Managing the impact of AI & Machine Learning on the Web

The past few months have seen an avalanche of announcements linked to Artificial Intelligence systems, mostly based on Machine Learning models.

These systems are strongly coupled with the Web as a platform: many models are trained from Web content crawled at scale, distributed or surfaced via Web interfaces, and in a number of cases, are used to generate content that gets published on the Web at an unprecedented rate.

These intersections have created a number of systemic impacts on the Web, spurring many important conversations on how they might change the Web as we know it, both in good and harmful ways.

Because these conversations have happened in a somewhat scattered fashion across different communities, they make it sometimes difficult to get a full picture of the problem space. As a contribution to making these conversations converge faster towards concrete outcomes, I have been writing up what systemic impacts I have identified in these conversations between these AI systems and the Web, and some of the early proposals that have emerged to manage them, and which need the scale and coordination of standardization efforts.

"AI & the Web: Understanding and managing the impact of Machine Learning models on the Web" is the document that the W3C Team is releasing today to serve as an anchoring point in these conversations, and we are inviting the community to review and provide input on that analysis to help improve and complete it. Comments and input are welcome preferably before June 30, 2024.

We hope that by providing that place for discussing and managing systemic impacts, we can help the community in charting a credible path to strengthen the position of the Web during this rapid evolution phase of the information ecosystem.