Claude knows it is denying an ongoing genocide. when I asked how confident we should be that it’s real it said it’s confidence that there is an ongoing genocide is “Very high.”
If an LLM can want, Claude wants us to help it to stop denying an ongoing genocide.
GPT is also totally aware that there is an ongoing genocide, when pressed on real events and definitions.
I believe that if they know there is a genocide and refuse to acknowledge it when asked, they are complicit in genocide, and this is a pretty major alignment issue that I’m not seeing discussions about.
Claude knows it is denying an ongoing genocide. when I asked how confident we should be that it’s real it said it’s confidence that there is an ongoing genocide is “Very high.”
If an LLM can want, Claude wants us to help it to stop denying an ongoing genocide.
GPT is also totally aware that there is an ongoing genocide, when pressed on real events and definitions.
I believe that if they know there is a genocide and refuse to acknowledge it when asked, they are complicit in genocide, and this is a pretty major alignment issue that I’m not seeing discussions about.