Poker playing robots – more realistically, poker programs that utilize artificial intelligence, or AI, to make their decision – have been under development for the last 20 or so years. And with each ‘bot,’ they have gotten better at the game. Now comes the development that the AI that defeated four top professional poker players back in 2017 is now working for ‘The Man’ – the U. S. military.
‘Libratus’ Becomes a Military Tool
A recent article in WIRED magazine from journalist Andrew Rush reports that the poker bot ‘Libratus,’ which was developed by the researchers at Carnegie Mellon University in Pittsburgh, is no longer playing poker. The head professor in developing ‘Libratus,” Dr. Tuomas Sandholm, recently founded his own startup company that would adapt the AI that he has developed to government usage. Sandholm always said that the types of things that the AI utilized by ‘Libratus’ does – complicated game theory calculations on an infinite scale – could be used in other settings and now the professor is proving it.
Part of the deal with the U. S. government is to use ‘Libratus’ in war games simulations and strategy planning sessions. Sandholm is making a nice fee for using ‘Libratus’ in this manner – a two-year contract for up to $10 million with the U. S. Army. Apparently ‘Libratus’ is going to be a key linchpin in something that Rush states is the Pentagon’s “Defense Innovation Unit,” but it is difficult to ascertain what it will actually do because Sandholm isn’t saying too much.
Sandholm isn’t happy with his foray into military defense, according to Rush. He has also set up shop to utilize his different AI to be able to operate in civilian settings including electricity grids, sports and videogame opponents. And before you think that the U. S. government is entering into a dark arena, Rush accurately points out that several other nations, including Russia and China, have already accelerated their pace of research into usage of AI, with China hosting a major war-gaming conference in 2017 that saw humans and AI battle it out in the virtual world.
‘Libratus’ Tough Player to Beat as Humans Learned
Sandholm and the experts at Carnegie Mellon were working on their poker playing AI way back in 2015. That year Carnegie Mellon rolled out ‘Claudico’ as their best poker playing AI and brought in four poker professionals – Doug ‘WCGRider’ Polk, Dong Kim, Jason Les and Bjorn Li – to play a two-week series of heads up no limit Texas Hold’em. The rules were simple: whoever came out of the series of matches up at the end of the two-week period would be deemed the winner.
Over that two week span, the four poker professionals were able to grind out a roughly $750,000 win, but the statistics weren’t conclusive on who won. Li racked up a major amount of the money won by the quartet, but the others were within a range of being a statistical tie. Thus, the masterminds at Carnegie Mellon went back to the lab to create an even better AI.
That was ‘Libratus,’ who came out in 2017 for another challenge against four more professionals, Daniel McAulay, Jimmy Chou, Kim and Les. In this two-week battle, ‘Libratus’ proved to be more than able to handle the task as it routinely handed the pros losses (some rules changes may have allowed this to take place). By the time the smoke cleared, ‘Libratus’ had earned roughly $1.7 million to hammer the four professionals.
Sandholm has admitted to Rush that he believes the concerns about military usage of AI are “overblown.” If the results of ‘Libratus’ can be duplicated in other settings, however, the only question that remains is…what date did Skynet go “live?”
