@CleverMoniker @glacierglider1 @WashedOutGundamPilot @matrix It’s not the first priority of the researchers in general but it is the first priority of the institutions who finance the (very expensive) research.
Some times it’s because of dumb shitlib opinions, more often it’s from a fear of the inevitable lawsuits.
But it’s a lost battle, deep neural nets are useful for one reason and one reason only, they are great at amplifying weak signals and interpreting them, which means that they will « discover » deep and hidden racisms that will escape categorization as such from the human observers, making them not racist always implies making them bad at their job for any AI used for decision making.
Sure they can still find use in information retrieval applications, but those, while impressive, have limited commercial value, a machine that knows how to answer a plain language query looks like science-fiction and impresses journalists but it’s not as useful as a good old search engine returning hundreds of results and letting the human read. It’s more normie friendly, but it’s less powerful because of it.