- A poker-playing AI capable of thinking like humans do is quickly becoming a reality
- Poker robots capable of doing perfect math and pulling bold bluffs, all at the same time, are on their way
Talking about games where humans can still hold their own and even outmatch computers, poker is somewhat of the last stronghold.
Other games, such as chess and even Go have been pretty much solved, and human players are no longer capable of holding their own.
This is true even for the best out there. As Adam Kucharski, a lecturer in mathematical modelling, wrote in Wired this week, two decades ago, Garry Kasparov, the reigning world chess champion, lost the legendary match to IBM’s Deep Blue super computer.
It was the end of an era, as for a long time, it was believed that computers could never outsmart humans over the black and white board.
One fact that Kasparov lamented back then was that Deep Blue played exactly as you’d expect a computer to play. It was good, more than good, in fact, but it lacked the human approach.
In a way, chess playing programs brute-force their way to victory. With a limited number of possible outcomes and all the information readily available, it is as all about who can calculate further.
It is a battle that humans could never win – it was just a matter of time.
Players need to extrapolate from the information they have and build on that data set to create the best guesses about the information that isn’t laid bare – such as defining opponents’ hand ranges and interpreting their actions.
However, the times they are a-changing, and poker robots are starting to catch up, getting to behave exactly like Kasparov wanted – more human-like.
Artificial intelligence taking on poker
After being able to solve games such as chess and Go, researchers have turned to poker as the ultimate challenge for an artificial intelligence.
WATCH: Poker bot takes on the humans.
Because of everything mentioned above, poker represents a true challenge for the AI, as it requires a real thinking approach. The software actually has to make decisions based on what they have and the best decision isn’t always clear-cut.
Back in 2015, University of Alberta scientists came up with the bot called Cepheus, which basically solved the limit version of heads-up Texas Hold’em. This wasn’t that big of a surprise, though, as the limit variation has a limited number of betting options, making it much easier to solve.
However, earlier this year (2017), the unimaginable happened. A poker AI called Libratus managed to beat several high-profile poker players in a No Limit Texas Hold’em heads up match.
The breakthrough that came with Libratus is a very significant one, as it clearly showed that modern poker robots are capable of thinking out of the box, so to say.
Bluffing, which has long been considered such a human trait, has become an essential part of the AI reasoning.
Sometimes, a bluff is the optimal decision in a given situation, and robots aren’t afraid to pull the trigger.
Perhaps the science isn’t quite there just yet, as solving the poker problem when there are more players at the table is infinitely more complex.
However, all signs are indicating that it won’t be long before we have an AI capable of outplaying human opponents on a regular basis.
Now that PokerStars also seems to be tackling the idea, looking to develop their own poker-playing AI, the progress could be even faster.
Some years ago, the idea of an AI capable of crunching the numbers like a computer but still thinking like a human when necessary seemed like a pipe dream. Today, not so much. Poker robots capable of pulling bold bluffs and keeping you honest are on their way.
Want to try an online casino?
Choose an approved casino from our carefully selected list. VIEW CASINOS