Machine Learning is about teaching the computer how to learn. There is a great difference between how we learn and how machines learn. Think back to a time when you were a toddler. You learnt the english alphabets from A to Z. How much time did it take you to read and write them. Maybe an year or two. Later you learnt the nuances of the language – the building blocks – words and the grammar to make sentences. It took some time, but you learnt it. What happened was that your brain took in this information and converted it into patterns. You see every day we are learning. Our neurons are plastic and hence malleable to new information. Either they get recorded in new neurons or existing ones change themselves to accommodate the influx of information. The thoughts that we linger upon, create deep neural pathways which are found in our subconscious minds.
Machine learning (deep learning) happens in a similar way. There are three ways a machine learns :
The first one (supervised) is a way of teaching the machines with human intervention. Take for example, a classification problem. Say there are oranges and apples in a bag and machine needs to identify which is which. For this problem, we may further divide it by looking at the weight and texture of these fruits. So we can evolve a rule which says that apples weigh less than 150 grams and oranges are above 170 grams. Now this has to be taught to the machine. There may be several such rules that are handed over to the machine.
In unsupervised learning, we throw a pattern at the machine and it learns from the algorithm that we feed in. For example, there is K-Nearest Neighbour algorithm in Clustering. To cut it short, the machine learns to predict the answer to a difficult problem, by looking at the nearest neighbour in a cluster of same data. The more distinct the cluster, the better the prediction of the algorithm. However, things in the real world can get messy.
To accommodate the variations that we come across in our day to day life, we need something that learns just like we do. A neural network. Something that works like our brain. The brain that we have consists of billions of neurons arranged in a layered manner. By that I mean an image like a dogs’ is stored in many layers. For example, the first layer may be the color, the second may be shape and so on. When we look at a dog, all these neurons fire and we understand that it is a dog. Further, if we have misjudged, the error gets back into the input, so that next time our judgement is better. This is how reinforcement works – for a human or a machine.
This concept of layers of perception is what drives deep learning. The more the number of layers, the more granular the understanding is and the more the data that we expose the machine to, the better is the prediction. This kind of learning is similar to how we learn, the only difference being that the computer is nowhere close to our parallel computing capabilities. However, they are picking up. With advances like using GPU’s and specialized AI chips like Tensor Processing Units (TPU), the rate of learning is going up.
So where do we go from here ? We all know that Machine Learning is behind driverless cars, the Amazon Echo and the new breed of games. Whatever our brain can do, will be one day be exceeded by machines. But do they really think ? It’s just an algorithm – maybe a self modifying one. Whereas we are not just our brain. That is just the hardware – the mind is the software where all the magic happens. Our emotions for instance (Call it from the heart or the mind) are an integral part of who we are. Our souls are much awakened than the machines’. We are in short, a much more complex pattern than a machine. What we have to remember is that, a machine will never be able to replicate a human being fully. Our authentic self is a limited edition. Nobody in this universe carries our distinct signature. The machines will have their own place and we will have our space. Co-existence will be the key to the future.
Some folklore say that Kalki (the next avatar of God) is half man, half machine. You see, these are the things that we find fascinating, which a machine can never. We are made up of stories and machines simply are ‘0’s and ‘1’s strung together. The Earth loves the touch of your feet and the wind likes to play with your hair. Will a machine ever be able to feel the depths of these lines – you tell me.