March 22, 2017

Artificial Intelligence (AI) is surely among the top buzzwords of our time. Some might even say that we're living in an AI bubble of sorts. Indeed, it is impossible for us to stay untouched by AI of some kind.

Machine Learning, Artificial Intelligence AI has an in-built power to learn by itself. Image Source
AI has had a long and checkered history. It's been around for a few decades in science and for a considerably longer period in science fiction. Over the last several years, AI has made great strides in solving real world problems and it seems to be accelerating at the current time. When we speak of AI, we tend to cut a wide swathe both in the technology and in the application domains. Deep Learning, Natural Language Processing, Human Machine Interfaces, Machine Learning and even Expert Systems, and many others get clubbed under the AI umbrella.

Just as computers are very good at making large number of calculations (or other repetitive tasks) quickly, it turns out that machines are excellent at solving specific problems that are the building blocks of "intelligence". This is enabled by sophisticated algorithms that crunch a large amount of data in order to achieve a desired level of accuracy. If the problem is of recognising, say, a cat, you start out by giving the machine a large number of pictures that are classified into cat and no-cat categories. You then train specialised algorithms that extract some features out of all these pictures which it considers to be salient features of cats. It turns out that the machine can then use these salience computations to figure out if a new picture is that of a cat with very good accuracy. For this particular problem, you might use Deep Learning with Convolutional Neural Networks. But the point is that this algorithm is good only for recognising a cat. It is not equipped to play chess. This is an example of specific AI, and a vast majority of AI, we come across, falls into this category

Artificial Intelligence Certain AI's may even be witty and smart ; 
Image Source 
In recent times, we've made many breakthroughs in specific AI fields. One exciting announcement from Microsoft is that their speech recognition software is now at par with human accuracy. Or, the example of Google's DeepMind being able to beat the winner of the World Amateur Go Championship. Whenever we see an uncannily good recommendation from Amazon, or an accurate tagging of a friend by Facebook, it is specific AI at work. These achievements definitely deserve our unstinted admiration.

Human beings are really good at multitasking. Artificial General Intelligence (AGI) is an emerging

Artificial Intelligence What does the future behold? Image Source
area that strives to match and eventually surpass human intelligence in a broad sense. AGI is the stuff of science fiction. It can reason, solve complex and previously unseen puzzles, and perhaps even have a sense of humor. What other, more critical tasks it can do is up to our imaginations. This is the holy grail of advanced AI research.

While some futurists are optimistic about the future of human beings in an AI-enabled world, others see a risk that a sufficiently advanced AI may harm humans, or at least render us useless. The jury is still out on how that will play out.

Text by Chetan Vinchhi

Chetan Vinchhi is a Technologist, Entrepreneur, Mentor, and Consultant based in Bangalore.