Machine learning is ‘predictive analysis’ in very simple terms, agree? Arrive into a conclusion by analyzing the data. Can this possible only with computers? We, human beings do it on a daily to basis, to catch a bus, to drive a car, to shop and what not.
Don’t be confused. Let me try to explain and don’t blame me at the end if that doesn’t work, a pre-bail has been taken :-).
Let me take the example of how we learn driving. Excuse me for those who never tried it 🙂.
The driving instructor gives you first set of data like usage of steering wheel, gear shifting, clutch, brake, accelerator etc. Initially, with those basic data you mess with clutch, gear, accelerator and brake often. Then you slowly correct the mistakes by practicing which means you are learning by feeding your brain with additional data on how to use it effectively. Eventually you get it correct when you have more and more data; and create your own algorithm to drive your car. Result of this you start applying the break softer, start shifting the gear smoothly. That is exactly the machine learning does with the help of your intelligence (it is currently being replaced with Artificial Intelligence –AI 🙂).
You struggle to get your driving correct with basic instruction because you had only limited data. You probably learned from the mistakes but in Machine Learning mostly you are fed with lot of data already available.
Once you start driving on the road, you do lot of Machine Learning there as well. You then start predicting what a pedestrian is going to do, which vehicle could probably cross the road, what are the chances of animal crossing the roads etc. You make your driving decision with the predictions based on your previous experience and observations. Good predictions help you to reduce the chances of accidents. So, what is the basis of all these, it is DATA. If you are driving in India, you are dealing with lot of algorithm and data 🙂.
You get this data by watching around, observing, hearing, touching etc and you are feeding this into your brain and you process the data and get the output make a decision. We often call this as Experience. Can we call the experience as Machine Learning? May be not. Let’s take another example go bit closer.
You are driving in the rural areas and you often see animals on roadsides to cross road. Those can be dogs, cow, goats, cats, buffaloes etc. Which animal does make you nervous and slows you down? It is Dog for me. How did I make that call? Because I know the characters of each animal and we have seen lot of cases where dog jumps in front of the vehicle. Each of us have lot of data of each animal on our brain. That data helps us to arrive to conclusion. I guess it is what computer does and we call it as Machine Learning.
Machine Learning output makes more sense when we feed it with lot of data. So, as it says ‘ Machine Learning’, it is Learning. And obviously, the algorithms that you create to process the data.
As per Intel, an autonomous car generates and consume 40 TB of data in 8-hour long drive. That is a lot of data? That is almost equal to 5000 full HD movies (8GB each). Lot of data being captured when we use internet, workout (by using gadgets), internet searches. Every individual will use/generate 3 GB of data everyday by 2020. We need lot of space to store all these and hell lot of processing power. If you ask any technology company about their future, they say it is on ‘ DATA’. That is where you have to look at the google who developed a new processor called ‘ Tensor Processing Unit’ (TPU). It is for Machine Learning. That is the kind of intense processing power required for Machine Learning in the future and specially designed ones.
With the evolution of the cloud computing, you can get the Machine Learning as a Service from AWS, Google, Azure etc. Using these services, it is much easier to make things work and use the API calls integrate it with your existing application.
I hope it helps, I am happy to take your feedback on improving the above example.
But, I know nothing about the Machine Learning.
One Reply to “Machine Learning: I do it, you do it everyday without a computer”
Thanks anoop nicely explained about MI. I know it’s little confusing topic and difficult to make other to understand. But you have done nicely with good example…