10th Annual RoboCup

Aryabhata writes "As soccer fever continues the 10th RoboCup also got to a start. 400 teams fight it out in 11 different leagues including onces designed for humanoid to four legged robots. "The organizers of the tournament hope that in 2050 the winners of the RoboCup will be able to beat the human World Cup champions".
Beyond the novelty value, the cup enables 2,500 experts in artificial intelligence and robot engineering to meet and test their latest ideas. The championships is followed by a 2 day conference where the teams can dissect their play and work."

Comments

2050

[rbarreira]

In 2050, the question "Is a team of robots capable of beating a team of humans in football?" will be irrelevant (or at least very different from what it is now). What is a human? Do "cyborg-like" modifications to one's body allow him to be considered human? Etc etc...

Re:2050

[mustafap]

Well lets look at what current footballers are like. Mostly dumb, and very vain.

We have dumb robots, so we just have to work on the 'vain' bit over the next 44 years.

Re:2050

[Alsn]

You wont be playing in the world cup as a footballer if youre a dumb, vain one. Regardless of what the average is the "best" will always be the smarter/stronger ones.

Re:2050

[Anonymous Coward]

Robots already have a degree of self awareness. Position sensors, battery charge monitors, etc are all designed to let a robot know about itself in relation to the world. As we develop more sophisticated robots, they will require a greater degree of self awareness. Right now, industrial robots are basically programmed at the "goto position x1,y1,z1; close gripper; goto position x2,y2,z2; release gripper;" level. If you want them to work at the "Pick up part X from conveyor belt; dip part in solvent tank;" level, the robot is going to have to be able to coordinate vision and arm motion. In other words it will have to have a greater degree of self awareness. When you get into higher level stuff (same robot, multiple tasks) the robot will have to keep track of which tool it has, what loads it is capable of manipulating, etc.

In short, the more self aware the robot, the higher the level of abstraction you get in assigning tasks to it.

Re:2050

[mustafap]

I wonder what people 45 years ago would have made of the technology we use today? Why shouldn't technology 45 years in the future surprise us as much?

Press release for Sango and Ami

[technoextreme]

http://www.cmu.edu/PR/releases06/060608_robocup.ht ml The competition is not just about robots preforming soccer. There are two other events that are completely unrelated to soccer. One event is search and rescue and the newest competition involves domestic applications. PS. This is probably the only time I will ever watch a soccer event.

Robots will still have the advantage

[technoextreme]

In 2050, the question "Is a team of robots capable of beating a team of humans in football?" will be irrelevant (or at least very different from what it is now). What is a human? Do "cyborg-like" modifications to one's body allow him to be considered human? Etc etc...

Well think of it logically. With robotics there is no limit to how powerful you can make their sensors and motors without causing harm to anything. It's just a matter of technology. WIth humans you can't just start attaching parts in a slapdash manner. That arm which can lift a couple thousand pounds will rip the socket and pretty much kill you if you use it to the potential. Robots don't have that problem. PS. If you can guess where I learned this Ill give you a cookie.

Re:Robots will still have the advantage

[RsG]

Let me guess, you are now typing with one hand?

He is, but that has nothing to do with robotics. :-P

On a more serious note, given that the only part of the body that's really needed for an otherwise mechanical entity to be considered a "human" cyborg is the brain, who says that robots will have the advantage?

I'd say a 800 pound cybernetic football player with a metal body and a human brain (augmented by microchips) would have an advantage over the same metal body governed by a computer. After all, computers can't improvise as well, and I suspect that football is simple enough that the extra processing power of a computer isn't needed.

Of course, that's assuming that we can't make an AI that is equal to a human mind in that regard, but I'd actually think that such an AI would be harder to develop than the cybernetics involved in creating a robotic body with a human brain. After all, a robot physically capable of playing football would probably be possible with modern technology, and I'd imagine that we'll have the technology to turn nervous signals into computer data before we have the technology to make a software program think like a human.

Re:Robots will still have the advantage

[roman_mir]

well, if you don't have a limit on the poundage (excuse me, weight,) then the robots can win by default by dropping a trillion ton anvil on the opponent's goal.

But seriously, will we also allow instant communications and radio-coordination between players on the field? This can make all the difference. Of-course if the cyborgs can also do this then the outcome is not certain.

