On one hand behaviour like this is deeply unacceptable for such a huge commercial product but on the other hand it's comforting that even the biggest AI companies has no idea how to solve catastrophic failures in their models
add a skeleton here at some point
about 1 month ago