Machine Learning and Adaptive Markets

Posted by

You’re watching:

Machine Learning and Adaptive Markets

user avatar

Lucas Greenwood

Elite Pro

Video Player is loading.
Current Time 0:00
Loaded: 0%
Duration -:-
Remaining Time 0:00
  • descriptions off, selected

Resume Video

Start from Beginning

Watch Video

Replay Video

10

You’re watching:

Machine Learning and Adaptive Markets

user avatar

Lucas Greenwood

POSTED Aug 15, 2017

Luc Greenwood delves into the complex topic of machine learning, first giving a brief overview on the subject including the history and application before discussing how it relates to poker.

3 Comments

Loading 3 Comments...

nittyoldman 7 years, 7 months ago

A few days ago, an AI developed by Elon Musk's OpenAI company crushed several pro gamers in DOTA. I don't personally play this game, but I did watch this matchup happen, and the players (best in the world) had absutely no chance. The difference in class between machine and human was enormous and this point was agreed upon by both the players and developers. The bot exploited weaknesses it saw in the players' games (so it was not simply following a so called video game GTO strat). The machine learned by playing heads up against itself "over thousands of lifetimes" of matches which happened in a couple weeks of real time. The developers additionally did not give the machine much of any input other than tell it the point of the game. In other words, the machine learned its adaptive strategy by itself through simulating everything that could be done against it and then finding the exploit for that and implementing the exploit instantly. The bot was loaded into a regular looking (size) desktop computer through a USB drive. I can't be sure on this, but I don't think it is out of the realm of possibilities that DOTA is less complicated than HUNL or even 6mNL. It scares me to think that we are on the precipice of online poker truly being dead. What's stopping them from creating a super poker bot that is capable of having the best mix of intuitive and optimal play that we've ever seen? I think it's only a very short matter of time.

SuicideSpree 7 years, 6 months ago

I really have to point this out since I DO play DotA.

What the machines excelled at wasn't actual DotA, but a hugely stripped down, exponentially simplified version of 1 vs 1 DotA which is itself much much simpler than real DotA. As such it was much easier than chess or GO. They are deliberately conflating two things: A) DOTA is much more complicated than chess, poker, GO: Absolutely true and B) The Machine won at DOTA, absolutely false. What the machine did was about as impressive as a bot beating a human at a 25 BB push fold game.

Lucas Greenwood 7 years, 7 months ago

I think these are legitimate concerns, To be optimistic I'd say that bots are still explicitly outlawed in online poker and if the penalty for using a bot is all your funds get seized and you are banned that makes using a bot extremely risky and likely a bad decision even if you have an elite bot.

Be the first to add a comment

You must upgrade your account to leave a comment.

Runitonce.com uses cookies to give you the best experience. Learn more about our Cookie Policy