However if we also don't bother too much with the definition of a player, then the robot team can just roll out a cannon, load the ball into the barrel and shoot straight into the opponent's goal.

Re:Robots will still have the advantage

[RsG]

True, but all that assumes that whatever sport the robots are playing has no rules.

If we said today that the only requirement for playing pro sports was that the player be human, then what would stop the athletes from doping themselves to the gils before play? I would imagine that if there was a "cyborg league", there'd be some sort of rule set for what equipment is allowable on the player, equivalent to the rules we have now about steroid use.

power

[roman_mir]

Actually I don't doubt that from point of view of mechanics and programming robots that beat humans in all kinds of sport can be built, but will these robots have power cables running to them? Or will the robot team have to replace the batteries on each robot every 10 minutes, that is what I would like to know. How will these robots be powered? For the longest time it has been a tradition in sci-fi stories to have autonomous robots that don't need to recharge every 10 minutes, it is assumed that in the future the problem of battery capacity is somehow resolved. Some robots use built in fission plants, some use fusion plants, some use batteries of unexplained nature, but they can run for days or even years without recharging. If we could actually do something like that, then the life on this planet could become interesting again.

Re:power

[YA_Python_dev]

I was in a team that built a humanoid robot for the robocup a few years ago.
Yes those robot have all the batteries on board and they have a very short duration: the absolute maximum was more or less 20 minutes, but some robots had batteries that lasted less than 60 seconds (the duration required for a single kick: the competitions in the humanoid league are pretty simple for now ;-)).

Re:power

[roman_mir]

But a human cannot push 3 tons of steel at a constant speed of 150km/h for hours and hours and in fact for as long, as there is fuel.

Sure, living organisms are very efficient at what they do, but we are not talking about living organisms.

Calculating chemical power requirements for living cells and organisms that comprise of living cells is not that different from calculating power requirements for a 100W lightbulb. Sure, living organisms are efficient in using chemical energy, but these chemicals in themselves are not the best storage mechanism for machines, that need lots of power instantly. That's why excavators burn petrol byproducts for power and not potato chips.

By the way a car for example can in principle use solar power directly, but a human cannot. A human has to wait for a plant to use the solar power and then a human can eat the plant. This is an inderect way to retrieve power and it is not the most efficient way.

2050? Yeah, right.

[Anonymous Coward]

Maybe AI will advance to the point that a robotic team would win (more easily if they 'share thoughts'), but who foresees that in 40 years there will be robotic machines fast, light and flexible enough to play with humans? Soccer is a contact sport (maybe not as much as rugby, I know). Would you play against a plastics/metal body? Would any country or soccer club risk their players? Some of these guys will be worth more than the robots.

As a goal to encourage scientific progress it may be a good idea. As a practical matter, I don't think it is.

Re:2050? Yeah, right.

[CrazyJim1]

Would you play against a plastics/metal body?
Well after that yellow card, things are getting rough out there on the field. And look, the robots are putting in their hooligan who used to be a battlebot.

The question still stands

[giorgiofr]

Can a working human mind ultimately be reduced to a complicated algorithm? Will we be able to emulate it, given the necessary computing power?

Re:The question still stands

[Kjella]

Can a working human mind ultimately be reduced to a complicated algorithm? Will we be able to emulate it, given the necessary computing power?

I think a more interesting question would be what would happen if we assume the existance of such a working brain. Our experience has been that when we can make it work, a computer is damn fast and damn precise at it. A computer can never die, never forget, be an expert at almost everything. Imagine a computer that could comprehend and set in system all the information in the Library of Congress. That could read up and be a doctorate in every subject. That could learn all languages of the world.

It could very well skip several steps ahead of man simply by being able to see all of it at once. It would not only be an expert in the field, it'd also be an expert coder - that can create and run simulations and test scenarios at speeds we only can dream of. It can corrolate data and try to find logical explainations for them - to find new causations we haven't seen yet. Basicly the scientic method on steroids.

I have read a few books in my life, I have a Master's degree and I speak three languages. But when it comes down to it, I have only a microscopic part of all human knowledge. Even the true geniuses may be extremely impressive, but they too operate within their own little niche. It's simply a limitation of being human.

Another thing that would be very interesting would be ethics and morality - I doubt you will be able to code in anything like that into a thinking brain. In order to function it will have to be create it by itself to apply to the situation.Will it evolve one? Is it natural? Will it go Skynet on us?

