Robots Aim To Top Humans At Air Hockey

An anonymous reader writes "You probably knew that the Deep Blue supercomputer beats chess masters, and that last weekend a software robot defeated four poker champions. But you may have missed this one: a GE Fanuc robot is taking on humans at air hockey. The robot is powered by a special PC-board that can instantly switch between 8-bit and its 32-bit modes. The 8-bit version lost to most human players, but the 32-bit microcontroller has defeated even the best human air hockey players by a ratio of three to one."

I for one welcome...

[gnosi]

Oh that is just getting so old. In this context however it could become so real.

On to the real subject...

"If droids could think for themselves we would not be here"

The day is coming when most if not all the routine and skilled functions of life will be carried out better by robots than by humans.

The last bastion for the human mind will be pure abstract thinking.

I do not even pretend to know what that new day will bring to the meaning of mankind when computers become better than the human mind at pure abstract thinking.

--
It is all in the sig. The rest is just window dressing

Re:Boring...

[elemnt14]

Although I agree that seeing hockey teams full of robots would be somewhat exciting (only to see them go up in sparks by a hard hit), I have to say that i personally would enjoy watching the sport with human players. Robots can not think for themselves (well, not yet anyway), so you would not see strategy that can change due to the changes on the field, or see some really great shots that only a human could pull off.

Re:I for one welcome...

[khellendros1984]

That's what's commonly called a "singularity". A point that changes things so much that it's impossible to predict its effects with any certainty. Depending on how we handle things, it could be a violent occasion worthy of a large-budget action sci fi movie, a quiet fade and disappearance of humanity, or a metamorphosis of humanity into a new form. Or it might not ever happen. *shrug*

Am I the only one who wants to take it on?

[Dr. Spork]

I would love to see this in an arcade. I'd pay a dollar to play the arm - bring on the 32bit mode! If they could make the arm fold itself out of the way while two people are playing, this would make an excellent arcade machine.

What's more, if the arms were standard and mass-produced, there's a great excuse for a little coding competition: Whose program will win when it's robot v. robot?

Lots of cool AI, artificial learning and computer vision would go into it, and the result would no doubt be fun to watch!

Re:Robots also top humans at arm wrestling..

[nfk]

"Honestly, it's not as if some robot is paintaing abstract art or writing poetry here."

You picked a couple of interesting examples; I'm sure robots could paint abstract art and write poetry that would match some of today's offerings by human beings. Anyway, I have no idea how complex it is to program a robot to play air hockey, and whether it involves only strength and precision, but there was an idea I read in a book by Douglas Hofstadter that I find amusing: artificial intelligence is always defined as whatever a machine cannot do yet.

Re:Yeah, and?

[RzUpAnmsCwrds]

It's almost cheating for a machine here. Much of the game is based off of your opponents meat-weaknesses and reading their hand from their faces. The computer doesn't have a face and is using pure probability. Even the best poker player cannot read them. So unless they're better at math than a computer and and their poker prowess isn't based on reading people, they're never going to win. It takes the game out of the game really. The perfect odds playing machine against the perfect odds playing player would come out even.

If you think that this is how poker works at a professional level, you don't know very much about poker.

Tells are a small factor at a professional level.
Good players already have a firm understanding of the pot odds and expected value of a call.

The major problem with writing a good poker program is that it can't be exploitable. If the program is too "tight", an experienced human player will realize this and can always raise with trash (because the program will fold), with a minimal chance of getting caught. If the program is too "loose", the human player can play good hands far more aggressively, knowing that the program will call.

This is a very general example. In reality, poker strategy is far more nuanced. But the basic problem that computers face is not being predictable. And, no, playing randomly doesn't help - it merely substitutes poor play for predictable play.