Can Artificial Intelligence(AI) bluff ?

Artificial Intelligence has passed a new benchmark. Now it knows when to hold and also knows when to fold.

Recently a computer system called libradis running on a supercomputer did what no other AI program had managed to do. It defeated three human world champion caliber poker players in a series of  No Limit Texas Hold'em games.

This is a big deal one of the ways we frame artificial intelligence, how does this AI match up against humans and frequently we use games to make this measurement.

Back in 1997 IBM's deep blue chess computer defeated world champion Garry Kasparov in a series of six games, Machine defeated man. Not all games are equal games like tic-tac-toe and Connect Four have a finite number of moves and we humans figured out the algorithms to solve these games.

That means we know the outcome of any game assuming all players are making no mistakes

Tic-tac-toe will always end in a draw assuming perfect play Connect Four will always see the first player win the game under those conditions.

But some games like chess have so many variations and possibilities that solving the game is a non-trivial task and others, like poker are even more difficult because perfect play isn't as easy to define.

A good poker player is aware of the statistical likelihood of his or her hand being the best at the table, but great poker players know how to analyze how their opponents play and then they capitalize on any weaknesses they might have.

For example, let's say we're playing Texas Hold'em and the three flop cards have been revealed, you hold two queens in your hands. The flop cards are a jack, a king and a queen that gives you three of a kind, Now what are the odds I have a better hand if I held two kings of my own or a ten and an ace, you'd be in trouble. But are the odds with me or against me. And could my betting behavior give you hints as to what I actually held in my hand.

It's not easy for humans and it's really hard for computers to master poker. There's a lot of psychology involved, going with a purely statistical approach might get you through the early stages of a tournament. But it's probably not enough to win the grand prize, how did the AI do it.

First libredis is a collection of three artificially intelligent processes one is called reinforcement learning, which is just what it sounds like. Libredis displayed trillions of games of Texas Hold'em against itself over and over. This is how human players start to get good though typically we don't have the time to fit in trillions of games in our days.

If you don't have an innate sense of odds playing games and paying attention will give you a feel for how likely in a given situation might be

On top of learning the game libredis developed its own betting strategy, it made unpredictable choices which kept human opponents off balance with no apparent rhyme or reason connected to its decisions, human players couldn't get a bead on whether libredis held legitimately awesome cards, or was bluffing it's transistors off.

A second system would analyze gameplay during actual matches. This allows Libredis to narrow down its place now catering it to go up against whatever humans it faced,

and a third system helped keep Libredis play style a surprise to other players.

It would analyze how it played throughout a day and identify any patterns that were popping up then it would actively instruct the other systems to avoid those patterns in future play. 

The Libredis of tomorrow won't play like the one you face today making it even harder for human players to find an advantage.

But perhaps the most interesting element to consider here is that Libredis could bluff.

It could bet in such a way that human opponents couldn't be sure their cards were superior.

A computer that can bluff and not be caught out could be useful in many situations not necessarily the deceptive ones.

If a computer can tell when someone else is bluffing, it could aid in business or politics. However a bluffing computer also reinforces the need for us to develop systems that require machines to explain their decisions.

How can you trust a computer's decision if you know it can lie to you, but I believe as long as we are careful and responsible this evolution in AI can be a huge benefit to us

What are your thoughts about computers that can Bluff ? You can share in the comments below.

Thanks for reading the article.