Do not collect equipment, do not look for bunkers, gun law odd poor, unresponsive ...... In a game of human organs count, these special players are like sending courier equipment, unusually unexpected. Even if it is out of consideration for the novice gamer experience, the same kind of "silly sweetness" is encountered in the high-end game, which makes one wonder: can not these AIs be smart?
Today, we talk about the past, present and future of AI players in the game history.
Game AI evolutionary history
The beginning of the video game has the shadow of the game AI, whether it is the earliest oscilloscope tennis or tic-tac-toe, especially in multiplayer online games are also trapped in the network technology is difficult to popularize in the past, stand-alone game AI became the player Reliable companionship, has become the absolute overlord of the game world.
In a sense, archetypal video game AI may have appeared in the classic arcade game Pac-Man. At that time, the AI in the game was starting to rise all over the world. At the time, Pac-Man made a wonderful attempt in this respect. There are a total of 4 different color monsters in the game, and each monster has a different pursuit algorithm, and for all the monsters set a unified target of the pursuit of enchantment. So, in the game at the time, those monsters did not line up to chase the player, but "chase" the player from different paths and ways, with different options at each juncture. This mechanism allows "Pacman" game for modern video game AI made an extraordinary attempt, but also to Pac-Man became the hearts of gamers classic.
Since then, more classic games began to unfold around the AI characters, "Super Mario Bros.", "The Legend of Zelda", "Prince of Persia", "Locker 2" in these all-rage games, AI Have become the key to guarantee the game experience and control the difficulty of the game. At that time the core of the game, in fact, players and AI confrontation, but only through a different way.
At that time AI's ultimate goal is to let the player defeat, and then get a sense of accomplishment. However, the AI at that time still belonged to the stage of "pseudo-intelligence". The actions of related virtual characters in the game were completed based on a series of pre-configured scripts based on conditional language judgment.
But as the genre and style of play evolve, AI's roles and capabilities are changing. The classic example of this is Metal Gear, 1987, in which the first appearance of AI in the game as a character equivalent to the player is patrolled and for the first time added hostility to the player Reaction mechanism, this mechanism in the future also gradually formed a "alloy equipment" in the core of NPC's adversarial alarm mode. The emergence of this AI mode has also further expanded the mission and role of AI in the game.
And as the video game industry has grown by leaps and bounds in hardware devices, algorithms and computing power over the past 30 years, the evolutionary speed of AI in games has gradually diverged from that of human players. In particular, most of the time the game mechanics were testing player reactions and actions. In these respects, AI soon debuted, and first person shooter games are a good example.
From the 1999 "Counter-Strike" to the later "Survival Path", Valve laid the foundation for FPS-based game AI, but in the early years the robots in the CS were always one of the places where the player was going. At that time the robot in the game is an important task is to fill the vacancies, in addition to become a cannon fodder and meat shield has no other effect, because they are basically in accordance with the law of easy to grasp, non-new players can most easily kill them .
However, people who are familiar with FPS games can easily find that the AI in many FPS games has become increasingly "cruel" in recent years, and the gradient has more gradients. In many cases, the most difficult AI combat has far exceeded most players of. In the recent "CS: GO" online Classic, there was a spectacular scene - a robot called BOT Connor, single-handedly killed the members of the rival team.
Nowadays, AI has started to surpass humans in all aspects of different games. Even chess games that are very proud of mankind have been hanged by a number of technology giants such as IBM and Google. Not long ago, Go unexpectedly completely fell to the ground It's
Humans are responsible for inventing the game, while AI is responsible for the conquest of games and players, which seems to be an irreversible trend. But in fact, in the field of the game there is always a battlefield, where AI can not completely overcome the human race.
Real-time strategy, the last piece of mankind "high ground"?
After AlphaGo's 0-0 victory over Kejie in May last year, many people lamented that on the road to "Game," humankind has completely become a chaser. But AlphaGo does not think so, because it has the next goal - "StarCraft."
