This thing AI (part b: machine learning)

In the previous blog we discussed about what artificial intelligence is and I am assuming that you understood. I also gave some examples of already existing machines that we can consider intelligent.

Heads up!

Some heads up first, this blog is the second in a series of five blog posts that are meant to just introduce you to AI. From The sixth blog onwards, you will actually be learning Artificial Intelligence in deeper details and this will require commitments. Don’t be scared anyway, you can be a freelancer.

Diving deeper

Now, today I want us to have a broader view of AI. AI is wide, very wide but very interesting. To maintain simplicity we’ll just focus on the main areas of study.
The major areas of study that drives the AI field are Machine learning and deep learning. However, you should know that deep learning is a special type of machine learning. After we’ve studied both, we’ll see how.

So... What is machine learning?

In AI people usually talk about machine learning a lot .But which is bigger? Machine learning or AI? You see machine learning consists of those algorithms, the methods or the ways of enabling machines to learn .Machine learning as the phrase suggests focuses on enabling machines to learn, yeah you got me right, TO LEARN. How do they do that? Let's use the human brain as the reference here. Basically machine learning is a section of AI.

Then, how do machines learn?

When you are born, we can say you come knowing nothing of this world, but with time, you learn how to speak by listening by listening to people around and imitating them. You then start learning the language, in no time you can speak and communicate with ease. How this happens is that you ears working with your brain recognize patterns in language and its usage, where it is used, when to say what and how to say it. Similarly, machines can learn but now instead of brain cells to process the information, we use mathematical functions. We then feed the machine with the data relating to the task we want it to learn. The machines then goes through a training session. It is told what to do with what data and how to interpret it. The machine is also shown patterns in the data and ”told” what to do when any of the patterns is identified (supervised learning). The machine is then presented with new data but from the same task, it is then left to make decisions based on the experience gained during training. It is the work of the machine now to identify new and learnt patterns in the data. At this point we say we are testing the model. After training, we can decide that a machine knows how to perform the task based on the accuracy. If the desired accuracy is achieved the model is ready for deployment.  The more the training, the higher the accuracy.

A few examples of learning machines you already use/know.

Your YouTube app recommends videos for you and you can agree with me that a bigger percentage of the videos recommended are of your preferences. The app learns you, what you search, watch and kind of channels you are subscribed to. It then finds videos of that kind (identifying patterns) and then recommends them to you. Simple, right? ….but believe me the math behind all this is beautiful.
Look at the kinds of adverts you receive while you are browsing. Many will agree that you are majorly shown adverts that relate to what you search for or kind of sites you visit.

Other applications

Machine learning algorithms are used in prediction extensively, for example in predicting stock exchange rates, probability of success or failure of a company based on its performance, possibilities of economic crisis in a country, how probable is it that a country will be overpopulated in years to come based on birth rates, possibilities of food shortage, etc. We cannot exhaust them.
Machine learning is applied is so many ways in today’s world, especially the online world.  Keep in touch to learn more.Next we'll focus on tge three versions of machine learning.  Namely;

• Supervised
• Unsupervised
• And reinforcement learning