Firstly, poker is a game of imperfect information. A game of perfect information means that all players are aware of all events, information, and knowledge in a game. This means that when a player makes a decision, they’re well-informed of all the events that have previously taken place. Things get interesting in a game of imperfect information like poker where an opponent’s hand remains an unknown variable. Things get even more interesting when an A.I. program tries to overcome the obstacle of imperfect information in a game of no-limit Texas hold’em. Libratus had to develop the perfect strategy of bluffing and not bluffing while also simultaneously randomizing play.
While on the surface, the AI victory over humankind in no-limit Texas hold’em may seem trivial it actually has a massive implication on the future use and development of AI. Since it’s now discovered that AI can do well with unpredictable behavior and incomplete information, it could potentially be used in detect and fight against fraud and cyber attacks. AI could also potentially be used to diagnose medical problems inpatients (although this does seem quite far into the future). Probably one of the most apt uses for AI would be utilizing it in complex business transactions.
Libratus’ victory in poker is not just a frivolous game win for a machine, it means that AI can, in fact, solve and interpret a range of problems and situations with incomplete information.