General technology trends

Artificial intelligence is one of the most popular terms of recent years, but let’s figure out what actually stands behind this phrase, and whether the “terminator” will destroy humanity one day during the robot uprising.

There are three concepts that you will most often come across in the same sentence, namely AI, ML and Deep Learning. The first mention of artificial intelligence dates back to 1956, when the American computer scientist, the founder of functional programming, John McCarthy, first hypothesized that machines can perform tasks that are characteristic of human intelligence. These tasks can be divided into planning, speech recognition, sound and pattern recognition, learning and problem solving.

At the same time, all artificial intelligence can be divided into two large categories — general and specialized. General artificial intelligence should be able to solve any intellectual tasks that a person can do, and specialized, as the name implies, specializes only in solving one problem, like driving a car.

Machine learning is a way to achieve artificial intelligence using various approaches, for example, when we train algorithms to recognize images. At the same time, Deep Learning is just one of the approaches of machine learning (in addition to this, there are also decision trees, inductive logic programming, clustering, reinforcement learning, and others), in which algorithms mimic the behavior of human neurons in the brain.

One of the first industries where machines began to build their own intelligence was games — perhaps because of human curiosity and the desire to defeat not a living player, but a machine algorithm. In 1952, a computer program was created in the USA that can play 6x6 chess, without bishops. Three years later, a program for playing checkers was presented, and in 1957 the first computer program appeared that played a full-fledged version of chess. However, there were still several decades left before the victory of the machine over man.

In 1985, the United States began the development of the ChipTest program, which later became the basis for the creation of the Deep Blue chess computer from IBM — the same one that beat the world chess champion, Garry Kasparov, in 1997 in a rematch (the first match took place in 1996 and the computer lost).

There was also a significant breakthrough in checkers. Canada developed the Chinook program in 1989, which was able to win the world championship by competing with people. Chinook is currently the best checkers player on the planet, so far no one has been able to beat this program. Can you give it a shot :)

Go took the longest. It’s an ancient and complex game with simple rules: you can learn to play it quickly, but it takes decades to hone your skills. However, the AlphaGo program, developed by the British company Google DeepMind, nevertheless defeated the Korean 9-dan pro Lee Sedol in 2016, and in 2017 was able to win three out of three matches against Ke Jie, the best player in the world ranking.

The research results obtained in the study of game algorithms are also used in medicine and chemistry. For example, the algorithms used to create AlphaGo are now being refined to calculate complex protein structures. It is assumed that research conducted using game algorithms will help defeat Parkinson’s disease in the future.

The computer already plays checkers, chess, go and poker better than humans. According to researchers from Cornell University (USA), artificial intelligence will surpass humans in all other games in the coming decades. OpenAI has been actively developing the project for the last few years when machine intelligence competes against professional Dota players.

Is the list of specialized programs limited to games only? Of course not. One of the most active industries is image recognition is Computer Vision. In its simplest form, the program can recognize one or several images in a photo (like in the Hamburger / Not hamburger episode from Silicon Valley). And in more complex cases, this is face recognition, car license plates in real time, as is already done in China. In addition to image recognition, artificial intelligence developers are working on voice recognition. When we turn to Google, Siri or Alexa, this is the technology that is used.

People’s main fear of AI is well portrayed in cinematography. We are afraid that artificial intelligence will begin to develop on its own and at some point will destroy humanity. But even the generalized algorithm AlphaZero, which uses deep neural networks to evaluate the game, and is able to play not just go, but also shogi and chess, is just a program, albeit one that has learned to play independently, but performs only one function. It is too early to expect the creation of general artificial intelligence in the near future. It’s up to you to decide whether it’s good or bad.

...

Impact on society, economy and state

The widespread introduction of artificial intelligence — even if only its specialized version — will have, and by the way, already has a strong impact on society and the economy. First of all, AI will affect the labor market. It will become an assistant in complex professions and take over simple ones where people can be easily replaced by an algorithm.

Artificial intelligence is already becoming a full-fledged partner for many professions — for example, in medicine, they use the power of the IBM Watson supercomputer, which can understand human speech, which in 2011 took part in the Jeopardy! and defeated 2 strongest players, receiving $1M. At the same time, the computer recognizes tumors in the images, which is what it specializes in now, with fewer mistakes than professionals of the highest level.

Deep Dee is another example. This startup of Belarusian roots works at the intersection of medical technology and artificial intelligence and creates a mechanism for the early diagnosis of diseases based on photographs of the fundus.

It is believed that by 2030, personalized medicine using Augmented Artificial Intelligence will become a reality, and 5 years later, around 2035, the first hospitals without doctors will appear. To do this, four stages must go through (two of which are being implemented right now!):

1. image recognition;
2. forecasting epidemics;
3. creating personalized medicine;
4. automated hospitals that will function autonomously without human intervention.

Interestingly enough, software shows better results than humans even in making managerial decisions. Will it ever be able to completely replace managers at various levels? Hardly. But there is no doubt that a person and expert software will work side by side, complementing each other, because business, like no one else, is interested in increasing its own efficiency.

Another field where artificial intelligence can be used at full capacity is security. In China, a system is already actively being tested that will allow law enforcement officers to instantly receive information about the location of a suspect, if at least one video camera identified him.

However, since the pattern recognition system focuses on the eyes and eyebrows, as well as the general outlines of the face, it can still be misled so that it will not even notice the person in the photo or in the video. You can, for example, wear a medical mask and dark glasses. But it’s getting harder every year as the algorithms improve. And when a social network offers to tag your friend on the photo, think about what that actually means. Algorithms can find anyone in a photo or video not just for fun but for law enforcement agencies.