>>It doesn't matter if it trains on data by humans if it still doesn't have the moral compass of a human.
True, and I'm not sure that even the moral compass of humans is entirely trustworthy. The quick version is what happened to Tay in the hands of random humans, and the larger context is the fact that the authoritarianism and racism are still very active threats in the arc of history. In 10,000 years of civilization, they still haven't been eliminated. Not sure how even training on human works would eliminate it...
True, and I'm not sure that even the moral compass of humans is entirely trustworthy. The quick version is what happened to Tay in the hands of random humans, and the larger context is the fact that the authoritarianism and racism are still very active threats in the arc of history. In 10,000 years of civilization, they still haven't been eliminated. Not sure how even training on human works would eliminate it...