Search

Evolution of Artificial Intelligence in E-sports

*Sanjana Chib & Tanmay Garg



Introduction


E-Sports is that area of sport where professionals train their mental or physical competitive skills through information and communication technologies.[1] Initially, video games were created without the use of artificial intelligence until somewhere between the year 1940-1950; Atari Game Corporation created NIM, the first-ever electronically computerized game in which the two players had to alternately remove one to three objects from a set of objects and player who removed the last piece would be the winner. However, the game lacked conventional artificial intelligence, but still, it wooed the gamers due to its competitive nature.[2]


AI was taken to another level when games like Pong, Pac-man, Space invaders and Donkey Kong worked on simple AI by paying heed to the enemy’s sensations. Artificial intelligence in gaming started as Finite State Machinery[3], a model of computation based on a hypothetical machine that can exist in one of several different and predefined states. However, only a single state can be active at a time, and hence the machine must transit from one state to another to perform different actions.


Functioning of Artificial Intelligence


There are two types of artificial intelligence in the gaming arena; Simple AI is when the game has been codified with simple programming, which follows the algorithm of ‘if and then’, such as Mario attacks if the player is in range otherwise, it turns left or right. Most of the AI works on decision trees wherein the game developer programmed all the possible outcomes in respect of the player’s move, but it has been designed in such a way that the player can look for loopholes in order to beat the AI, and it becomes complex the moment the AI becomes invincible like the IBM’S Deep Blue which defeated chess grandmaster Garry Kasparov.


Genetic neural network: It is a learning algorithm that states that crossing two good neural networks will result in a better neural network [4]. The weights are generated at random and the agent is subjected to a series of tests; based on these tests, the agent receives a score. To make a population, this process is repeated several times and to be available for crossover, the top 10% of the population is to be chosen. Two parents are randomly selected from the top 10% of the population, and their weights are crossed. There is a small chance of mutation every time a crossover occurs: a random value that isn't in either parent's weights and as the agents adapt to the environment, this process gradually improves their performance.


The snake game is considered one of the most sought after AI games in which the player controls the snake by moving it in different directions, eating apples that are spawned randomly to maximize its size. One apple means an increase in size by one grid. The snake dies if it collides with any part of itself. The researchers have developed a refined deep-reinforcement learning model to enable the AI to play this game. The researchers also employed a convolutional neural network trained with a variant of Q-Learning. According to the results, the AI surpassed all human-level performance with exceptional game score and survival time.


Finite-state machinery: A finite-state machine is a model used to represent and control the execution flow. It is perfect for implementing AI in games without the use of any complex codes. For instance, the ghosts in Pac Man are finite state machines as they wander in the scattering corners of the maze, either chase or evade the player. In each state, they behave differently, depending on the player’s actions. They chase as soon as they are nearly approached; otherwise, they evade when a power pill is eaten. Using FSM, a game developer generates all possible situations and programs each response for every possible situation. If a VR Game is equipped with a neuromorphic chip, it can automatically collect data to make the NCP adapt according to the gamer's actions in the game.[5]


Recent trends in the AI Arena


Non-playing characters: At present, the gaming arena consists of AI bots that control the sequence of actions by assigning pre-conditions, mainly in single-player games where the AI controls everything except the main player, which is also known as GOAP (Goal-oriented action plan) The test dataset can be used to create ‘bots’ that can be made to play games as if they are players, thus exercising it. These bots can be as simple as an external application that generates in-game events, or they can literally take the place of players and adjust in the virtual game environment. These Non-Playing Characters (NPC’S) are often based on intuitive AI, which shows quick reflex as per the player's actions, such as in Call of duty, the enemy takes the shield when a player throws the grenade. “If you touch an extremely hot glass which burnt your hand, would you do that again?” The answer to this will be determined by experience at first and then predicting its outcome lately.[6]


Virtual and Augmented Reality: Virtual reality is the creation of computer-generated simulation of an alternative virtual world where the player is taken inside the game in a way where he does not feel that he is outside the game. The player uses specifically designed hardware such as headsets, consoles, sensors etc. that helps the player to immerse in the virtual world. Half-life’s Alyx and Biohazard’s resident evil are examples of virtual reality.


