What is Artificial Intelligence (AI)?

According to Gartner statistics, the number of enterprises using AI grew by 270% from 2015 to 2019. And the AI market will grow by 12.3% to $156.5 billion by the end of 2020, despite the COVID-19 crisis. In this blog post, we will define what artificial intelligence is, and why AI and machine learning are not the same thing.

Definition of AI Artificial intelligence is a technology within the domain of computer software and hardware development. It allows one to create programs that can perform tasks that have traditionally required human intelligence to perform, tasks such as speech or image recognition, decision making, information analysis, and translation from one language to another.

To better understand how AI differs from conventional programs, consider the following example with a chatbot.

Conventional Chatbot (e.g. Without AI) The programmer enters specific replicas and triggers, to which the chatbot responds with prepared phrases. For example, to the phrase “I want to order pizza” the chatbot would provide a pre-programmed answer: “What size: 10 or 12 inches?”

However, phrases like “Bring me a piza, I’m starving” or “I want a 12-inch pizza” will put such a bot in a stupor. The programmer will not be able to predict all possible user requests and prepare a response for each of them. AI-Powered Chat Bot Artificial Intelligence enables a chatbot to analyze vocabulary, compare similar queries, and “catch” the general context of a phrase. The bot will understand that “piza” is just a miss-spelling of “pizza”, “bring” is the equivalent of “I want to order” and “I’m starving” implies the urgency of a request for food. Such a bot can also learn over time, so that it will better understand users and respond to their requests. After previous interactions with a user, or through analysis of their previous food orders (or even their browsing history), an AI-powered bot may actually predict what kind of pizza a person would like.

There are tasks that conventional programs cannot cope with at all. For example, only artificial intelligence can recognize faces and voices, sort images and solve other more “creative” tasks. When it comes to chat bots in particular, AI’s key advantage is that it can recognize nuances and emotions for the user, such as in the above example where the AI did not offer any more “upsells” to the frustrated user and went on with finalizing the order.

Artificial intelligence is utilized for various tasks. One of the first well-known AIs, Deep Blue, was created in 1992 to play chess, and in 1997 it beat the reigning world champion, Garry Kasparov. Later on, other AI-powered programs began to appear for playing games such as Go, poker or the computer game “Dota 2”.

AI tools are also being developed to automate production, forecast demand for goods, or block suspicious banking transactions. These are commonly specific applied challenges and tasks that used to take a lot of manpower in churning through large data sets to accomplish (or we simply impossible/impractical to do), and an automated AI-powered tool brings tremendous value to businesses.

AI is often better at doing routine data-intensive tasks, is faster than humans, and, when well-trained, generally makes fewer errors. But he still does not work well in non-standard or unpredictable situations and can only solve very specific, applied problems. Artificial intelligence is still far from able to completely replace humans.

Types of Artificial Intelligence

As of right now, any existing artificial intelligence is far from human in terms of cognitive capacity. Even the most advanced AIs are unconscious and not self-aware. They can learn, but these are just algorithms, capable only of solving specific problems, as opposed to displaying true creativity of the full spectrum of human ingenuity. Such AIs are called “weak” or “narrow”.

A common theory is that humans will develop a “strong” or “true” AI someday. It will be close to a real person: become self-aware, be able to work in different conditions, solve atypical problems, feel emotions, and be capable of creation, inventing something completely new. Such AI does not exist yet.

Some scientists believe that it is impossible to create “strong” artificial intelligence. On the other hand, the AIs currently being developed now seemed impossible before: they can communicate by voice, write texts, draw, and recognize faces. While these are still comparatively simple algorithms, when compared to the complexity of the human brain, they are improving every day.

Weak AIs can be divided into two further groups:

Limited AI

Such programs are capable of solving only one specific problem. For example, a chatbot may be able to communicate with customers but is unable to monitor production sensors or predict demand.

Universal AI

Such AIs are capable of solving several different problems. To do this, they do not need to be reprogrammed – it is enough to teach them a new trade. One such AI is IBM Watson. It is famous for winning the Jeopardy intellectual quiz, but it is also used for diagnostics, linguistic analysis, financial advice, and a host of other tasks. In the below video, IBM Watson beats connoisseurs in an intellectual quiz.