Well, Max, how would you explain what artificial intelligence is to people unfamiliar with it and without programming experience? 

First, I want to say there is nothing in common between programming experience and ML understanding. I think it’s closely connected with your analytical skills and knowledge of working with a large amount of structured data. Because ML is,  first of all, working with some information with data. Well, people who work with data are also called data scientists. Maybe you’ve heard it before. How do I explain it? In general, what AI does is simply convert some structured information. It can be anything, images, summaries, text, voice, etc., from one specific form to another. The main difference between AI and other programs is that AI learns how to transform from one type to another. To understand it only, let’s imagine a model of our human brain. Our brain has a lot of connections, which are called synapses. Well, they hold, they transform, and they can change some different signals in our brains. And, well, in general, it affects our decisions in other cases. And these connections were built in the past, and we made these decisions based on our previous experience. For example, a professional basketball player can easily catch an object flying at him because of his trained reflexes. And the same way AI works, it trains to work with unknown data using other known data. And it builds new connections inside to compare this data. And so the input matches the output. And this new iterative model is usually called AI.

What is the difference between artificial intelligence and machine learning? Some key aspects. 

Well, it’s an excellent question. It’s indeed a brilliant question. AI terminology and itself are overhyped nowadays. Well, scientists, especially data scientists, don’t like to use the term AI, to be honest, because it doesn’t have an exact meaning in maths, so it doesn’t have any strong definition. To describe it in general, it’s the program you created after learning. But scientists prefer to use the term maths model, and we will use maths model in our conversation, not AI. ML  is a set of specified algorithms based on learning to find some solution. You can mainly use ML in terminology without AI, and you won’t lose anything.

What are the basic principles of machine learning? Can you provide an example of a simple machine-learning model?

Well, as I’ve already said, the main difference or main feature of ML is the learning part. As an example of a model, the simplest model to understand is called a decision tree for classification. Decision Tree is an ML algorithm based on conditional statements. In programs, it’s based on many ifs. So, if you try to Google it, you will see some diagrams. It looks like something similar to a block diagram, but it seems like a tree. And it can be easily explained and understood. Because classification is based on many ifs, we made decisions based on some created statements.

What opportunities does AI open up for humanity? Can you provide a few examples of its real-world applications? 

Yes, the main advantage is that, based on its basic structure, ML can gain and copy some human experience because our decisions are made using our experience. ML can achieve it faster and use it better in different cases. For example, a few days ago, I saw a presentation from Google that introduced a human sign recognition model for deaf people. So what does it mean? It got your webcam and tried converting your signs from video to text. Well, how to explain it? It’s just like some translator. Well, if you use a translator, it translates from, you know, English to Ukrainian, for example. But this type of translator is for video, from your signs to text. And indeed, it’s very inspiring, and it’s a highly complex model because you know that the signs may be different, people are different, and they use their hands in different positions. So, first of all, we need to catch an image, and then we need to convert it into an understandable format for the program. So yeah, that’s what a great invention I see nowadays.

Are there any dangers or limitations to using AI and machine learning? It will give more answers to your previous question.

Yeah. Of course, it may be dangerous. The most dangerous part of ML is models. They have unexplained connections inside. So, after learning, you receive a model — we call it a black box. So you don’t know what happens inside. It’s difficult to manually change it because nobody knows what happens inside or why these connections are built this way, even for inventors. So it can be a big problem for powerful models. We call it deep learning, like, for example, chat GPT or maybe copilot. They can give you some unexpected results, which I think inventors did not expect. So, they add some manual limitations to their program. Some of them are not yet ready for powerful models like this. So it should be controlled, of course. But, well, I want to say that ML can only give you advice. I think that humans should make final decisions in some sensitive areas, like medicine, for example.

Сan AI and machine learning completely replace humans in the workplace? If not, are there tasks that AI still cannot perform, and maybe some more words about humans and AI?

Well, yes, as I’ve already said, first of all, it’s a very sensitive area like medicine. Maybe something with dangerous jobs like piloting, for example. What can I say? If your job is routine, non-creative, and has not dynamically changed for years, then, well, yeah, I have excellent news for you. Well, I don’t want to affect artists on this topic because art is a very complex topic to discuss. Everybody has a vision of what is art and what is not. It may be different. Yeah, I’m not so familiar with art, so yeah. Some routine jobs may be replaced by ML, even if it requires some experience. But in some dangerous areas, no. Of course, it shouldn’t be.

And about something new, what developments in AI are you looking forward to in the near future? 

Well, of course, we expect new inventions in deep learning. Deep learning is a powerful model. Deep learning is powerful. But as I already said, it’s overhyped in AI, and I expect a decrease in interest in ML. Because, well, as for me, many people and ML enthusiasts need to take into account some ML disadvantages. For example, how it works with large amounts of data requires some resources to learn, but ML does not guarantee you 100% true results, etc. So yeah, of course, it has great potential, but it’s limited.

Okay, and about some resources, what would you recommend for people who want to learn more about AI and machine learning? 

Regarding the sources, I want to recommend only one site. It’s kaggle.com. That’s the best site for ML engineers and data scientists I have ever known. It’s not an app. Yeah, but it contains everything needed. So, first of all, it has the data set you need for learning. And finding data sets can be a big, big challenge, sometimes a big problem because you need huge data sets with information that should be well structured. It also contains discussions between ML engineers. So it’s also a social network for them. Different models, explanations, and competitions, and I sometimes see big companies come to this site and create their own with Price Fund. And, yeah, it has everything needed. I saw many references to this site. And, yeah, so Kaggle.com. The best you need.

What sparked your interest and motivation to pursue the field of AI and machine learning?

My journey is specific; I don’t recommend trying it my way, but I started by learning maths. So I graduated with a degree in applied mathematics, and I wanted to know in what form we can use maths, in what areas, and what perspectives it has. And I had a great course in machine learning. So I started with maths and then went to ML. So it wasn’t like, Oh, it’s a very hyped thing that we can discuss and start learning and investigating. It was from maths to ML. And, nowadays, it hypes a lot, so… it matches my knowledge.

But it’s an excellent way to follow because maths is the basic thing, and it’s always better to have basic knowledge and then develop it into something new.

Well, as I see from my experience, people need to understand why they need maths in the beginning. They are like, Well, I want to create something great like chat GPT, like Copilot, etc. Why do we study this simple algorithm? What does it mean? What happens there? But in maths, we start with something abstract. For example, from line, vector, array, points, and then it goes from simple to hard. And many people, at this step, drop their learning. So that’s why it isn’t very easy to study ML, I think. And I’ve spent five years studying it. And I still need to learn a lot of logic and terminology.