New algorithm overcomes the problem of AI being too “curious,” utilizing intrinsic rewards when they are helpful to save practitioners' time on deciding which algorithm to use on any new task: http://
bit.ly/3fStdhe

→ View original post on X — @mit_csail,