Google’s artificial intelligence bot made history early this year when it defeated the existing Go World Champion Lee Sedol. The bot was developed by Google DeepMind, and its victory had a glimpse of the future, which is not so distant.
Imagine the day, when computers need not be programmed, but they will learn by themselves how they must do a given task. And not just this, the computers will grow beyond machine language and understand the natural language we speak, like Jarvis in Iron Man, a primitive version of which is Apple’s Siri and Microsoft’s Cortana. And because they can learn new things by themselves, they will start recognizing human emotions and will develop their own conscience.
Current Scenario of Artificial Intelligence
Currently, most of the research is going on in the field of ‘machine learning’ and ‘deep neural networks’, both are precursors to artificial intelligence. Machine learning is all about making classifications and predictions. For example, a numerical series – 1, 2, 4, 8, …, calculating the next term in the series is a cakewalk for most of us. Let’s make it – 1, 3, 7, 15, …, tougher but still easy, isn’t it? In both the series, the basic approach is to observe the series, find a pattern and calculate the next value, and this is exactly how computers are used to make predictions.
The same problem can be extended to predicting stock market prices or weather conditions, where each term in the series represents the status of stock market or weather at some instant in time. And each term will be much more than just a single number; it will be a vector or a matrix. Machine learning can also help us find structure in unstructured data or useful information in large data sets. Even past criminal data of a city can be used to gain valuable insights about why and what kinds of crimes take place in an area and what are the factors behind it. Gaining so much information is possible only because computers can churn in large amounts of data in seconds, the same amount of data will take years to process using human brains.
Artificial neural networks (ANN) is another promising area, which is paving way for further research and progress towards artificial intelligence. Human brains are made of millions of neurons, their interconnections and arrangements being unique to every individual, and this uniqueness sets every human being’s thinking apart. As we learn new things, the interconnections and arrangements change in order to store the new information and at the same time try to retain the older information. These rearrangements are faster when we are infants and gradually become slow as we grow. This configuration of neurons in a human brain is known as a neural network, and artificial neural networks aim at mimicking the same learning process of the human brain. The research and progress in this area is limited by existing computing power available to the researcher, and hence the research in the field stagnated in the late 90’s. But late in the previous decade, because of an explosion in computing power and new developments in areas of computer architecture, artificial neural networks resurfaced. Now even the graphic processors (that go by the brand names of NVIDIA and ATI) installed on typical computers can be harnessed for training such networks. They give massive performance improvements over the regular processors made by Intel and AMD. When things became more complicated, ANN gave way to ‘deep neural networks’, the technology behind Google DeepMind’s Go bot.
The recent advances made in a deep neural networks is used by Google in Google Now, Apple in Siri and Microsoft in Cortana. Deep neural networks reduced speech recognition error rates immensely and also helped in natural language understanding. So, when we say to Google Now, “Wake me up tomorrow at 9”, it is deep neural network in the background, which recognizes speech in English and then converts it into machine understandable language.
In 1996, IBM’s computer Deep Blue defeated the then chess world champion, Garry Kasparov. In 2011, IBM’s Watson defeated humans in Jeopardy. Watson can interact in natural language and can process massive amounts of structured and unstructured data like it churned in the full text of Wikipedia for Jeopardy. Watson is now used by many in healthcare for better diagnosis and treatment of diseases. The game ‘Go’ was a special challenge for engineers as it has millions of possible moves, but Google’s bot triumphed humans in that too.
Some other related fields like Data Analytics, Computer Vision and Natural Language Processing are also contributing to the advances in artificial intelligence, and there will be a lot of demand for engineers skilled in these technologies.
Many see the coming of artificial intelligence with optimism and as many see it with pessimism. Will computers take over the world as is shown in the movie Matrix? We can keep guessing. But there is something computers would never do – the random.
Artificial intelligence is an amazing technology! Read more about amazing technologies invented by science fiction authors here.