The thing is though, current AI safety checks don't stop actually harmful things while also hyperfixating on anything that could be seen as politically incorrec

The thing is though, current AI safety checks don't stop actually harmful things while also hyperfixating on anything that could be seen as politically incorrect. First two prompts I chucked in to make a point: https://chatgpt.com/share/69900757-7b78-8007-9e7e-5c163a21a681 https://chatgpt.com/share/69900757-7b78-8007-9e7e-5c163a21a6... https://chatgpt.com/share/69900777-1e78-8007-81af-c6dc5632df98 https://chatgpt.com/share/69900777-1e78-8007-81af-c6dc5632df... It was totally fine making fake news articles about Bill Clinton's ties to Epstein but drew the line at drawing a cartoon of a black man eating fried chicken and watermelon.