An even stranger question is the goal - to what end would a computer brain exist? To reproduce? That's a rather stupid concept for a computer, and only exists in biological life because producing offspring is the only way to survive. Could it, like human-ethics try to do, reason itself to a purpose? Does it need a purpose? Other creatures have instincts but how could a computer have that?

In practise I think we're ages away from creating a functioning AI. But I think all it takes is that magic spark, the creativity and the ability to evolve. From there it'll take itself to where it needs to go. Simulating behavior, even very complex behaviour is good for creating simulated intelligence. But I don't think it brings us much closer to a truly intelligent computer.

For the 10000th time,

[Ireneo Funes]

It's called FOOTBALL.

Re:For the 10000th time,

[MamiyaOtaru]

Someone should inform the Socceroos.

Games as an AI research platform.

[Black Parrot]

> Beyond the novelty value, the cup enables 2,500 experts in artificial intelligence and robot engineering to meet and test their latest ideas.

FYI, though RoboCup has been around for a long time, the past few years have seen a sudden surge of interest in the use of games as a platform for AI research. In addition to the now vast literature on RoboCup there are several new conferences dedicated to AI and games, usually covering non-RoboCup topics. Grep the net for Artificial Intelligence in Interactive Digital Entertainment (AIIDE), Computational Intelligence in Games (CIG), and the Special Session on Games at the Congress on Evolutionary Computation (CEC). I've seen some of the proceedings on line, and you can find some pretty interesting papers about applications, if you're interested in that sort of thing.

Re:Games as an AI research platform.

[Avian visitor]

FYI, though RoboCup has been around for a long time, the past few years have seen a sudden surge of interest in the use of games as a platform for AI research.

Some time ago I had some interest in AI research. I visited an international Robocup event in Slovenia because I thought I might see some interesting concepts being used there.

I talked with several teams and I was quite surprised when I saw how primitive the their programs were. They basically had thousand nested "if" statements. No neural nets or anything remotely advanced. When I asked them if no one uses such things they said that there were some experiments but "if" statements just work better in practice.

The Robocup competition I saw there didn't require any special AI or engineering skills from team members at all. All teams had identical robots that were mass-manufactured by some company. Image analysis software (for determining the position of the ball and robots with a video camera), communications software, etc. were also already written for them. The software didn't even have to be written with embedded systems in mind because it ran on a powerful PC (the robot itself was only a radio-controlled black box).

As far as I know, the whole tournament could have been played entirely in software. The little robots were there only for the audience to see something.

Re: Games as an AI research platform.

[Black Parrot]

> I talked with several teams and I was quite surprised when I saw how primitive the their programs were. They basically had thousand nested "if" statements. No neural nets or anything remotely advanced. [...] The Robocup competition I saw there didn't require any special AI or engineering skills from team members at all.

Kind of like the "beat Kasparov" approach to chess-playing AI. When competition is involved, people resort to hardware and hacks. That says a lot about the state of AI after ~50 years of research!

Yeah, right

[Clovert Agent]

"RoboCup 2006 is the first step towards a vision," said Minoru Asada, president of the RoboCup Federation.

"This vision includes the development of a humanoid robot team of eleven players, which can win against a human soccer world champion team."

Even granting the somewhat unlikely prospect of a robot team that can match the skill and tactical experience of a human side, I can't see them overcoming the obvious safety problems.

Call me when Minoru Asada is willing to demo what it's like to be slide-tackled by a robot, and I'll reconsider.

robocup 2006 home page

[martyb]

The BEEB's blurb was interesting, but here is a link to the RoboCup 2006 home page

There are pics, background, schedules, leagues, etc.

Re:Silly northamericans.

[roman_mir]

Just thought of something. The robots that play this game, are they playing with feet or with hands? Can we even apply those terms to robots' body parts? They are definitely playing with lower appendages, maybe they are lower manipulators? So maybe robots are playing ManipulatorBall or Manipuball. Eewww, sounds dirty.

Re:Silly northamericans.

[zippthorne]

Me either, but it certainly is odd to call a sport football in which the ball is a nearly perfect sphere and not in any way shaped like a foot.

It's over already

[SmilingBoy]

"As soccer fever continues the 10th RoboCup also got to a start."

