How does artificial intelligence work? How can it be used? Why is everyone suddenly talking about this?

We have heard about the use of AI in self-driving cars, smart home devices, medical analysis and customer service over the past seven to ten years.

What is artificial intelligence and how is it used in different fields?

In fact, artificial intelligence or AI does not refer to any one specific thing at all, but covers a whole range of technologies. Although AI has been around for decades, it has only become commercially viable in the last few years. It's like a story with a steam engine. When the first steam engine was invented, it was a failure in itself. It was not until years after his death, when a new type of boiler was invented, that the steam engine became popular.

The same can be said about artificial intelligence. The real rise of computer intelligence occurred when the exponential increase in computing power allowed us to process huge amounts of data quickly.

At its core, AI mimics human decision making, that is, the ability to predict outcomes based on input from the environment. The real value of artificial intelligence is that it makes complex predictions efficient and inexpensive.

AI replaces skills, not people. It automates time-consuming tasks that are more efficiently handled by AI technology.

So what are the different types of AI?

In traditional programming, data and rules are passed through the machine to produce an output or response.

In machine learning (a subset of AI), the computer is given large amounts of data (thousands, sometimes millions of examples) as well as the required output, and the computer itself figures out which particular rules lead to those results.

For a real-world application of machine learning, we can look at how credit card companies deal with credit card fraud.

In this situation, the data transmitted to the computer will be the entire purchase history on the body of credit cards issued by the company. The output or "responses" will also be sent to the computer and will include the option to select a particular transaction as fraud or not fraud. The computer will analyze data from millions of transactions to predict rules such as where purchases are typically made, the cost of a typical purchase, and the frequency of purchases to be able to pinpoint rule deviations and flag them as fraud.

If you live in Moscow, and suddenly a purchase for ₽253,000 made in Turkey appears on your card, without an associated purchase (for example, a plane ticket), the program will regard this as a deviation from the rules.

What is Deep Learning?
Let's take a classic example of sorting photos of dogs and cats. As humans, we can easily identify several traits that distinguish cats and dogs: overall size, ear shape, and muzzle shape/length. But it's actually an incredible feat!

In deep learning, the machine is given many thousands or millions of examples of cats and dogs, and the machine chooses the features (or rules) without human intervention, using multi-layered "neural networks" that mimic the neural networks present in the human brain. In a "learning process", the machine's results are compared with the human's to confirm the machine's attempt to generate rules.

In short, a neural network is a series of algorithms whose goal is to understand the underlying relationships and patterns in data, and to make predictions.

Today, deep learning is being used to create efficient and accurate customer support, revolutionize healthcare (including cancer cell detection and MRI image analysis), drive self-driving cars directly, and power the future of education, both in schools and in the corporate world.

The exponential increase in the processing power of machines will increasingly introduce AI into our daily lives.