Artificial Intelligence, Machine Learning Neural Networks: Beyond Buzzwords

All of us heard about AI and Machine Learning. Those are real buzzwords nowadays. And to add to the confusion, people tend to use these terms interchangeably. What do they mean? What are the most common examples? And how mature is the state of these technologies in our everyday lives today? Let’s dive a little deeper.

Artificial Intelligence

Many of us have some idea what AI looks like. We’ve seen smart robots in fantastic movies, read in books and comics more about them. We can say that AI is designed to act as a human; in theory it’s able to handle any task and make decisions as any living person. A synthetic brain created by humans. It can think, analyze and decide. It can repeat the whole thinking process like a real person. From this description it seems that AI is some sort of intelligent robot we can meet only in cinema or in the very distant reality. And that’s what we call Artificial Intelligence. Although it’s not developed to that point yet, we still encounter smart robots in everyday life today – just in different forms. Below, we’ll talk more about them and the latest advancements in this field. Let’s first move to the subset of AI referred to as Machine Learning.

Machine Learning

The idea of ML is to give data to a program and let it learn for itself. Thus you do not need to teach a program everything you know, only to teach it how to learn. The program then gets the information, analyses it and makes conclusions. The definition of ‘Machine Learning’ was first used in 1959 by Arthur Samuel, American scientist and Artificial Intelligence pioneer. He created a self-learning checkers program and it was the first example of a smart machine. His program assessed the number of pieces and kings on each side, the number of pieces that could become kings, and measured the chances of winning from any given position. The second push of ML development was with the advent of the internet. It gave access to huge stores of the information and made it possible for this data to be analysed.

To teach the machine, you need three things:

  • data – as much as possible, hundreds of thousands of examples (more = better)
  • features – the program needs to know what to look for
  • algorithms – how the program will analyse data to make informed decisions

Simplest example: your email service knows features of spam letters, and when you receive a suspected spam message, it goes automatically into the relevant spam folder.

Neural Networks

Part of AI for which we could not fail to mention is neural networks. These two words became buzzwords as well. Neural networks are mathematical sets of algorithms, whereby its software or hardware implementations aim to mimic the way the human brain operates. It is built on the principles of functioning for human nerve cells. It is a deeper concept of Machine Learning designed to classify and analyse information the way humans do.

For people, it takes a moment to process information and make decisions. For example, you are driving and see that road maintenance caused a traffic jam. You understand that on the way back, you’ll choose another route to get home faster. This analysis and decision making process took less than a second in your head. The thing is to teach that to a machine.

Not today, but perhaps in 10-20 years, we will reproduce neural connections of our organism and then we’ll say that true “Artificial Intelligence” is here.

Artificial Intelligence in Everyday Life

No, no, no. There is no real Artificial Intelligence in our everyday lives yet; at least not for the majority of us. Sure, the smartest machines today can remember, forecast some things, choose the best option and reproduce it. They can transform your voice into text, transform text description into pictures, write articles, poems, prose and answer the questions based on given data. People even taught robots how to dance! However, these programs are generally constrained, and cannot grow wiser and grow beyond the frames we have set. All these examples are rather AI prototypes, based on Machine Learning.

Some AI functions & frameworks have penetrated into our everyday life. Let’s review a sample-size of them.

  • FaceID or FingerprintID. Yep, your phone is called “smart” for a reason. It remembers your face or fingertips then when somebody tries to unblock it, and your smartphone analyzes whether it’s you or not. We can add here other functions, i.e. voice assistants; they all process the natural language and AI generates an answer for you.
  • Streaming services. They gather information about videos, music, movies, genres, performers, bloggers you watch/listen to and suggest something similar or new in terms of tracks/clips/videos/movies you might like also.
  • Chatbots. These guys analyse phrases, words, and answer options you choose to provide you ultimately with the most appropriate response. Sometimes chatbots speak in such a realistic way you believe you’re speaking with a human. They are widely used on messengers by different social media and commerce service providers.
  • Advertisements. When you are looking for something on the internet, a particular program will present you with relevant ads to help you find the right product or service; sometimes it works better than other times!
  • Online stores. Similar to ads, but the stores’ algorithm remembers your preferences and can suggest to you not only one particular product you were looking for, but some useful supplements, accessories, or add-ons.
  • Maps. We all use them to check traffic jams and to find the best possible route home.

Concluding Thoughts

To conclude, we can say that Artificial Intelligence, as an intellect artificially created by humans, does not truly exist at the moment. What we have now is Machine Learning, built on the basics of programming: “if – then” type logic. We give a dataset based on real events to a program and it can make predictions on this basis. Nowadays, even the cleverest  machines cannot go beyond the terms of tasks; and this is precisely the edge where ML ends and AI begins.

Talking about the current state of Artificial Intelligence, we can draw another human analogy. First we are born, then we learn some elementary things: how to walk, eat, talk. Then we go to school to learn more, later attending some form of university to learn some particular things like medicine, economics, etc. Artificial Intelligence now is at the stage of a newborn child; it can do nothing without our help. Our brain processes a lot of information momentarily, you could say it has an interface of getting data from everywhere: we can hear and know where the sound came from, we can see, we can touch and we can feel. To reproduce neural networks, we need to understand how machines will receive these loads of information. Today, robots and programs cannot process disordered data; they need it structured, and we structure it for them. But this is just the beginning. Today, we teach programs like children. Tomorrow, Artificial Intelligence will be capable of teaching itself.