20 years ago, IBM's computer, Deep Blue, defeated at the time the undisputed World Chess Champion, Garry Kasparov (who is considered by many to be the greatest chess player of all time). On May 11, 1997, the overall score between Kasparov and Deep Blue was equal when game six was played. The winner of the game would be the winner of the tournament. Deep Blue won the game through a daring move that wrecked Kasparov's defense and forced him to resign in less than 20 moves.
This was a giant leap in artificial intelligence (AI), the technology that tries to mimic and beat the human intelligence. In spite of this big progress in beating humans in the game of chess, which was considered by many as a symbol of intelligence for a long time, the belief that AI was close to human intelligence in general was rather weak. The game between Deep Blue and the world champion was still fairly even, and as a matter of fact, Kasparov defeated Deep Blue a year before in 1996.
The next challenge for AI was to take on the ancient Chinese game of go. In a 1997 New York Times article, Dr. Piet Hut, an astrophysicist at the Institute for Advanced Study in Princeton, stated that the computer's ability in defeating humans in go wasn't very close.
''It may be a hundred years before a computer beats humans at go -- maybe even longer. If a reasonably intelligent person learned to play go, in a few months he could beat all existing computer programs. You don't have to be a Kasparov.''
Almost 20 years later, in March 2016, Googles AI system, AlphaGo, defeated the World Champion Lee Sedol four times in a five-game tournament. Unlike the almost even tournament between Kasparov and Deep Blue, the victory of AlphaGo was very clear with three wins in arrow proving the superiority of AI. With the recent advancement of AI in the last 5-10 years, of course the estimate of the time horizon for beating humans in go was shorter than a hundred years as opposed to the prediction 20 years ago. However, it wasn't believed to happen before the year 2025.
Games like chess and go are symbols for human intelligence and intellect, which is the main reason for the obsession of creating computers that outsmart humans in these games. Perhaps there is no direct benefit of such computers, but the learning gained in understanding intelligence, the human brain, and the capacity of creating AI is immense.
Now what is the next challenge for the coming 20 years? There is no doubt that it has been established that AI will revolutionise our way of living, and this is predicted to happen at a pace that is faster than our brain can grasp, so called exponentially fast development. The exponential rate simply means that the development of the next year is much larger than last year and of that in two years from now is even larger than the development of next year. There are many avenues of research and development that are taking place currently. Autonomous driverless cars (and perhaps flying objects) will most likely see the light in our streets as a way of transportation. In 50 years from now, we will probably look back and think what savages we were 50 years ago driving the cars ourselves.
Natural Language Processing (NLP), Natural Language Understanding (NLU), and Natural Language Generation (NLG) are other topics that are making huge progress. About six years ago, IBM Watson impressed on the AI community and the world when it ended victorious against the World Champions in Jeopardy, Ken Jennings and Brad Rutter. Watson learned by itself by reading a huge amount of documents, roughly 200 million pages of structured and unstructured content, including Wikipedia. The challenge here is not the information storing as it's just a matter of infrastructure. But the real challenge is to understand the contexts and how to link the information together. Not only did Watson demonstrate its ability in understanding contexts linguistically to provide the correct answers (or questions as it is the format of Jeopardy), but it also made relevant jokes in the show that surprised everyone and put everyone into laughter, including its competitors.
We have just started applying AI in our everyday life and there are yet more interesting inventions to come that will change our way of living for good. Companies like Google, Facebook, Amazon, and Netflix are few examples that most of us are familiar with where AI is used extensively, and the list can grow longer. These companies, although much further ahead in the AI development, are still in their infancy compared to where the technology is going. Looking closely at what these companies are doing, one realizes that much of their success is based on understanding personas and making recommendations depending on the context. Google search learns about us, the users, by understanding what we search for. Depending on our search history, Google search tries to make recommendations on similar topics or items we might be interested in. Facebook builds its own understanding of its users based on the information stored of the activities on its website such as posted text, pictures, or likes. Of course, Amazon and Netflix are doing similar things. Based on the activities of their consumers, profiles and segmentations of their products and markets are classified to give their customers a better experience.
The battle of getting more information about people and their behavior has just started. Understanding humans and their behavior and forming different classifications and segmentations of their needs and interests will grow ever bigger. With relatively simple use of the big data that is available today, one can make a huge difference. The challenge for the near future will be to build even more efficient recommender systems. As of today, there is no generic recommender system that provably works very well. In addition, much more work needs to be invested in understanding different personas from data where text and other activities of the personal activities should be aggregated and included in the optimizations. This will be the topic for our next post.