From the course: AI Algorithms for Gaming

Some history as motivation - Python Tutorial

From the course: AI Algorithms for Gaming

Start my 1-month free trial

Some history as motivation

“

- [Instructor] Before we talk about the big blue chess-playing computer, let me tell you about Supercomputers. You may already know this, but to succeed at the challenge of beating a human chess champion, we need way more computational power than just a desktop computer. So, Supercomputers are computers with much higher performance specs than regular desktop computers. They usually have many, many CPUs in blocks known as compute nodes, which perform massive parallel computing. They also have huge specs for storage and RAM. All of this computational power comes with a price. Supercomputers are very costly and highly demanding in energy and space. As examples of Supercomputers, I can mention those by Cray Research. This picture shows the HLRN-III Cray system at Zuse Institute in Berlin. So what are the applications of Supercomputers. Well, Supercomputers have been used for Weather Forecasting, Aerodynamic research, Astronomy, Biology, Pharmacology and most of these applications involve highly-demanding simulations. In addition, Supercomputers have been used to tackle problems by brute force, where computation would take too long on a regular computer. The picture shows a Blue Gene Supercomputer by IBM. Now let me tell you some interesting historical details related to Deep Blue. Deep Blue is a chess-playing Supercomputer that was developed by IBM with the purpose of winning against a human player, a big milestone of Artificial Intelligence at the time. The development of Deep Blue started at IBM in 1989, coming from an earlier project from Carnegie Mellon University After countless hours of research and development, Deep Blue finally defeated the Chess World Champion, Garry Kasparov, in a series of matches between 1996 and 1997. So, what was the software running in Deep Blue. Well, Deep Blue admittedly used a brute force approach, and as you'll see, this is not necessarily bad. Furthermore, one of its developers even denied having used Artificial Intelligence at all in Deep Blue. Today, however, all of the techniques used by this computer are considered part of the numerous tools in AI. To be more specific, Deep Blue ran a variation of minimax, a tree search algorithm. Now, about hardware. Deep Blue had a special architecture designed to implement some form of the alpha-beta search algorithm which you'll see in this course. The technical specs include a mainframe computer, massive parallel computing, and custom processing chips. But a more significant detail is that it could process 200 million games per second. If we compare this rate of computing with modern computers, I'm happy to report that today, you can get this processing power from a smartphone. However, remember that this was a Supercomputer back in 1997, when cellphones weren't smart at all, and today's Supercomputers are much, much more capable than today's smartphones.

Contents