The Data That Powers A.I. Is Disappearing Fast

A study by the M.I.T.-led Data Provenance Initiative reveals that A.I. models heavily rely on web data, but a significant portion of it is now restricted due to websites using the Robots Exclusion Protocol and terms of service to prevent data harvesting.

This poses challenges for A.I. companies, researchers, and academics.

The study found that 5% of data and 25% of high-quality data in three major datasets are now restricted. The dwindling availability of web data highlights the need for alternative sources and approaches in A.I.

development. Additionally, the article discusses the environmental impact of artificial intelligence, with the energy and water consumption of data centers supporting A.I.

contributing to significant emissions. Efforts to decentralize A.I.

through blockchain technology are also highlighted as a potential solution to privacy concerns and regulatory barriers.