We need this, yes, but current AI doesn’t have enough theory of mind to pull it off, reliably, and I don’t expect it will in the near-term. The idea itself is hardly new, eg https://
technologyreview.com/2007/01/01/227
178/anything-you-can-do-i-can-do-meta/
… from 2007, and the dream is older
Current AI Lacks Theory of Mind for Reliable Task Performance
By
–