AI in Gaming: Where Do We Wanna Go With This?
Recently, I encountered what might be one of the most significant advancements in AI for the gaming industry this year: GameNGen by Google.
Google’s GameNGen
(Image captured from a video of Doom created in real-time by Google’s GameNGen)
Using Stable Diffusion, GameNGen analyzes player input and previously generated frames to create new ones in real-time, delivering impressive results.
How Was It Trained?
To gather the training data needed for GameNGen to accurately create its own Doom levels, the Google team trained an AI to play Doom at various difficulty levels and imitate different player abilities. This training produced AI players capable of playing Doom and provided hundreds of hours of visual data for GameNGen to learn from.
How Good Is It?
As a result, GameNGen's version of Doom delivers an authentic gaming experience, including movement, shooting, and realistic damage from enemies and environmental hazards. As players explore, the game dynamically builds a fully interactive level around them in real-time. Running at 20 FPS, it can be difficult to distinguish from the original Doom in short video clips.
This has me thinking: what other AI cases have we witnessed so far in gaming?
Ubisoft's Neo NPCs
At the Game Developers Conference earlier this year, Ubisoft introduced 'Neo NPCs,' a new type of AI-driven NPC (Non-Player Character) capable of holding real-time conversations with players. Unlike traditional NPCs with scripted dialogue, these characters, like Bloom, use AI to create more natural and dynamic interactions
As shown in the image, the player asks Bloom a question, to which he responds in real-time.
Diverse Personalities for Realism
Ubisoft also created NPCs with different personalities to add variety and depth to the game world. Some are friendly like Bloom, while others are more standoffish or challenging, offering a range of behaviors that make the environment feel more alive and unpredictable.
Epic Games’ Real-Time Monitoring
While Epic Games hasn't branded its entire anti-cheat strategy under a specific name, it offers custom AI solutions to maintain a fair gaming environment.
The system actively tracks gameplay data as it happens, enabling swift detection and response to potential cheating. It examines player movements, actions, inputs, and in-game events to spot unusual behaviors in real time.
Its AI algorithms employ pattern recognition to detect suspicious activities, such as aimbots, wallhacks, speed hacks, and macros or scripts.
By combining these technical capabilities, this artificial environment creates greater unpredictability for users, making it feel more like reality.
While these advancements could significantly impact the gaming industry, players in games generally have specific objectives, like hunting monsters to level up or competing against other users.
On the other hand, the metaverse faces a major challenge with its limited content for users to explore, which is why games have often been featured as a primary content offering.
The environment of the metaverse and the game content within it often feel disconnected. Players typically have to enter a separate environment specifically designed for the content and are then returned to the main space once they are finished. However, in the near future, we can expect these two elements to merge into a seamless environment, similar to what we see in movies.
For instance, above is an image from a game called BLADNET. In this game, players must identify real players hidden among AI-controlled NPCs. Players can also disguise themselves by behaving like AI characters to avoid detection by others.
While ethical standards and age regulations must be upheld by creating designated zones for such content, I envision a world where everything is so seamless that your imagination becomes the only limitation.
Connect with me on LinkedIn by clicking the button below, and share your vision for AI development in gaming
