Great report, especially the selection of papers grounded in real-world impact. MuZero is extremely though-provoking: can we really have the super intelligence of AlphaZero consistently applied to messy problems, such as societal ones? I am thrilled to see how, if that happens, it cascades across organisations and institutions. Although, the hardness of cracking the missing pieces is very hard to estimate, I am curious to hear if anyone here has any informed guess about it.
Thank you for commenting, Riccardo. I don't think we'll see RL applied to complex 'real world' problems any time soon, but I'd be glad to hear others' thoughts on this question too (and more examples of live deployments of RL more generally)
Very useful update. Many thanks Azeem and Libby.
Great report, especially the selection of papers grounded in real-world impact. MuZero is extremely though-provoking: can we really have the super intelligence of AlphaZero consistently applied to messy problems, such as societal ones? I am thrilled to see how, if that happens, it cascades across organisations and institutions. Although, the hardness of cracking the missing pieces is very hard to estimate, I am curious to hear if anyone here has any informed guess about it.
Thank you for commenting, Riccardo. I don't think we'll see RL applied to complex 'real world' problems any time soon, but I'd be glad to hear others' thoughts on this question too (and more examples of live deployments of RL more generally)