With the term, Artificial Intelligence (AI) usually we describe the ability of a Machine to do functions, reasoning and having a behavior that is typical of a human brain. AI was at first described by McCarthy during a famous interdisciplinary seminar happen in 1956 in New Hampshire. As Marvin Minsky, one of the AI pioneers, described, the scope of this new discipline was to allow machine to do processes and actions that would require intelligence in case a human had to do them. The very final objective of such technology was to create an artificial brain able to emulate a human brain up to the point that it would be undistinguishable from a human one.
Games at first (and videogames subsequently) have been identified as a fertile ground on which AI could be applied and could try to evolve. This is mostly because the environment is fairly simple and limited, with a small number of parameters and variables that have to be observed and taken into consideration for the decision making.
The first prototypes of videogames were created in the late ’40s, but only in the early ’50s the first videogames came out in the market. In 1952 A.S. Douglas while working on his thesis at Cambridge created OXO, a graphic version of tic-tac-toe, in which the opponent of a human player was an artificial intelligence player. However, only during the ‘70s, with Atari’s PONG, videogames reached the mass and became a popular pastime. This videogame was a simplification of ping-pong, with two bars, one at each side, that could be moved up and down, trying to hit a bouncing ball and shooting it towards the other side of display. This game could be played both by due human players facing each other as well as a single player (facing therefore the AI). In this case, AI will play the game as a human player, trying to move the light bar in the position in which the ball was expected to land. Given the fact that the ball direction was computed by the game as well, many times the AI would have end up cheating, since it knew precisely were the ball would have landed.
Over the years, videogames, as well as the AI, became profoundly different, with graphics and gameplay that emulate the reality so much that they are used for training in many scenarios (e.g. Formula 1). The AI in particular tries to play the game as efficiently as possible, using human behaviors and avoiding standardized actions.
For instance, Real Time Strategy games (RTS), as the well-known Age of Empires, the AI proved to be able to create every time different games although the macro-variables (map, civilizations, difficulty and number of players) are the same. This is possible since each AI player has to manage its initial resources and accumulate even greater amounts so that it can support the war effort required by the game, it must be able to form and manage its army, defending strategically important positions such as its extraction and production facilities and systematically attacking the enemy. This bundle of tasks can be done in different ways and with different strategies, factors that will create very different outcomes of the same game. However, these AI could not beat in a match professional (or just highly skilled) players, since no matter how complex their behavior might be, a human brain will be always able to find a way to prevail thanks to tricks and triggers that will lead the AI to an inefficient decision.
Yet, the latest development of AI is incredibly powerful. Thanks to Machine Learning technology and Artificial Neural Network (Adversial Neural Network to be precise), Elon Musk’s OpenAI Bot (https://openai.com/the-international/) after only 2 weeks of learning was the first AI that was able to beat professional players in a complex game as DOTA2, a Multiplayer Online Battle Arena, without AI cheating involved. All the top players of the Internationals (the “World Cup” of such game) challenged the bot in a 1-vs-1 match, but none of them was able to overtake it. Surprisingly although they are the best players in the world, the bot was so powerful that some games that were not even tight matches. The result was so surprising that the developing team decided to allocate a prize of 1.000$ for anyone able to beat such bot. It is no surprise however the fact that no player was yet able to win such prize.
Is any of you going to accept the challenge?
