Optimality and NP-Hardness of Transformers in Learning Markovian Dynamical Functions | Read Paper on Bytez