As early as 2016, AlphaGo beat Lee Seok, Deepmind announced that the future will be the AI challenge "StarCraft" confrontation with the player. After one year has passed, when AlphaGo has been lonely defeated by the Go community, AI is still "crushed" by the top players in the "StarCraft" arena. Google DeepMind practiced for a year, even simple NPC; at the end of 2017 a game, CherryPi from Facebook AI Institute is also 0: 4 swept by former South Korean professional player Song Bingju.
AI in other areas cumbersome momentum in the face of "StarCraft" seems to come to an abrupt end, the reason for all this is the "StarCraft" game-specific mechanism - the fog of war.
Like StarCraft, StarCraft belongs to the category of games, but because of the addition of the "fog of war" mechanism, it has become a glitch that can not cross the AI.
In Go's game, both parties can see each other's current situation at any time, which means that the player does not hide the information and all the information in the past is clear at a glance. The so-called mutual knowledge and understanding depends on the player's thorough understanding and imagination about the rules of the game. Well known as Chess, Go and so on, they are called "complete information game."
In real-time strategy games such as "StarCraft", the fog of war leaves the parties involved in the game unaware of all the information. In this case, the dynamic situation is judged and decided based on the limited information collected, and when should it be done? What, often need to rely on inference and speculation. This is a headache for AI's "incomplete information game", such as Warcraft, Dota, etc. belong to this type of game.
Top AI There is a common phenomenon in this type of game with humankind - winning the battle and losing the war. If you analyze the gap between AI and human beings, intuitively, it mainly lies in two aspects: "big picture" and "rhythm sense".
Like Geek Park in the previous article, "Go win the first move, and then tower, with the game more artificial intelligence how to catch up with humans? "Mentioned in the changes in the strength of these games more dynamic, no one is absolutely strong and weak, the real triumph of the party may be caught at a certain point in time each other's weakness, and this point in time Judgment is a complex decision made through investigation and empirical judgment. Both sides of the game can change the direction of the battlefield through induction and feint. Even in the weak, but also through some psychological games to achieve the transformation of the situation.
When to attack, when to retreat, where to attack, how to divide and disarm, attack and harass, feint and temptation ... In incomplete information, AI is difficult to sort out the logic of the absolute established to support decision-making, it is difficult to correct The timing to make the key judgments.
So, in the past, many real-time strategy games could only raise the AI level by increasing the income of AI resources, building speed, and full-map, etc. Under the same restrictions as the human players, it is hard to rely solely on the operation to win the entire victory .
Where is the future of game AI?
Maybe AI will soon be able to defeat humans in real-time strategy games, but that's not the future of game AI.
With Google, Facebook, IBM and other technology giants in this input and guidance, AI future is likely to defeat humans in real-time strategy games, it is not surprising. More importantly for humans, it is better to take advantage of AI's development to give players a better gaming experience. Just as AlphaGo defeated Ke Jie, AlphaGo did not destroy this 3,000-year-old game, but instead opened the door to a new world of Go to humans through its own advantages.
About the future of game AI exploration, the famous game masterpiece "GTA5" is a very perfect case. This open game with a high degree of realism and complexity is hugely random, giving players an unprecedented gaming experience. Even more interesting is that this game has also become the experimental field of many automatic pilot AI. Intel Labs, Darmstadt University of Technology and Princeton University all chose artificial intelligence to harness the "GTA5" game for autonomous driving.
You know, smarter AI in the game can do more than just lonely defeat opponents, more importantly, based on story development and player characteristics, to provide a more perfect gaming experience, which can be dynamically increased difficulty, of course, can also be Reduce, everything is to make the game play more interesting.
There's no need to worry too much about whether AI will ultimately beat humanity, after all, we have also killed werewolves. If one day AI can really proselytize, interpret expression, tone, and then fool all human win, that is the end of the AI ......