Making AI Headlines · Page 3
Wikimedia Wants AI to Read the Web Without Reciting It
· New York time
The Wikimedia Foundation is asking courts to protect AI training on copyrighted works while not protecting chatbots that reproduce those works for users. The operator of Wikipedia says an overly rigid ruling in major copyright cases could threaten the open web. Its position separates two uses that can get bundled together in arguments over AI. Training a model on material is one question; a chatbot repeating that material to a user is another. Wikimedia wants the courts to make room for the first without giving the second an automatic pass. The foundation also warns that a sweeping ruling could deprive Wikipedia contributors of useful AI tools or endanger the online environment on which their work depends. Wikipedia itself is a prominent example of a project built around making information accessible, while relying on people to assemble and maintain it. The cases involve competing interests in creative work, AI development and access to information. Wikimedia's proposal does not settle where a court should draw every boundary, but it gives judges a more precise distinction than simply declaring all AI uses good or bad. An encyclopedia famous for footnotes is asking the legal system for a ruling with room for one. In a debate that often offers two enormous buttons marked yes and no, Wikimedia has submitted a request for a third: it depends.
Read this story in the full newspaper →Satire. Not factual reporting. Read the original reporting from the linked publishers.