I see, thanks. Is any AI research center/company trying to model this idea, instead of applying RL?