Bender Rodriguez has a defender on him, unleashes
with his right foot! Gooooooooaaaaaaallllll Bot-celona! Hello everyone, Julian here for DNews. Artificial
Intelligence is the holy grail of computing and robotics. A machine that can intelligently
analyze and manipulate its environment to achieve a goal is no small feat, and at first
it was believed that the ultimate challenge for AI was winning a game of chess. It makes sense at a glance: chess masters
are brilliant strategists who carefully analyze and react to their opponents in order to win.
If a computer can beat one of them, then surely we’ve cracked this nut, right? Well a computer has beaten a chess master.
18 years ago. So why hasn’t Skynet taken over? Because chess is not the perfect problem
for AI that we once thought it was. Chess is actually relatively easy for a computer.
Chess is only against one opponent, and everything happens sequentially. A computer knows that
when it makes a move, exactly what it intended to happen will happen. And even though there
are millions and millions of possible moves, the slow pace of the game means the computer
can calculate exactly all of them and cherry pick the best one. Basically winning chess
for a computer is less intelligence and more brute force calculations. But in 1997, the same year Deep Blue defeated
chess champion Gary Kasparov, another tournament was started, and it involved a game much harder
for robots. I’m talking of course, about soccer. Yes, soccer. The tournament is called the
“Robot Soccer World Cup,” so soccer is the correct term. Plus soccer is old British
slang for “Association Football,” so I don’t want to see a bunch of comments about
how it’s called football. Anyway, soccer poses problems for AI that chess never could.
Instead of reacting to one other player, there are entire teams. It’s not turn based, but
constant and dynamic. The robot cannot predict exactly where the ball will go when it kicks
it; it could take a strange bounce or another robot could intercept it. That means the robots
have to react to unexpected events, and they have to do so quickly, none of this faffing
about while it decides what to do. Plus robot soccer is tractable; it can easily be recreated
over and over anywhere in the world. It’s not as challenging as creating a robot that’s
almost human, and it’s not too expensive. It also helps that soccer is the most popular
sport in the world, so robot soccer would probably gain some attention. And indeed it has. When the first RoboCup
was held in Nagoya, Japan, there were 38 teams from 11 countries. Last year there were 358
teams from 45 countries. The number of categories has grown immensely too. Originally RoboCup
had only 3 leagues; small robots that used an overhead camera, medium robots with their
own sensors, and AI playing soccer solely in a simulation. Then in 1999 RoboCup introduced
the first league with legged robots. The robots they used? Specially programmed Sony AIBO
dogs. Yes those little plastic dogs your aunt got you for christmas because she had no idea
what to get you actually served a purpose advancing science. Now the legged divisions
are humanoid, and can either be a standard platform or entirely custom built. They’ve
also branched away from soccer, and have competitions for search and rescue robots and robot dancers.
Most importantly though, all the robots are autonomous and independent. The eventual goal
is to have a robot soccer squad capable of beating the human world cup champions by the
year 2050. Watch out Messi, Robo-naldo is coming for you. Competition is one of the best ways to drive
innovation, and when brilliant dedicated people come together, they can achieve amazing things.
In the same way the United States Air Force is powered by Airmen and fueled by innovation,
and we’d like to thank them for sponsoring this episode.