Jesus, I thought Wikimedia at least would understand its editors. Who on earth thought this was going to go over well?
SE was at least a for-profit company, WM has no excuse for this. There are plenty of places where they can use AI really fruitfully, but none of them are to do with generating user-facing content.
https://www.404media.co/wikipedia-pauses-ai-generated-summaries-after-editor-backlash/
Wikipedia really needs coherent position on AI. They have a copyright claim on any model that spits out WP text verbatim without attribution.
That position is going to be severely undercut if they cram their own product full of LLM-generated stuff that disrespects other people's copyright.
In this case, the model is Cohere's Aya, which was pretrained on mC4 (a cleaned version of CommonCrawl), which is general internet content. In short, full of copyrighted data.
@pbloem yuck indeed. Glad they stopped but wtf were they thinking. What kind of tone deaf echo champer are these people in?