This is a discussion on Libratus within the online poker forums, in the General Poker section; Ok, all heard how cool it is and how it has beaten pro's (though I have my version) but is it any chance to try
Ok, all heard how cool it is and how it has beaten pro's (though I have my version) but is it any chance to try it? Just to test is it really so cool or it has won only once (accidentally)? Since noone tested it second or third time.
If not then for what it was created?
Yes, Libratus sounds incredible, however, does it exist as an independent and playable entity?
To build the program (Wikipedia) it took 15 million core hours of computing and during the one tournament it played it used another 4 million hours of computing to analyze the in-tournament play. I presume this is all via super-computer so the program is not set up to play anyone, anytime.
However, I agree it seems a shame Libratus has not been put to more use/challenges. That said, has it simply been accepted with this much AI/computer power it could run circles around any human regarding NLHE? So it would likely be a waste of time and resources even if they sat the likes of Stephen Chidwick, Justin Bonomo and Steffen Sondheimer down to face it as they would also be chewed up and spat out within a few days?
I am sure there are countless research scientists who would love the opportunity to have several days of that much computing power at their disposal and it might simply come down to the idea they do not need to repeatedly show they can beat humans at Hold'em.
Ok, I got it. I thought maybe they sell it as challenge like: only $90.
My version was: players were tired. Yes, 120 000 hands are enough to prove something (especially that players were tired) but pro's played with one stack. And if one loses all lose. Creators of Libratus (or newsmakers) perform it as they beat all people. They don't give to try it second or third time. While if computer lose they just go and build new one.