Why Artificial Intelligence is booming now?

Why Artificial Intelligence is booming now?

The term artificial intelligence was coined in 1956 by computer scientist and professor of mathematics John McCarthy at Dartmouth College. Many scientists have been contributing to the field till date, but why did AI boom occur now, why not in mid of 20th century or few decades back? What possible impact can AI have on our society? What types of jobs will be eliminated/created, check From First Industrial Revolution to 4th Industrial Revolution.

Basically, the convergence of three driving forces have sparked AI boom now.

  • Computing Power

To build high-functioning systems, you must have the right hardware and infrastructure in place. Early personal computers didn’t have sufficient power. Supercomputers were highly expensive, out of the range of even small and medium organisations. Now Nvidia GPUs cluster (worth few hundred thousands dollar) can match the capabilities that of Supercomputer. Plus, there are GPU cloud services easily accessible to every individual. Implementing the machine learning algorithms on a GPU could speed up the training process by 10x – 100x.

Google has rolled out TPU (Tensor Processing Units) which it claims to be 15 times faster than a GPU (graphics processing units) and designed specifically for machine learning. IBM is working to develop a quantum computing system to power their supercomputer Watson.

  • Data availability

We are generating more data today than ever before. Social platforms like Google, Facebook and all major organisations have all the basic personal data about you. Uber knows all your locations. A single modern car has 100 or more sensors that monitor functions such as fuel level and tire pressure. To put the rapid growth of data into focus, 4 years ago, IBM reported 90% of the data in the world today had been created in the last two years alone.

  • Better algorithms development due to availability of data

Previously, there simply wasn’t enough available data to train a machine, let alone build algorithms that allowed machines to train themselves.  The more data we have, the better the algorithms do. Artificial Neural Network algorithms have been developed in 80’s. But the data and computational power was not available. This explosion of data+computational power has made it possible to refine algorithms and develop more extensive datasets algorithms can consume for machine learning. For example, speech recognition is a prime example of this: progress in speech recognition is closely related to the size of the data sets available for training – it requires several hundred thousand hours of speech to train. Over the past decade or two, large data sets have become available. It’s become possible to train algorithms using every image on Flickr, or video on Youtube, etc. That wasn’t possible two decades ago. Deep learning has been quite successful for last 7 years. New algorithms have been invented which gave speed and accuracy to DNN – autoencoders, stochastic gradient descent with mini-batches, ReLU activation functions, drop-out for regularization, convolutational neural network structures and more.

According to Andrew Ng, “Artificial Intelligence is new electricity”. The way electricity brought change in every sector, Artificial Intelligence (AI) will bring in the same way. Data is becoming new oil. Machine learning is becoming the new combustion engine.

How much progress Artificial Intelligence has made so far, and what possible impacts it’s going to make on our society and jobs in future, check From 1st Industrial Revolution to 4th Industrial Revolution.