The Evolution Of Machine Learning

The Evolution Of Machine Learning

About 50 years back, machine learning was just a concept in science fiction movies or books. Movies like Terminator 2 and The Matrix which were released in the 90s gave us a very good idea about machine learning. Even then ‘machine learning’ was still a subject of research without much real-life application. But now, we see a lot of examples of machine learning in our daily lives. Virtual assistant, facial recognition, and online customer support are some examples of machine learning. As machine learning has a great impact on our lives, it is worth knowing how machine learning evolved over the years.

Machine learning uses algorithms and models to help computers improve their performance. The machine learning model was created based on the brain cell interaction model developed by Donald Hebb in 1949. His theories about neuron communications were presented in the book ‘The Organization of Behavior’.

In the 1950s, Arthur Samuel, who worked in IBM, developed a program for playing checkers. He developed a scoring function that determined the chances of winning of each side by knowing the position of the pieces on the board. He used the ‘minimax algorithm’ to decide on the next move. In 1952, he was the one who termed this type of program as ‘machine learning’.

After Donald Hebb and Arthur Samuel laid the foundation of machine learning, it was picked up by Frank Rosenblatt in 1957. He created the software ‘perceptron’ which was installed in a machine built for visual pattern recognition. This computer called the Mark 1 perceptron’ is known as the first neuro-computer. Though ‘perceptron’ seemed very promising, it failed to recognize faces and other things. The research was continuously going on to solve the problems, but it wasn’t until the 1990s that the researchers found some fruitful results. 

The concept of ‘neighbor algorithm’ was found in 1967. This algorithm was developed for mapping routes. The salesperson delivering products to customers could use this algorithm to find the best possible route to the destination. During this time, the discovery of multilayers created new opportunities for the field of neural networks. It was found that using multiple layers in perceptron increased the processing power more compared to using a single layer. This ‘multilayer’ concept led to the findings of other neural networks like feedforward neural network, backpropagation, and artificial neural network.

In the 1970s and 80s, machine learning was used for understanding artificial intelligence (AI). But AI used logic and knowledge-based approaches rather than the algorithm. So, machine learning and AI took separate ways. The concept of ‘boosting algorithm’ was developed in the 90s which showed how weak learners could be transformed into a strong learner. This was an important breakthrough in the research of machine learning. After many years of research, speech and facial recognition are now possible through the ‘deep learning technique’.

At present, ‘machine learning’ is defined as the science of making computers perform without any specific instruction. Modern examples of machine learning are self-driving vehicles, chatbots, and the Internet of Things (IoT). The machine learning models are now adaptive and keep on improving with the changing scenarios. Machine learning combined with new technology can solve many organizational and daily life problems.

Leave a Reply

Your email address will not be published. Required fields are marked *