How Artificial Intelligence can change the way we play sports
We Humans
Can see, talk, listen, understand, comprehend, create, react, walk, run, jump and last but not the least reproduce more humans. What is that AI humanoid robots cannot do today in the above mentioned list except for reproduction(At least for now)? Looking at where AI is headed these-days its no surprise if AI actually did reproduce in a couple of decades (Do not get any ideas yet, I am not talking about the traditional way). I mean AI designing and programming another AI.
What is this article about?
As I was just browsing through the internet, I came across this video clipping of a robot named ATLAS built by Boston Dynamics doing a 180 degree jump and a backflip. SERIOUSLY!!!! A BACKFLIP!!!!!!
When I saw that video, I was thinking about the amount of training that the robot went through in perfecting that kind of a stunt which is still a dream for most of the people around the world if I am not wrong. Ok, if it can pull of a backflip with such an ease, it should obviously be able to move forward, back, and side with very less effort. Then I got this idea of using the same technology in sports which I believe can completely change the way we play sports and this is what I am going to explain in this article. But before I express my ideas I want to make one thing clear. I am in no way an expert in the Reinforcement learning and Artificial intelligence, I am just going to express the ideas that I have and they might as well be wrong.
How does a robot actually learn?
How many of you still remember the first time you rode a bicycle and the first time you hurt your elbows and knees while falling off of it? I sure do.
And how did we learn? We ride, we fall, we get back up and repeat until we get better at doing that specific task, in this case riding a bicycle. But how do we know we got better at that. So if we fall down while riding we can get hurt, and we do not want to do anything that hurts us. So we try not to get hurt again we constantly try not to repeat our mistakes and eventually we end up having a smooth ride from point A to point B. In a brief essence thats how a robot and reinforcement learning works together. Every time the robot falls of the bicycle lets say we give it a reward of -2 and every time it balances itself by not falling off of we give a reward of +2. Just like humans the robot tries not to hurt itself by gaining negative rewards and tries to gain as much positive rewards as it can by learning from its mistakes and trying not to repeat them. Even if any external forces acts upon it while doing the specific task it learns how to deal with them with out getting hurt. This needs an immense amount of training and it does not happen in one night. The robot contains a lot of sensors which collect the data and work together in order to make this possible.
Look at ATLAS
Below is the robot named ATLAS, a humanoid robot developed by an American robotics company Boston Dynamics. The pictures below just give a glimpse of what ATLAS is capable of.
This is just one of the robots built by Boston Dynamics. They still have a lot more. For more info check out Boston Dynamics. A few more companies which work on the similar tech are Hanson Robotics, Rethink Robotics, etc.,
OK, ALRIGHT, What are you trying to say???
What I am trying is, we can make a robot jump, run, and even crazier to do a backflip. What if we teach the robot to play as a goalie in Soccer? A referee in Basketball? A player in ping pong? You name it.
SOCCER
Just imagine that the soccer player training while trying to make goal against a humanoid robot which understands the pattern of the player and adapts the game from the person it is playing against, and gets better and better each and every time a player kicks the ball. All the set of actions that a robot is required to do in this case are move to the right, move to the left, move forward, move backward and ability to extend its arms to stop the ball if necessary. Do you think for a robot which can do a backflip with such an immense level of perfection, is it impossible? No. Definitely not.
We often see players running into referees in the games of soccer, foot ball, basketball, and what not. What if we train game specific robots which are trained to every minute rule and the process of decision making with respect to that specific game which has the state of the art computer vision and in turn can project the decision through a holograph. For example it will be able to showcase its decisions by lifting the red cards, yellow cards and projecting through a holograph the name of the team and the number of the player. One robot can be a referee, umpire, head linesman (or down judge), line judge, back judge, side judge, center judge. What could all the robot referees with their combined computer vision can do? Perfect decision making.
Cricket
How many of you know about the game cricket? I am sure some of you do, but for those who don’t, check out this link about an umpire in the game of Cricket. Basically its a referee for the game of cricket. All that the umpire does is take decisions, standing opposite to the batsman. These decisions include whether the batsman is OUT, whether the ball hit the six runs or four runs and many more, of which every single decision is based on the observation.
“Across several industry benchmarks, our computer vision algorithms have surpassed others in the industry — even humans,” said Harry Shum, Microsoft’s Artificial Intelligence (AI) and Research group EVP.
Just imagine a robot standing in the place of an umpire and making key decision which otherwise require a 3rd person who has to review the decision as human vision is not that capable of observing every details of the ball, but a computer vision can even calculate the angle and speed of the ball with up to n decimal points of accuracy. Is it hard for it to make decisions? The decisions which were once ambiguous for human could be obvious for the robots.
The game of Ping pong
How many times have you asked someone to play ping pong and got refused? Lets say you are a beginner in ping pong and really want to learn, but the person who you are playing with was so aggressive and always judging the way you play. You don’t have to deal with that anymore. Imagine that you have a robot which is trained so well in ping pong that you can even set the difficulty level of your play, and i am talking about the physical ping pong, not a computer game. How cool would that be, if that robot can make you a better player every time you play with it and its advanced computer vision analyses your style of play and learns your weaknesses and in turn gives you an option to work on your weak side of your play. There is a working robot arm which does this but I am not sure whether it has advanced analytics and difficulty level setting and all.
Leagues which are completely played by robots
We might even have RSL(Robot Sports League) which will be conducted across the globe, where companies from different states, countries and engineers who build these kind of robots can showcase their latest state of the art robot players in the competitions of Soccer, Football, Tennis, Ping Pong, Basketball, Kick-Boxing you name it. Things once thought as science fiction could become a reality. But after looking at the skills that the robots are developing and the tech that the companies are building, its no surprise if we see this happen in as soon as the next decade. And the funny part is we barely scratched the surface of the abilities of Artificial Intelligence.
Are we thinking too much?
If you ask me I would say NO. A big NO. It could very well be the reality and who knows, the companies might already be working on this kind of stuff. You know what, I 100% believe that if anyone working on this tech accidentally come across my article, he would be like “Why are people talking about the tech which we developed back in 2008 in 2018?”. If AI which is not even close to human intelligence can perform super human tasks which no humans can perform by themselves, what could Artificial Intelligence when reached and surpassed human intelligence can conquer? How many of you are exited to witness the Artificial Intelligence Singularity? For sure I am.
Robots ‘will reach human intelligence by 2029 and life as we know it will end in 2045’. This isn’t the prediction of a conspiracy theorist, a blind dead woman or an octopus but of Google’s chief of engineering, Ray Kurzweil.
Is it just the the beginning of an era?
or the end of an era?
