This post is not referring to the rise of Skynet or the spectacularly dull witted AI from Matrix. The real, and more plausible question, is: will machines be able to truly replace humans? Will we be able to program Artificial Intelligence (AI) with the intellectual level of a human being? The answer is yes!

Humans have developed intelligence that is so advanced, it remains second to none in the known universe. We have achieved this through a slow and rigorous process known as evolution. Human evolution can be traced over a period of almost 7 million years. As compared to this our computers have evolved barely over 7 decades, a remarkable achievement. The rapid growth of computing power has been exponential.

Today, we walk around with supercomputers in our pockets or strapped to our wrists. We drive ordinary cars packed with more computing power than NASA used in the Apollo 11 mission. We have the internet, a massive neural network linking billions of devices together, allowing them to learn from each other. At the touch of a screen or flick of a wrist, we get more things done than we could have dreamt of a couple of decades ago. With this in perspective, does AI still come across as farfetched?

A popular misconception still remains that AI of any form is a thing of the future. As of now, a major chunk of the world runs on AI. Automated systems have been employed across the globe for several tasks. These programs learn from past behaviour and act upon stimuli received from sensors. AI charts out train schedules and manages most utilities in developed countries. AI identifies when you will reach home and ensures your laundry is done on time. Automated home systems receive real time data from you and make sure your lights are turned on, the thermostat is set at a comfortable temperature and the garage door is open.

These are, of course, a branch of AI known as Artificial Narrow Intelligence – or ­­­­­­­­ANI. Their intelligence is restricted to highly specialized tasks only. Your computer has a program loaded that can beat you at chess hands down. But the same program, when loaded onto a washing machine, would just stare at you blankly – similar to when Charlie Harper was forced to do laundry in Two and a Half Men.

Google’s driverless cars have a revolutionary form of ANI. Using data from a vast number of sensors, the car is able to showcase phenomenal driving skills – albeit at a slow pace. But that incredible piece of software can do little else apart from driving cars.

What our highly renowned scientists, in their crisp lab coats, aspire to now is Artificial General Intelligence (AGI) which can be considered near-human levels of thinking.As you have probably figured, AGI is unlike ANI in terms of what it can do. This form of AI will be far more complex and not restricted to one particular task. Akin to humans, AGI will be able to learn and adapt to any environment or situation it encounters. This is much closer to Skynet than your washing machine, I’ll admit. But like it or not, we’re not that far away from developing such a sophisticated form of AI. Here's an estimate by renowned futurist, Ray Kurzweil:

IBM has made several strides in this field with Watson, an AI that beat the world’s two best Jeopardy players at their own game. So what makes Watson different? This AI is the first ever to combine natural language processing, hypothesis generation and evaluation, as well as dynamic learning. To state it simply, Watson can speak like a human, make its own judgements and learn on the fly, from its own mistakes as well as yours. Put all these together and you get a Jeopardy player like none other.

A more recent and jaw-dropping example is that of AlphaGo. This is an AI developed by Google’s DeepMind. To be fair, AlphaGo is focussed purely on playing Go – the world’s most difficult board game. What is truly astounding here is, this game cannot be played by calculating moves. This is because there are more possible combinations in this game than there are atoms in the universe. This means AI cannot calculate all moves using brute force like it would do in a typical game of chess.

What makes AlphaGo awesome is that it relies on something an AI has never used before – intuition. Humans rely on intuition or gut-feeling for a lot of complex tasks. No unmanned drone can fly as well as a trained pilot because the pilot can do complex manoeuvres relying on intuition.

AlphaGo has astounded the world by beating 18-time Go world champion, Lee Sedol 3-0 in a 5 game series. As I write this, the fourth game is being live-streamed on YouTube with over 50,000 people watching.

The constraints we face for developing AI are slowly melting away, such as storage and computing power. Adaptation of Moore’s law to storage space hypothesizes that cost of storage space halves every two years. This might not be an accurate statistic, but storage space is rapidly increasing while its cost continues to drop. Google’s advances in quantum computing will blow open the space for the next generation of hypercomputers.

Today, several tasks which humans used to do, are completely automated. Humans are superbly intelligent but have mediocre hand-eye coordination at best, with poor accuracy and repeatability. Accuracy and repeatability are essential in production lines and therefore, automation has seen maximum implementation in this industry. With exponential levels of innovation, technology moves closer and closer to more advanced forms of AI. With this, we might be able to eliminate human error altogether someday.

UPDATE:

Lee Sedol succeeded in defeating AlphaGo on his fourth attempt, but AlphaGo came back with a vengeance and beat him in the fifth and final game of the tournament. The final score of the legendary Man versus Machine tournament stood at 4-1 in favour of AlphaGo.