rene_mobile’s Twitter Archive—№ 9,118

@thai101 If the designers of ChatGPT could technically answer which of the training inputs led to this behavior, ML would already be in a much more trustworthy state - for many applications.