Augmented reality is when physical elements of the real world are used to create a virtual game experience. Play Station’s Vita having features like 3G gaming, cross-platform play, computer vision, map tracking which collects, sends and processes data which augments reality, revolutionizing the way players play games. Pokemon Go has made use of augmented reality wherein the pokemon are pinpointed on a real-world map. When the player turns on the location using GPS, the camera in the phone screen detects and helps to collect them, giving a realistic experience to the player.


Ray tracing: The technique that makes light travel in video games by simulating the effects of its encounter with virtual objects, just like it does in reality, providing a real-time gaming experience, is called Ray tracing. Primarily, Minecraft was popularized due to its exceptional ray tracing techniques that can make virtual rays of light appear to bounce off objects, cast realistic shadows, and create lifelike reflections.


Gameplay analysis: Some AI platforms utilize artificially intelligent models to analyze the gaming pattern of the player, identify the shortcomings, suggest ways to hone the skills and accordingly train them. Senp-AI provides feedback to the E-sports players and competitively trains them.[7]


Conclusion: The way forward


Artificial intelligence moved forward in leaps and bounds when in 1950, Alan Turning devised the ‘Turing Test’[8] to test the capabilities of a machine that exhibits intelligence, and if the judges are fooled, the machine is considered to be intelligent.


Soon there will be a time when AI bots will outperform humans in E-Sports as AI agents will be programmed using deep-reinforcement learning Google’s deep mind with its Alpha-go did the exact same when it had beaten the strongest go player in history[9]. The gaming bots will be taught various skills such as generalized learning, decision- making and adaptability to its surroundings. The psychological aspect of how an AI will react to the user outcome will be decided by the user engagement, reducing the strenuous effort to interact with it by providing a smooth gaming interface.


Speech recognition and voice assistants will make a hefty contribution towards the bright future of gaming that will run on the commands of one’s voice, and the only goal is to make the interface compact so that it becomes easier as well as stress-busting to play games, answering multiple questions to chatbots can be tiresome, speech recognition could help convert voice into text and recognize the player based on his voice.


AI is neither good nor evil; it is merely a technology that will soon consummate the gaming industry by becoming its staple.


*The authors are scholars from Vivekananda Institute of Professional Studies (VIPS), New Delhi.


(The image used here is for representational purposes only)


References

1. Fanni Banyai, Mark D Griffiths, Orsolya Király, Zsolt Demetrovics.“ Psychology of Esports: A Systematic Literature Review.”Journal of Gambling Studies, 5th March, 2018, pp. 2. Springer Link, link.springer.com/article/10.1007/s10899-018-9763-1.


2. Jon-Paul Dyson. “ What was the first video game?.” Play Stuff Blog, 24th November,2020,

www.museumofplay.org/blog/2020/11/what-was-the-first-video-game.


3. David M. Bourg, Glenn Seemann. “AI for game developers.” 2004, pp.228.


4. Peter Bingesser.“ Designing AI: Solving snake with evolution.”, Medium,25th September 2017, becominghuman.ai/designing-ai-solving-snake-with-evolution-f3dd6a9da867.


5. Shikhar Gupta. Neuromorphic hardware: Trying to put brain into chips. Jun.2019, towardsdatascience.com.


6. Guy W Lecky-Thompson. “ AI and Artificial life in video games”. Course Technology, 2008.


7. “How AI is used in e-sports to keep pro players competitive”. Esportsus. 6th May 2020, www.esportsus.org/blog/how-ai-is-used-in-esports-to-keep-pro-gamers-competitive.


8. Java-t point. “Turing test in AI”. javatpoint.com.


9. David Silver and Demis Hassabis. “AlphaGo: Mastering the ancient game og go with machine learning, 27th January 2016,ai.googleblog.com/2016/01/alphago-mastering-ancient-game-of-go.html.