It got to a start four days ago and finished at about the same time as this story was posted!

Anyway, I was quite impressed - watched lots of it through an internet live stream. The humanoids still have a way to go, but in a few years, it will look much better.

There are lots of videos on http://www.robocup.zdf.de/ (in German).

SmilingBoy.

Re:Wrong. StrongAI will still have the advantage

[vertinox]

Well think of it logically. With robotics there is no limit to how powerful you can make their sensors and motors without causing harm to anything. It's just a matter of technology. WIth humans you can't just start attaching parts in a slapdash manner. That arm which can lift a couple thousand pounds will rip the socket and pretty much kill you if you use it to the potential. Robots don't have that problem. PS. If you can guess where I learned this Ill give you a cookie.

Actually, you are falling into the fallacy (as Ed at the Singularity institue calls it) anthromorphic assumtion. We assume that we as humans will be the same now as we are then and we assume that robots and humans have the same limitations today.

Lets say in 2050 we have a robot that can rip a regular humans arms of. Chances are that will be a given.

But whats the difference between a robot and a robot with a human brain inside of it that can rip a normal humans arms off? (think Ghost in the Shell)

Well besides the life support system and neural interface, changes are the robot and the cyborg are on equal footing. Heck... You could not even get rid of the human body, but have a neural interface to the humans mind while he sits in a room somewhere and controls the robot remotley.

But... The robot or should we say... AI (if they pull it off) will that the advantage over the human speed and tactical wise.

Lets say the goalie human cyborg being on equal strength of the robot can only guess and predict X amount of moves in X amount of seconds in determine where and where the robot is going to optimally kick the ball (and from which direction). If his mind is still organic without enhancements, he'll have to think at the speed of his synapses (1 to 120 meters per second) even with an electronic interface to robotic eyes or radar or whatever cyborg soccer players use to see in 2050. Then he has to use those neurons to fire off and communicate with his robot body.

The robot, having the disctinct advantage of being electronic through and through can use his computing power at the speed of electrons running from his eyes to his CPU and to his arms (which is near speed of light) and has the speed. Not only that the average human mind can not simply make more than 5 guess on the next best move. (Kasaprov the chess champion can do something like 12 next best moves).

So while the human goalie is trying to guess what the AI is going to do, the AI has already formulated all possible moves and has found the move he can make that will have the highest percentage of scoring a goal. Not only that he communicated this wirelessly to his robotic teammates and they are doing moves together in real time. That would be really hard for humans to do.

However, a cyborg human with AI assistants will have a better chance of finding the next best move.

Now of course you may wonder, how do we interface a human with a machine in the first place? We are doing it today small scale and the human body is fragile, but what is to say that in 45 years that they have figured out safe ways to interface the human mind to a machine and can even build life support systems that no longer needs the human body to keep a brain alive.

We have those kind of life support systems today and could make a complete "brain in a jar" if we really wanted, but it wouldn't be fun for the brain since we don't have thing to interface for communication.

Otherwise think of it like this... Today is 1906.

If we compare all the changes to life that happend from 1906 to 1946... We heavier than air flight, atomic bombs, mass production, radar, trasnatlantic flight and rockets. Who is to say that by 2050 we are going to have the limitations of what we have now when we are dealing with robotics and human interfaces.

The real question is...

[Alicat1194]

How well will the robots be able to fake injuy?

Ow! Ref! The human just kicked me in my power coupler! The pain! The pain! ::convincing limp::...

Starbucks' RoboCup?

[antdude]

SomethingAwful's amusing doctored image with StarBucks' RoboCup!

"Careful, the beverage you're about to enjoy is extremely hot, Creep." --Jonah in RoboCop Archive forum thread.

Re:Silly northamericans.

[roman_mir]

I am Canadian and you are an idiot. Soccer is a later invention than Rugby Football, that's first. Secondly I am only talking about North America and what this sport is called here, you dimwit.

MegaHAL is great

[Nurgled]

A bunch of friends and I used to run a bunch of MegaHAL bots on an IRC network. A couple of them ran for several years. We let them talk to one another on channels sometimes, with appropriate rate-limiting. After a while the longer-running ones started to seem more and more insane as their databases grew larger and larger. Eventually one of them exploded and corrupted its database somehow; we couldn't be bothered to fix it, but it was a fun experiment while it lasted.