Hello guys. Did you have good weekends? Me too. I had a good weekend playing some games. I am really into a specific game these days.
The game names ‘Detroit: Become Human’. Its background is two thousand thirty eight, U.S.A. At the time, Android has been commercialized as technology advances. U.S.A is a dystopian world where every androids is discriminated as black slaves. Anyway, this game show us advanced future where artificial intelligence is commercialized.
Today, I’m gonna talk about two things that I’ve already told. Game. And Artificial intelligence.
It would be unquestionable that everyone knows AlphaGo. Deepmind, the British artificial intelligence（AI） company which is now belonging to Google, stunned the world when its AlphaGo AI defeated Go world-champion Lee se-dol. Go was considered the most difficult “perfect information” game for computers to crack, but AlphaGo managed it with a revolutionary system built on neural networks and machine learning.
Now DeepMind is turning its attention to a game that will pose an even bigger challenge: Starcraft. Blizzard Entertainment’s real-time strategy hit is one of the most fiercely competitive games played professionally around the world. Also the company is working together with DeepMind to release it as an AI research environment. Similar to Go, Starcraft is a game that we should change strategy according to opponent’s movement.
However, Oct. 31. 2017, Starcraft progamer Song Byung-gu win a brutal victory versus AIs. RTS games is difficult field for computers to conquer. According to researches, RTS AIs have some problems to overcome induced by RTS’s characteristics.
First of all, According to team Ontanon’s research, Uncertainty is certainly a problem for computers. Adversarial planning under uncertainty in domains of the size of RTS games is still an unsolved challenge. In RTS games, there are two main kinds of uncertainty.
. First, the game is partially observable, and players cannot observe the whole game map, but need to scout in order to see what the opponent is doing.

Chess is an example that view whole game map whenever you want. This type of uncertainty can be lowered by good scouting, and knowledge representation to compute possibilities from scouted data.
Second, there are also uncertainties arising from the fact that the game is adversarial, and a player cannot predict the actions that the opponent（s） will execute. team Ontanon’s another research says that For this type of uncertainty, the AI, as the human player, can only build a sensible model of what the opponent is likely to do. So, researchers need more replays and competitions to enable themselves to train and compare different techniques.
Another problem that caused by RTS games’ characteristic is that there are not enough data and research environment to develop AI.
According to Khan Adil, and his team claims the most important thing is strategy and builds in RTS games. If you lose your strategy, no matter how fast you steer, you cannot win.
Ben Weber, Data Science manager at Twitch, says four things to need. First, Open APIS for researchers to build and evaluate bots. Second, Competitions to enable researchers to test their programs. Third, Abundant replays and advanced deep-learning process to train. Lastly, Human opponents to evaluate bot’s performance.
Do you know Artificial intelligence development status? Lately, Google developed calling AI, Duplex and demonstrated to people.
The reason why Google continues to apply artificial intelligence to playing computer games after Go, is to create a universal artificial intelligence that can be applied to many areas of life. This is because algorithms that can learn it oneself that the complicated rules of the game like StarCraft, can be used for example, Medical care or interpretation. Finding efficient techniques for tackling these problems on RTS games can thus benefit other AI disciplines and application domains, and also have concrete and direct applications in the ever growing industry of video games.