The deep blue story changed the way AI is being developed for a few reasons. It taught researchers that there are many ways to approach a complex problem. The human approach is intuitive and pattern-recognition-based, while the machine approach is search-intensive and involves looking at billions of possible outcomes. Often, the two approaches are complementary, and Deep Blue gave researchers a clear example of how they could work together.
The IBM team has been advancing the field of artificial intelligence (AI) for 20 years. The team developed Deep Blue, a computer program that can analyze 200 million chess positions per second. It then selects the move with the highest likelihood of success. This advanced AI system relied on machine learning approaches to learn about chess. The game is well-defined, with 32 pieces and 64 squares, and Deep Blue had a clear goal.
When compared to today’s systems, Deep Blue used less machine learning than its competitors. In fact, it used a chess accelerator chip called a board evaluation function. It learned its parameters from analyzing thousands of master games. Before Deep Blue became the champion, IBM computer scientists had been interested in chess computing. They had developed a chess playing machine, ChipTest, during the 1950s.
While Kasparov and his supporters believed that Deep Blue was controlled by a human grandmaster, many outsiders argued that the program was artificially intelligent and not a real person. The program’s rigid adherence to logic, while ignoring human emotions, made it unhumanlike. But that is not to say that Deep Blue did not use any artificial intelligence. Instead, it was simply a computer program implementing simple rules on a grand scale.
When it comes to chess, the results are still a long way from the best human players. Deep Blue’s computer is more narrow than the human brain, so it’s unlikely to win a tournament against an opponent who is more diverse. This is where the question of whether Deep Blue used Machine Learning comes in. Deep Blue was successful in many aspects, but still fell short in others. And even if Deep Blue was the perfect chess player, it still has some major flaws.
While it may not be clear how much Deep Blue relied on Machine Learning in its development, the computer did learn from the grandmasters. During its development, the team consulted with grandmasters and played against the program to determine its weaknesses. The program was able to beat the world champion Garry Kasparov, and despite its flaws, Deep Blue’s win proves that humans are not the only ones who can build great computers.
AlphaZero, a rival of Deep Blue, has proven to be more successful in the game of Go. Both AlphaGo and Deep Blue have evolved from earlier algorithms that were designed for Chess. AlphaZero, a more complex computer, has been developed more than 25 years later. Despite this, AlphaZero has already been able to beat Deep Blue in the same game and has also shown that artificial intelligence is more advanced than human intelligence.
In 1997, IBM’s Deep Blue computer beat Garry Kasparov in a six-game chess match. This victory sparked a debate about whether the machine used Machine Learning to beat the grand master. Deep Blue’s rise to the top in the game mirrored the Rocky Balboa story. In Philadelphia, it beat Kasparov and won the match a second time. The next year, it defeated him again.
A similar process was used to develop the new version of Deep Blue. Deep Blue researchers created a new chess chip that enhanced the machine’s ability to evaluate pawn positions. This version of Deep Blue could evaluate up to 200 million possible moves per second. It could explore 40 moves into the future. Its name – Deep Blue – was an obvious reference to the technology behind the supercomputer. Deep Blue utilised a new method of training the machine.