Will we simply put in a hard override to get the answers we want to get, the answers we find comfortable Yep. That's already happening. See Google's papers on

Will we simply put in a hard override to get the answers we want to get, the answers we find comfortable Yep. That's already happening. See Google's papers on "debiasing AI" (if you look at their example goals, they mean biasing it away from what it learned and towards what Googlers wish was true).