ways they’re used. I loved working with Elena, who is much more optimistic and much more tuned into how people are using these tools, to take on the challenge of coming up with a balanced piece together.”
Elena Lape, co-author, added: “Standards are meant to last. And on the Web, that matters even more — because decisions made here can affect billions of people, for years to come. Hidde and I came at this from different perspectives, and I think that helped the Advisory Board arrive at a shared view that is neither uncritically open nor reflexively skeptical.”
The AB welcomes feedback and would like to hear how people in the W3C community are using LLMs today in the context of producing web standards, other factors to consider.and whether their observations resonate.