Can learning machines save us?

The long answer is YES - provided we remain smarter. Before i tell you how, let me give you some insight into learning machines. 

The ultimate machine learning dream - to make a human like machine.

Machines either know or don’t. You could call it binary thinking. This is what used to be though. Like humans, machines now have a new dimension to their mechanism. It’s the dimension called ‘samay’ (means 'time' in Sanskrit). This is helping them learn specially to identify patterns that humans can’t see.

By definition, a machine that gets better at performing a task with experience is called a learning machine. I would think ‘experience’ is at the least equal to ‘time’ spent in performing that task. For discussions sake let’s say Roger Federer became a champion by spending ten thousand hours of playing tennis (I won’t argue with the idea of it being his destiny). This obviously excludes the times he thought about the game which can potentially be his current age in hours. In absolute terms ten thousand hours would translate to 417 days or, 1 year 1 month 3 weeks and 1 day.

I would imagine Roger to have reached this mark in several years including times of injury, puberty, binging on fries and coke while watching Pete Sampras win a Wimbledon final followed by an upset stomach spoiling next day’s training, to name a few.

A learning machine however, could reach this mark in exactly 417 days. Could the machine have done the thinking about the game? - totally depends on the computers RAM. Jokes apart, machines can become very smart at a much faster rate than humans, but, on limited number of tasks.

What makes this possible? - the goal of neural networks is to make a machine algorithm think like a human brain. i.e., process layers of input information to produce a probability towards a decision followed by choosing the option with a higher probability. The choice is then compared to a training dataset that makes corrections to achieve desired accuracy levels. It's a lot like neurons in our brain receiving information from our senses and combining that with memory of past experiences to arrive at a decision leading to an action.

The brain is a network of complex algorithms which can over time learn to process any form of information. Just like, if you train the Auditory Cortex of the brain to see, it will learn and adapt to see. All you have to do is temporarily move the connection to the Auditory Cortex from your ears to perhaps an external camera. Much like what’s happening below:

Except here the attempt is to see through your tongue.

It’s an established fact that machines have a significantly more reliable memory and are better in identifying patterns than an average human. This machine skill combined with the power to learn is revolutionizing how many of us live.

For example, Gmail classifies emails between primary, promotional and spam with a learning logistic regression algorithm. Similarly, auto suggestions made by google search, product suggestion on Amazon and song recommendations on Spotify are personalization that most of us have enjoyed.

Who is missing out? - Eritria, Niger, Chad and Madagascar together have a poorer internet penetration than Afghanistan. Note that North Korea is not a part of this comparison as all their usage probably happens through one person’s 57 inch iMac, an outlier that we can't consider.  

On the brighter side, according to forecasts by eMarketer, by 2021 ~ 54% of the world would be connected to the internet (4.13 Billion) as compared to 47% at present (3.47 Billion) and this increase is expected to be driven by mobile device penetration in low-income countries across Africa and Asia Pacific. This should be supported by improved infrastructure for high speed internet allowing penetration of personalization through machine learning.

Is machine learning only for personalized experiences? - The answer is no.

There are several innovations coming up in the field of development that are solving problems for large groups, sub regions or even entire countries. Saving lives by providing timely flood risk warnings in Togo, creative usage of drones, the cloud and television channel frequencies for precision farming techniques are some such examples. These advances are definitely preparing us to meet challenges from climate change to food security.

If so far, its sounded as if I have given all the credit to machines, let me assure you that in this case the chicken definitely came first. After all, it is the effort of a global network of scientists, engineers, social entrepreneurs and businesses that have united to make it all possible for profit and non-profit. Critics remain skeptical and there are innumerable doomsday stories based on AI, some my favorite, but I truly hope the chicken came first story remains popular forever. I also hope that we succeed in healthy co-existence by leveraging machine learning to predict and prepare for ever-growing demands and risks to human existence.