AI Can Be Pleased With the Performance in 2019

This year, artificial intelligence has had multiple sources for pride. Fortunately or not, humanity has not come close to the point of technological singularity yet, which, apparently, will come with the creation of a strong AI. The 2019 achievements that bring this event closer will be covered in this article.

Technological singularity is a point in the history of civilization when human-created intelligent systems reach such a level of development that their further improvement will get out of control, ceasing to be explained and controlled by “natural” intelligence.

White flag

Some scientists believe that the onset of technological singularity will not be an intellectual explosion, an avalanche-like stream that instantly sweeps away everything in its path. They believe that development will go smoothly, and the main events that made the progress of AI uncontrollable have already taken place. So, we can consider the retirement of Lee Sedol, a professional Korean Go player, a kind of milestone on this path — the day when humanity abandoned attempts and surrendered.

By the time Lee Sedol declared that his further activity in professional sports no longer makes sense because computers cannot be defeated, he was no longer a world champion and was far from the top places in the ranking of the strongest players. His famous series of games with the AlphaGo computer program from DeepMind were played in 2016. Since then, humanity has not had much success in the game against AI, but the white flag, thrown by Sedol in November 2019, still looks symbolic.

The AlphaGo developer takes on other games: both old ones on the Atari platform and more modern ones. So, by the end of the year, the AlphaStar program became a Starcraft grandmaster, defeating the 99.8% of top-ranking people. If we consider the activities of DeepMind as a kind of hypothetical AI attack on humanity, then the next event is completely in line with this. It turned out that the DeepMind Health, a division of the company, specializing in the development of medical services, helping doctors to make diagnoses, in particular, collected and analyzed the health data of millions of patients without their permission.

Robodogs, the law enforcers

Google could be one of the protagonists of the next significant event in the world of AI, but its robot developer Boston Dynamics was bought by the Japanese technology giant SoftBank back in 2017.

The manufacturer offers its robotic dogs Spots to a narrow circle of customers and limits their use to certain frames. One of them is a ban on the use of these iron animals to harm humans in any way. This is not exactly what 20th-century science fiction predicted: smart machines were supposed to be limited to the three laws of robotics at the program code level. If Boston Dynamics additionally stipulates the non-harm condition during the sale, then it becomes obvious that their robots can potentially harm humans if free rein is given.

Police service is one of the most famous spheres where Spots were used recently. According to consumers from the law enforcement, four-legged robots managed to take part in several “real” police operations, but didn’t harm people in any of them, as they were busy ensuring the security of law enforcement officers in difficult conditions.

No luddism: AI gives work to people

2019 can hardly be called a peak year in terms of the labor market climate for AI specialists. The average salary of Big-Data engineer, for example, is about $126 thousand a year. The Japanese company was looking for AI geniuses and was ready to offer an annual salary of one hundred million yen, which is slightly less than a million dollars.

It is often agreed that the introduction of AI is a sentence to humanity: robots will gradually displace people from their jobs. The results of surveys, however, have shown that, for global companies, the opposite is true: the proliferation of intelligent systems has increased the number of jobs.

It includes marking up homogeneous data for compiling the sets of structured information required for training neural networks (datasets). The flip side of high-tech companies developing AI systems looks exactly like this — every smart machine needs thousands of hours of human time to be smart enough — monotonous, painstaking, unskilled and low-paying work.

The profession of a dataset "procurer" is gaining popularity and is in great demand nowadays. In many poor countries, there are entire data-labeling factories where people hired for pennies sort, label and pack intelligent food for smart machines.

Smart machine nationality

The heads of various states all speak as one about the priorities of national AI systems: it is necessary to speed up the development and implementation of smart machines since this is an important factor in the growing competition between states. What these big words will lead to and how this high-tech industry will develop? BNT will publish a new article on regulating the market for smart machines, where we will discuss the reasons why the laws of robotics fail in real life.

Image courtesy of: AI Trends