Re:Generalists vs specialists – Who has a greater chance of ...
by mratsim on Tuesday March 09, 2021 @08:24AM
(#26398517)
In reinforcement learning which is used for things like AlphaGo, there is a similar duality called "exploration vs exploitation":
- do you keep searching in breadth for new avenues
- or is it time to dive in depth in what you found so far.
And this is even optimized through something called "regret".
- example: https://rlss.inria.fr/files/2019/07/exploration_lazaric_1.pd...
(I'm not sure it has to be a two-dimensional spectrum in all scenarios —I believe there may be cases in which robustness and efficiencies are at least partially orthogonal/independent to each other.)
Taleb discusses this in his writing, if I recall correctly.