Will Machines Outbluff Us?

Twenty-three years ago, a computer Deep Blue won a formal chess game against Garry Kasparov, the world champion at the time. It was a six-game match, though, that eventually saw Kasparov as the overall winner, with one loss and two draws.

But the next year, the man met the machine for a rematch, and the (upgraded) computer won. It was a great day for artificial intelligence evolution and an eerie one for disquieted technophobes.

When Kasparov accused Deep Blue of cheating and demanded another rematch, the owners at IBM decided to dismantle the machine. But the research into AI that could beat humans in complex games continued thriving.

Games imply everything a scientist needs to review the computer against. Information collection, precise calculations, opponent estimation, decision-making, and strategy development to reach the desired goal are what make games what they are.

As such, they have always served researchers as the perfect field to test and demonstrate their machines’ capabilities.

Checkmate Consequences

Computers improving in chess did not lead to a machine Armageddon, though. The sensation at the time soon turned into common knowledge that machines could beat anyone, and, as such, they became excellent chess sparring partners.

They didn’t ruin the game, but they did change how it was played. Computers unveiled new strategies, and some would even claim they contributed to the game’s democratization.

Although the original machine was retired, new, simpler versions emerged, which made access to “machine training” available to anyone owning a computer or a smartphone.

However, as Kasparov himself recognized, the development of AI was not only about beating humans.

The contest might be over, but there’s still a lot to do in that field. One of the goals is to build a machine that could replicate the human learning process and strategize in a way a human would.

The way things are now, a computer can only point out your wrong moves and show you how to fix them. It cannot explain and specify the moves you (should) make. So it’s still far away from human-like reasoning.

Of course, it could always be the end of skill-based gambling if the machines truly start outplaying us at every corner. So if you fear the AI revolution, you can always just turn to purely luck-based games such as slots.

Nowadays, a variety of these are also mobile-compatible, with many downloadable apps to choose from. The bright side of this technological novelty is you’ll always know where it’s going.

The New Frontier(s)

In 2010, Kasparov declared poker the next frontier for computers to beat. And he was farsighted.

The experts from Carnegie Mellon University, the same research facility where Deep Blue’s development was initiated, started working on yet another machine. The new apparatus was designed to beat humans at their favorite gambling game. And in 2017, it did.

A group of researchers led by Tuomas Sandholm and Noam Brown developed a poker bot “Libratus,” which triumphed in the game of Texas Hold’Em over some of the best poker professionals in the world.

But there were further improvements to be made since the bot beat one opponent at a time, and poker is a game of multiple players.

However, this year marked a significant win for the machines. The same researchers joined with Facebook AI and created a computer that managed to beat five elite poker players in an incomplete information game of no-limit hold’em.

The two experiments involved an AI machine playing against five humans and non-collaborating five AIs playing against one human. Both tests proved machinal superiority. The computer, called Pluribus, outperformed humans through 10,000 poker hands.

Pluribus, similarly to its chess-related predecessors, was supplied with some interesting insights into the game of poker.

First, it devised its core strategy through self-play. That means it taught itself to play effectively by learning from its previous moves and working on minimizing the “wrong” or “regrettable” ones.

During real play, it continued adapting its core strategy to new situations. It didn’t need information on its opponents’ tactics, to study the poker champions’ biographies, look for information about efficient game strategies, etc.

Instead, it showed some unconventional moves and unusual tactics unknown to human players.

Reading the Poker Phase

Brown and Sandholm said to have chosen poker as the next step in AI research because it was superior in capturing the challenges of hidden information. As such, a game of poker could serve as a metaphor for many human traits and actions.

Take, for instance, a proverbial poker face or bluffing, or a patient but strategic approach to problem-solving that overcomes the scarcity of available information. It is a game of difficult “reading” skills and risk-taking, as well as a game of intuition and the ability to misguide your opponents. That is what machines outplayed us in.

If the machines can outplay us in such complex situations and even beat humans in what one of the professional players called “monster bluffing,” it raises some concerns. And these do not end in cheating in poker tournaments.

The developers claim that Pluribu’s findings can be applied to areas such as drug development, self-driving cars, fraud prevention, and cybersecurity, but the troubled skeptics could easily see past these benefits.

Can machines now out bluff us, a paranoid would ask? Or does their performance merely reduce bluffing to an efficiency calculation?

To dataists and proponents of AI, it is proof that some traits we deemed strictly human are nothing but a series of predictable and calculable patterns or chemical reactions within our brains and bodies. But how does that make it any less scary?

The Outplayed 

As far as our future is concerned, our biggest fear should not be whether the machines will outplay us or engage in a Judgment Day scenario to overtake the world from our feeble and corrupted hands.

They already did (outplay us), and they possibly could (take over), but not in a way imagined by a big Hollywood studio.

Our biggest fear should be that the development in the artificial intelligence field will make humans obsolete. As historian and writer Yuval Noah Harari explained in his books about our changing world, becoming redundant could be a fate worse than being exploited.

In the end, machines can beat us at games, but they do it without consciousness.

They don’t have the intention to beat us — they lack the will to do that. More importantly, their input depends on the will and wants of their human creator, and so will the consequences of their application.