[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums

Late April Fools' Day...

Posted By: Michael Petch
Date: Wednesday, 23 May 2012, at 2:29 a.m.

In Response To: The next major revolution in backgammon bots (Michael Petch)

The Adaptive Chimeric Network is a hoax.

The original idea was intended for April Fools' Day. The problem with a hoax on April Fools' day is because people expect a hoax, so what fun is it? No malice is/was intended against XG2. One could have argued that GNUBG likely would have fared worse.

This existing thread on the revolution/evolution of bots seemed like a natural place to place it.

The name was partially inspired by my favorite author Douglas Adams. In particular his "infinite improbability drive". The term "chimeric" was specifically chosen because it is both a reference to a genetic hybrid (which does seem appropriate for an artificial intelligence), but more importantly the other definitions of a mythical beast, improbable, or fanciful.

Although the chimeric bot doesn't exist, I had hoped that people would realize that today's knowledge is incomplete. I personally believe we are at a point where there is likely NOT going to be any major "revolutionary" changes. The changes we will see are likely to be incremental with regards to bots understanding of the game.

With that being said, since our knowledge is incomplete, there is still the *very improbable* idea that a true revolution in analysis could find (as an example) that Backgammon is not a race, but a Backgame. It would be revolutionary if we had it all wrong and we shouldn't make the 5pt on an opening 3-1!

Sure it is an extreme, and highly improbable, but our bots today are based on a bias of myopic lookahead. Bots today can get confused with certain positions, often related (but not limited) to backgames. In these situations an astute human can jack up the cube in a money session, and extract a lot of value from the bot.

A revolutionary bot of the future, likely will see incremental (but not earth shattering) changes in the understanding of the game. Outside of the AI, I believe these changes would be more revolutionary:

a) a Bot that understands the "Why", and can explain it to a human in terms that a student of the game should be able to understand. b) a Bot that maintains profiles of players and adapts to the playing style of the opponent. I believe that it is important for bots to understand that beating a human or another bot also should take possible weaknesses into consideration when making a choice. c) a Bot that can be told before a match what general tendencies a player may have. Overly aggressive with cube, weak opening checker play, poor at backgames etc. A bot could then attempt to adapt to such play by making it more difficult for the human in situations where they may be weak.

Fundamentally, the main area of weakness I see today is that bots are static, they don't adapt, and they don't learn beyond what the original AI has learned. Bots generally assume everyone is a perfect player, the single biggest false assumption that is made IMHO. The original hoax had intended to have a bot that does all this too, thus the "Adaptive" term in the name.

The hoax is not intended to make fun of people, or to test gullibility. It was meant to liven up the discussion on BGO, and to get people thinking about what we know (or don't know) about Backgammon.

Messages In This Thread

 

Post Response

Your Name:
Your E-Mail Address:
Subject:
Message:

If necessary, enter your password below:

Password:

 

 

[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums is maintained by Stick with WebBBS 5.12.