Actually the answer to this question you can find here (Machine Learning) and here (AI). Wikipedia is amazing and is the best tool to get some clue about the subject. But why am I writing this text? Just to give people an understanding that this is not some fiction movie – it’s reality. And the truth is – the reality is ugly. AI is a math – equations and formulas, logarithms and integrals. And a lot of data. Today AI is one-problem specific: image recognition or pattern finding or prediction based on regression or language processing or recommending systems. And all of this separately, not all at once. It’s far far away from Skynet :D.
Machine learning is a field of computer science that gives computer systems the ability to “learn”. What does that “learn” mean? It means that machine just eats a shit load of data and perform some specific task on this data: it’s being looking for patterns, correlations, features or anomalies. The simplest example is linear regression. Yes! Exactly! That weird thing you studied at statistics class (who did that). Machine learning is about statistics and math (again!).
So what is linear regression? Imagine you’ve been working at the same company, doing the same job for 10 years. The company was smart enough to rise your salary every year as you become more experienced. So if you take those numbers and put them on a plot you will get something like the graph below:
 |
| Years of experience – Salary graph |
And what we see? A line! Which we can put in between those dots, because our brain is incredibly amazing at pattern recognition. And what’s more interesting – this line is infinite. So taking that line in mind we can predict your salary for the next year by using simple mathematical formula. (and again I have sent you to Wikipedia)
And that’s it. This is how machine learns – the engineer gives it data and experiment with it. If I believe, that in data I have, the linear dependency can be found I try linear regression. If it doesn’t work I try another algorithm. And that’s it. Depends on task you want to accomplish you choose a model you want to try.
Those brilliant minds from Google who developed AlphaGo, that could beat world’s champion in a very complex game few times, have put an enormous amount of work and time on it and AI accomplished the goal, but that’s the only thing this machine is capable of. And it requires a lot of knowledge and natural intelligence to build such a thing, but it all starts here, with super simple linear regression, with basic statistics and math. And the most awesome part – you don’t necessarily need to know all those formulas – they all are already written in methods and functions, you just have to be aware of their existence.