Chips are down - Texas Hold’em-playing AI running on modest hardware wins tournament

[gary]

On 31st January 2017, it was reported (http://spectrum.ieee.org/automaton/robotics/artificial-intelligence/ai-learns-from-mistakes-to-defeat-human-poker-players)that a supercomputer running
software developed at Carnegie Mellon University had won the
“Brains vs. Artificial Intelligence” heads-up, no-limit Texas Hold'em
tournament playing against four human poker pros. It won
by US $1,766,250 in chips over 120,000 hands.

Decades ago I saw a cartoon depicting two guys sitting at a bar
and one guy, looking depressed, was saying to the other,
"I could understand if I had been made redundant by an entire
computer, but I have been replaced by a single transistor".

Well the news today isn't quite that bad for human poker players,
but a new Texas Hold'em AI called DeepStack has emerged with
a fresh victory running on nothing more than the type of GPU chip
found in a home gaming desktop computer.





Article at the Institute of Electrical and Electronic Engineers (IEEE) here -
http://spectrum.ieee.org/automaton/robotics/artificial-intelligence/texas-holdem-ai-bot-taps-deep-learning-to-demolish-humans

[alpal]

Computers can already beat the best chess player easily -
so I'm not surprised.

[xelasnave]

Well it will never win at blackjack cause it wont be allowed to count cards or they will break its legs.
How amazing.
Its sucess probably turns on not bluffing or being a le to be bluffed and the ability to work out percentages and not drinking during a game.
Alex

[gary]

The 1997 win by the IBM supercomputer, Deep Blue, against reigning
world chess champion, Garry Kasparov, was a major achievement.

However, the recent tournament wins by these new computers against
poker pros is significant for a couple of different reasons.

There is a field of mathematics known as Game Theory (https://en.wikipedia.org/wiki/Game_theory) that was
pioneered by John von Neumann during the 20th century.

Now von Neumann was a genius by any measure of the word. A common
description given by the many contemporary mathematicians,
physicists, economists, engineers and bureaucrats who had dealings
with him was that he was the most brilliant person they ever met.

In 1944, von Neumann published a seminal book along with economist
Oskar Morgenstern entitled "Theory of Games and Economic Behaviour".

As von Neumann would tell people who asked him, within the
discipline of Game Theory, chess is not a game.

Chess, he would describe, is simply a well-defined form of computation.

Just like most of us by the age of 6 or 7 can master noughts-and-crosses -
that is we learn how to play to win or force a draw - there are, in theory,
a set of moves that one could make to always win or draw in chess.

As there are many more possible moves in chess than noughts-and-crosses
it makes the game exponentially more difficult and not even the greatest
master's know the precise set of moves to win or draw.

However, in chess, though you humanly may not be able to work out
the right move to make, for any position there exists an exact procedure.

Real games, von Neumann would describe, are like real life.
Real life consists of bluffing and tactics of deception and asking yourself
what the other man is going to think I mean to do.

So it is no surprise that in Von Neumann's mathematically-intense
book there is a chapter entitled "Poker and Bluffing".

The triumph for von Neumann was to prove that though there are no
set of precise plays you can make in a game like poker to ensure one
always wins or at least draws, there are strategies that can be derived
through mathematics to determine the best outcome.

The mathematics behind these strategies is extremely difficult and
some of the most brilliant mathematicians are still employed by
stock brokers and investment companies to apply Game Theory
to the markets and business.

Von Neumann himself was employed by the RAND Corporation
during the Cold War to model the interaction between the US and USSR
both politically and in the event of war.

If you have ever seen the movie, "A Beautiful Mind", Nash extended
Game Theory to what are called "non-cooperative games".

So these major wins by computers at poker - a true game in a Game
Theory sense - represents a new milestone for computation.

The second significant aspect of the win reported in the article was that
the processor used was not a supercomputer by today's standards,
but a far more modest system.