I was interested about exactly what economic theories could be dislodged by this so I read the paper and... the article isn't particularly representative. The

I was interested about exactly what economic theories could be dislodged by this so I read the paper and... the article isn't particularly representative. The major finding send to be that learning algorithms are good at playing best-reply games, even though learning algos are based on memory but may have imperfect knowledge of the game and rules, while best-reply requires no memory but perfect understanding. As I understand it, they use that to find a strong inverse correlation in best-reply games between probability of finding equilibrium and complexity (or range of action), but barely mention economics in the paper.