It’s likely induced bias as a result of the alignment training, not the web data. The reason people are up in arms about AI safety to begin with is that they re

It’s likely induced bias as a result of the alignment training, not the web data. The reason people are up in arms about AI safety to begin with is that they reproduce humanity as it is, containing racism, sexism, etc.