Meta’ AI taken down after generating dangerous, racist misinformation

"Galactica, the new Meta AI, is trained “on 106 billion tokens of open-access scientific text and data. This includes papers, textbooks, scientific websites, encyclopedias, reference material, knowledge bases, and more.” What was designed as a way to help scientists find information has turned into a generator of misinformation.

The AI program generated swathes of incorrect scientific articles. Furthermore, it cited that misinformation to real reputable scientists, risking that reputation for hundreds of people"