AI and gaming are inextricably linked. They have been parallelly developed. Every time we see advancements in technology and AI, we can anticipate the same in gaming.
AI plays a crucial role in game development, for example, improving game environments for a more realistic experience, creating new games, testing the ability of AI by having them compete against humans in several games, ranging from chess to MOBA games.
Let's take a look at some of the most recent AI trends in gaming. If you're a gamer, don't miss out!
AI improves non-player characters’ ability and creates a virtual gaming world
Other than our player character, there are NPCs (Non-player characters). NPCs are characters in the game that sometimes play a part in our quest. The developers typically programmed their tasks, behaviors, and dialogues.
AI is used with NPCs to make them smarter and more human-like, allowing players to have a more realistic and personalised experience. The developers of Assassin's Creed Valhalla used AI to create NPCs that are difficult to predict and have their thoughts. Once, one NPC followed a player everywhere for no reason. It might do so because it wanted to. Who knows?
In Watch Dogs: Legion, the developers used Natural Language Processing (NLP) to create NPCs’ own side story. NPCs in that game all have their dialogues, relationships, and activities based on their ploy which doesn’t depend on the player. When we walk down the street, we will see NPCs doing their things. Of course, we may interfere but it also affects their story and certainly has consequences.
AI observes and learns how humans play to interact with us and play for us
Do you know that while we are busy trying to complete our missions, someone is watching us and gradually learning how we play? Some are here to play for us, some are here to learn about our characters and style, and some are here to find some news about us to gossip in their world.
Sony has recently published research on the AI technology that might be used in PlayStation 5. AI will observe and study how we play to mimic our style and even play for us if we want. If there is a part where our AI doesn't understand, it can also ask for help, just like it is our buddy on a battlefield.
There are also other developments to anticipate such as AI selecting the mission that suits our style. AI-controlled NPCs can also observe and learn about our player characters. In Red Dead Redemption 2, an NPC noticed a player character wearing weird clothes and said, "The tailor you went to must have been blind”. Another NPC from Hitman 3 said to a player who approached her “Imagine an invisible bubble around me. A bubble that makes me feel safe and comfortable. You’re in that bubble right now and I didn’t invite you”. These are examples of NPCs reacting to us in unexpected ways, demonstrating AI advancements in gaming. Next, imagine a character who can communicate with us in a conversation that hasn't been programmed yet, with a synthetic voice that conveys its emotion.
AI helps create video games that match our needs
Researchers from the University of Alberta developed AI that helps create video games the way we want. The AI model is trained by watching videos of people playing games that are the references of our ideal video game. For example, if you want a game like Super Mario Bros, show AI the videos of people playing Mega Man and Kirby’s Adventure.
AI will finally understand the rules after watching enough. Then it will use the information to create a Game graph and build the game based on that graph. This AI-invented game is novel but still in the same kind of video games we like because AI learns and builds the game from the references. We can assign AI to gradually build rules and patterns of a new game step by step and we get the whole game created by AI.
Even though you don’t know how to code, you can build a dream-like game by yourself if you have this AI assistant. The system is to diversify video game making and allow non-coders to participate in game development. It does not intend to replace the developers at all.
AI is so advanced that it can beat human players
AI can help us improve, play, and develop games. On the other hand, it has the potential to defeat us in the competition.
In 2020, Alphastar, DeepMind’s AI, achieved Grandmaster level, the highest level in StarCraft II, a strategy game that requires careful planning, analysing, and decision making. This was proved to be one of the biggest successes of AI as it beat almost 90,000 players in European serve and rose to the top 0.15% of all players. Alphastar also defeated TLO and MaNa, two professional players.
Initially, most people argued that Alphastar cheated because its machine-like abilities allowed it to see all of the maps while playing, which is impossible for humans to do. Then, the developers reduced its vision, and MaNa finally won. However, the previous win still demonstrated that AI advances are much more impressive than we thought.
The same thing happened to OpenAI Five, AI from OpenAI that competed in Dota 2 tournament. OpenAI Five beat OG, a world champion team, and also beat Dendi, a famous pro player, in one on one competition. OpenAI then allowed other teams to compete with OpenAI Five. The result was that OpenAI Five won 99.4% of all competitions.
Why wasn't 100%? Fortunately, humans still got a chance to win. Alpha Red, the Thai pro player team was the first to defeat OpenAI Five, followed by three other teams. AI might do a great job but our Thai team is equally competent. It has been two years since these updates so let’s stay tuned for new cutting-edge developments of AI in gaming.