Google DeepMind and Blizzard will turn StarCraft 2 on Wednesday for learning AI

A division of Google Inc., which is engaged in research and development in the field of artificial intelligence, announced a collaboration with Blizzard. This is a collaboration between DeepMind and Blizzard on a project that allows StarCraft 2 to turn into an environment to study the weak form of AI. The company plans, including the creation of a system that could win the StarCraft champions.

DeepMind previously developed the AlphaGo system, which was able to beat the world go champion several times . A few months ago, AlphaGo came out victorious in a battle with Lee Sedol. Jeff Dean, a spokesman for Google’s research team, even before winning AlphaGo, announced his intention to engage in eSports games.

Speaking about the development of AI for StarCraft, the authors of this project imply a system that will receive the same basic data as the person. That is, this system will not know the location of enemy units and map elements before they are opened. The conditions of the person and the computer will be equal. Now the situation is somewhat different - you can fight with a computer, but he knows everything about the map, so an honest battle cannot call a war with such an adversary.
')
Here is a demonstration of the game of two champions in StarCraft 2:


In practice, this means that Google AI will need to anticipate what the enemy is doing at any particular time in order to adequately respond to the actions of the “enemy”. In addition, you will need to quickly respond to what went off plan. All this will allow to test the next level of capabilities of artificial intelligence. to “In the end, we want to apply these abilities to solve global problems,” said Demis Hassabis, the founder of Deepmind (it was Google that bought it in 2014, and now, based on the achievements of the purchased company, AI is being developed).

Google chose StarCraft because for 20 years this game has been the benchmark among cybersportsmen. Professional StarCraft fights have been held since the late 90s of the last century, in fact, since the game appeared on the market. And now the prize funds, the level of cybersportsmen and StarCraft championships is only growing.

DeepMind has long been engaged in research in the field of artificial intelligence. Developers are trying to solve complex problems in this area, which were previously considered unsolvable. An example is the same go game. Experts have argued that computers will be able to start playing at a professional level in go no earlier than in 10–20 years. DeepMind was able to develop a system that emerges as the winner of the competition with the best of the best today. DeepMind now works not only with StarCraft, but also with other games, including 2D old games with Atari or Torcs, an open-source auto racing simulator.



StarCraft is one of the first real-time strategies. She appeared on the market in the 98th year. Here, each side is distinguished not only by its appearance, but also by the characteristics of its representatives (there are three races - terrans, protoss, and zerg), as well as a common battle strategy. Protoss have powerful but expensive warriors and equipment, and zerg surpass the enemy due to the number and speed. All other things being equal, no race has a significant advantage over the other. These conditions include the number and location of resources in close proximity to the first base, the defensive conditions of the terrain, the location in relation to other players.

In order to get the opportunity to win in StarCraft, the player needs to show the effective use of the resources of his race. Also, he should be able to plan, making short-term and long-term plans, which, moreover, should be flexible enough so that they can be changed when new factors appear. The AI ​​will have to learn how to control the mouse, camera, and how to work with game resources.

Joint developments DeepMind and Blizzard will be available to researchers in the field of games and artificial intelligence next year. Partners create an API that allows the use of bots, the principle of which is similar to those programs that some users use now. In this case, opponents will be able to work with the gaming environment in the form of pixels, but there is with the most simplified game cards in the RGB format. Below is an example of what a game level might look like when using an API.



“Computers can work very quickly, but this doesn’t mean intelligence,” says DeepMind researcher Oriol Vinyals. He also claims that so far the artificial intelligence developed in the company is far from the level of the game that modern champions show. But one should not forget that this company created AlphaGo relatively quickly, and after it was presented, many experts claimed that it could not play on equal terms with Lee Cedol. And indeed, it didn’t work as equals - the computer won 4 times out of 5, showing a higher class of the game.