In the past, AI was only partially implemented in titles, namely for smart NPC behavior and the dynamic feeling of an in-game environment. After the advancements in in-game visuals and overall gameplay, game studios like EA have begun exploring the integration of artificial intelligence to automate the core mechanics of games.
The prospects of using AI to automate core game mechanics mainly only exist in the form of patents for now. Many ways are being scoured to take advantage of machine learning models for video games, such as automatic prop placement to populate maps and generating in-game music unique to each player using trained AI.
We also conveniently bumped into a recently published patent by Electronic Arts dubbed “AUTOMATED REAL-TIME ENGAGEMENT IN AN INTERACTIVE ENVIRONMENT” that examines generating real-time events in games using machine learning models. “Real-time engagement” stands for real-time in-game events, and “interactive environment” means games, as stated in the patent.
- EA’s recently published patent seeks to automate generating various game challenges on the spot to simulate real-life events; in turn, players can earn in-game rewards like coins.
- Machine learning models will automatically generate events according to players’ preferences, such as difficulty, and player engagement, among other factors.
- The patent only talks about multiplayer titles, using NFL game examples in-depth.
- However, the patent can be utilized for various genres like sports-themed, simulation titles, first-person shooter games, or the like.
The patent by EA states, “The disclosed system provides for producing a series of game challenges that replicate scenarios of a real-life event (e.g., a live professional sporting event), and soliciting users to engage the series of game challenges to win an in-game reward. The reward may be a coin award that may be linked to a user’s game profile.”
The legal document discusses the hindrances of manually composing each challenge, “Game players want to play challenges relevant to what is occurring in real life in a professional sporting event, such as the NFL. However, gameplay experiences are typically crafted by humans and take hours to days to craft.”
It continues, “Given a stream of data that contains minute-by-minute real-time game activity (provided by the NFL), a system can be built that searches the stream for “interesting” events, such as touch downs, interceptions, important tackles, etc.”
EA describes the patent’s prospect by using NFL game examples. The patent seeks to automate real-time event generation to produce real-life events/scenarios to increase engagement. These real-time events will potentially resolve the feeling of monotony in each match in a sports game.
The machine learning model will be trained to imitate real-life events and scenarios, “An objective is to create challenges in an interactive environment that sets the game state (e.g., the field, the ball, the players, etc.) as they are set in real life and then challenge the user to accomplish the same goal that the real players have accomplished.”
The events will be dynamically generated by considering real-life occurrences. EA’s patent further cites, “the created event is intended to correspond to a live event that occurred in the real-life sporting event. In this regard, a measure of enjoyment can be determined from the real-life sporting event.”
The method will consider various variables to craft the most appropriate of events for each specific player. The AI system will produce events by assessing the players’ engagement. It cites, “Player engagement can be measured by determining which challenges appear most interesting to the players.”
EA will also keep the players’ experience and level in mind. It elaborates, “Player success can be measured and categorized based on player level to determine which challenges are most difficult for different players. Machine learning can be applied differently to customize which events are presented to which users again to optimize for the most engaging experience per user.”
It continues, “For example, easy challenges are given to more novice players and harder challenges for more experienced players, but based on a per-player measurement of level and engagement, so the events each player interacts with is entirely customized for them and optimized for their enjoyment.”
A detailed description of how a machine learning model is trained is also described. “In an approach, the events are provided to multiple users to determine which events are more engaging (e.g., more enjoyable) to users. The results of the testing can help produce an algorithm (or model) to decide which of the events are more interesting, more winnable, more engaging, more playable, etc., to the users.”
The events or challenges are vastly broadcasted to various users once they are noticed to be the most intriguing or applicable. “It is widely distributed. Based on the results, the interactive environment containing the events that are working (or causing greater engagement as decided by the algorithm) is widely distributed in an automated manner.”
The patent also gives a few examples of events that AI may generate by copying a real-life scenario, one being, “A notification of the engagement event is provided for display that reads “Pass for 20 yards to earn a Coin Reward!” The notification alerts the user during the gameplay to initiate and complete the game challenge.”
It is worth noting this system only discusses multiplayer games, and there is no mention of a single-player title. However, the method may be utilized for various genres, as explicitly mentioned, “The application [can be] a sports-themed video game, a real-world life simulation video game, a first-person shooter video game, or the like.”
This innovative event generation method may be seen in EA’s future Madden NFL games. This discussed system may also tackle the current criticisms of redundancy found in recent EA sport-genre games. With Electronic Arts divorcing Fifa, these AI systems may also influence its upcoming “EA Sports FC” soccer franchise.
What are your thoughts about automating real-time events on the fly in multiplayer games using machine-learning models? Do let us know your opinions in the comments below.
Thanks! Do share your feedback with us. ⚡
How could we improve this post? Please Help us. ✍