AI, soccer and chess
Recently I saw a video of a soccer game where three professionals scores goals on 100 kids. I noticed how this has been used in LinkedIn as a fun example of how quality is better than quantity. Here is the link: https://www.youtube.com/watch?v=s5f8hjzxmkA
As an example of the opposite I think the IBM built Deep Blue (first computer to beat a chess grandmaster) was amazing. Using lots of simple calculations, in a way that by most standards today wouldn't be considered AI. it beat Garri Kasparov in 1996. You could easily claim that Deep Blue had nowhere near an understanding of chess, but the game can be played in a different way by a computer. Without deep understanding of the strategy or opening theories it could win simply by looking at enough possible future positions and selecting the best.
There are several interesting observations to make about both these events even though they are very different. I think the first is about potential, the three soccer professionals may not be the best in the world but they are most certainly among the top percent of humans, some might be a tiny fraction better but as far as human soccer players they are about as good as they get. The 100 kids are nowhere near their potential, a few of them getting a little bit better would easily shift the results a year later. This was also displayed by the chess playing abilities of computers, today any desktop computer beats the best all the time, human chess grandmasters are not getting better but computers are not even close to their maximum potential.
While strength in numbers can be hard to understand when working computers imagine the kids in the soccer game replaced by small machines, then imagine a thousand of them filling the field. The skill level wouldn't need to be very high to always win. This is probably one of the most common way in which we underestimate the value of AI. we can simply produce more of it at an ever declining cost. The other way we underestimate machines is usually the 24/7 operation they can achieve. Many problems can simply be solved by being many enough working long enough, working around the clock and being many is already making up for many shortcomings of AI. When AI. grows smarter our biological needs will quickly be a hard limiting factor.
Another interesting aspect is how planning and strategy are powerful tools. The three soccer pro's and Garri Kasparov clearly demonstrate that even with inferior resources great efficiency can be achieved with appropriate long term thinking and understanding. This is even further demonstrated by implementing some strategic understanding into modern chess playing computers, allowing commodity hardware beat all humans at a fraction of the resource usage of Deep Blue.
AI. has not showed any of it's potential, it is usually not using efficient strategies, almost all of it is yet to come. Exciting, and a little bit scary, we will be as ants to them in the future unless we integrate or cooperate.
As an example of the opposite I think the IBM built Deep Blue (first computer to beat a chess grandmaster) was amazing. Using lots of simple calculations, in a way that by most standards today wouldn't be considered AI. it beat Garri Kasparov in 1996. You could easily claim that Deep Blue had nowhere near an understanding of chess, but the game can be played in a different way by a computer. Without deep understanding of the strategy or opening theories it could win simply by looking at enough possible future positions and selecting the best.
IBM Deep Blue
There are several interesting observations to make about both these events even though they are very different. I think the first is about potential, the three soccer professionals may not be the best in the world but they are most certainly among the top percent of humans, some might be a tiny fraction better but as far as human soccer players they are about as good as they get. The 100 kids are nowhere near their potential, a few of them getting a little bit better would easily shift the results a year later. This was also displayed by the chess playing abilities of computers, today any desktop computer beats the best all the time, human chess grandmasters are not getting better but computers are not even close to their maximum potential.
While strength in numbers can be hard to understand when working computers imagine the kids in the soccer game replaced by small machines, then imagine a thousand of them filling the field. The skill level wouldn't need to be very high to always win. This is probably one of the most common way in which we underestimate the value of AI. we can simply produce more of it at an ever declining cost. The other way we underestimate machines is usually the 24/7 operation they can achieve. Many problems can simply be solved by being many enough working long enough, working around the clock and being many is already making up for many shortcomings of AI. When AI. grows smarter our biological needs will quickly be a hard limiting factor.
Another interesting aspect is how planning and strategy are powerful tools. The three soccer pro's and Garri Kasparov clearly demonstrate that even with inferior resources great efficiency can be achieved with appropriate long term thinking and understanding. This is even further demonstrated by implementing some strategic understanding into modern chess playing computers, allowing commodity hardware beat all humans at a fraction of the resource usage of Deep Blue.
AI. has not showed any of it's potential, it is usually not using efficient strategies, almost all of it is yet to come. Exciting, and a little bit scary, we will be as ants to them in the future unless we integrate or cooperate.
Comments
Post a Comment