Last month saw the conclusion of a twenty-day long experiment at the Rivers Casino in Philadelphia. Libratus, an AI machine that runs on a supercomputer in Pittsburg, was built by two computer scientists at Carnegie Mellon. For the duration, some of the highest ranked world players played against the AI in no-limit Texas Hold ‘Em poker. A similar experiment was conducted two years ago where a version called Claudico was thoroughly trounced by human competitors; however, this time, the tables were turned, winning $US1.7 million from its four opponents over the course of the challenge.
Up against the supercomputer were Jason Les, Dong Kim, Jimmy Chou and Daniel McAulay: professional players with a plethora of experience. Kim discussed the experience, saying he had the feeling that the supercomputer-based machine knew exactly what cards he held in his hand. Chou agreed that it was a difficult challenge, describing it as playing more advanced versions of the human competitors.
Whilst its creators did not give all the specifics of how the system works, it was revealed that Libratus had learned how to play and win the game by playing trillions of games against itself in order to refine its performance – a level no human player could hope to match. In addition to this, the system also analysed the moves of its opponents whilst matches were ongoing, learning as it went rather than just relying on banked strategies and moves. However, as an extra level of difficulty, a third method was used, whereby, at the end of each day, patterns in the machines play were detected and removed, essentially meaning the human players could not predict the machines actions.
Whilst other event have shown that experts can be beaten by people with less experience – one famous example being where entertainer David Blaine played nine chess grandmasters simultaneously, beating more than half by essentially playing their moves against each other; however, this event marks a point where AI has managed to outplay humans at a game with a high variance of chance in addition to strategy.
The danger of course, as AI continues to develop and becomes more mainstream, is that human players who enjoy online gambling will come up against these unbeatable opponents in their games. In the future, will online gamblers be able to tell if a loss is due to a good opponent and a bit of luck, rather than a supercomputer programmed to play and win against all possible outcomes? It is perhaps reasonable to assume that, without assurances of fairness, the online gambling market will suffer as a result.
Fortunately, other technology could provide a solution to the issue. Blockchain technology has a number of benefits associated with its use which could go a long way into mitigating the threats that AI could pose.
As a transparent and immutable system, blockchain has the ability to provide an unchanging record of all results and transactions stored on it, available for anyone to examine. This would mean that the results of a ‘player’ on a game using the blockchain could be monitored and analysed to ascertain whether the win/loss ratio was suspicious or could allow player activity to be investigated if necessary.
In addition, by using a private blockchain with a requirement to register a user’s real ID alongside their blockchain ID, gambling providers would be able to control access to their platforms. Using this method, operators could also store video, audio or other records of users’ application processes on the blockchain to provide an immutable record in case of queries of identification and legitimacy. With the un-hackable nature of blockchain technology, players could also rest easily, safe in the knowledge that they are playing on a secure platform.