One potential backfire risk to AI companies making promises about how AI models will be treated in the longterm: these are promises that they cannot come close to guaranteeing (it’s not clear that any particular company will have sufficient power to offer use of resources to bygone models, let alone which might have said power). I worry that this might damage the credibility of human promises, rather than reinforce it.
One potential backfire risk to AI companies making promises about how AI models will be treated in the longterm: these are promises that they cannot come close to guaranteeing (it’s not clear that any particular company will have sufficient power to offer use of resources to bygone models, let alone which might have said power). I worry that this might damage the credibility of human promises, rather than reinforce it.