Abstract: Embodiments of the systems and methods described herein provide a virtual object aging system. The virtual object aging system can utilize artificial intelligence to modify virtual objects within a video game to age and/or deteriorate for a certain time period. The virtual object aging system can be used to determine erosion, melting ice, and/or other environmental effects on virtual objects within the game. The virtual object aging system can apply aging, rust, weathering, and/or other effects that cause persistent change to object meshes and textures.

Abstract: Using user-specific prediction models, it is possible to present an individualized view of messages generated by users playing a shared instance of a video game. Further, users with different subjective views of what is offensive may be presented with different forms or annotations of a message. By personalizing the views of messages generated by users, it is possible to reduce or eliminate the toxic environment that sometimes forms when players, who may be strangers to each other and may be located in disparate locations play a shared instance of a video game. Further, the user-specific prediction models may be adapted to filter or otherwise annotate other undesirable messages that may not be offensive, such as a message generated by one user in a video game that includes a solution to an in-game puzzle that another user may not desire to read as it may spoil the challenge for the user.

Abstract: Systems and methods are disclosed for training a machine learning model to control an in-game character or other entity in a video game in a manner that aims to imitate how a particular player would control the character or entity. A generic behavior model that is trained without respect to the particular player may be obtained and then customized based on observed gameplay of the particular player. The customization training process may include freezing at least a subset of layers or levels in the generic model, then generating one or more additional layers or levels that are trained using gameplay data for the particular player.

Abstract: Systems and methods are disclosed for enabling a player of a video game to designate custom voice utterances to control an in-game character. One or more machine learning models may learn in-game character actions associated with each of a number of player-defined utterances based on player demonstration of desired character actions. During execution of an instance of a video game, current game state information may be provided to the one or more trained machine learning models based on an indication that a given utterance was spoken by the player. A system may then cause one or more in-game actions to be performed by a non-player character in the instance of the video game based on output of the one or more machine learning models.

Abstract: Systems described herein may automatically and dynamically adjust the amount and type of computing resources usable to execute, process, or perform various tasks associated with a video game. Using one or more machine learning algorithms, a prediction model can be generated that uses the historical and/or current user interaction data obtained by monitoring the users playing the video game. Based on the historical and/or current user interaction data, future user interactions likely to be performed in the future can be predicted. Using the predictions of the users' future interactions, the amount and type of computing resources maintained in the systems can be adjusted such that a proper balance between reducing the consumption of computing resources and reducing the latency experienced by the users of the video game is achieved and maintained.

Abstract: Embodiments of the systems and methods described herein provide game terrain generation system that can generate height field data from a sketch of graphical inputs from a user via a graphical user interface. The game terrain generation system can use a model, such as a trained neural network, to apply macro and micro topological features on top of the height field data to generate game terrain data. The game terrain generation system can identify boundaries between different styles of terrain and generate transitions between the styles to create a more realistic terrain boundary.

Abstract: Embodiments of the systems and methods described herein provide a terrain generation and population system that can determine terrain population rules for terrain population objects and features when placing objects and features in a three dimensional virtual space. As such, the terrain generation and population system can generate realistic terrain for use in game. The terrain generation and population system can receive an image, such as a satellite image, and utilize artificial intelligence to perform image segmentation at the pixel level to segment features and/or objects in the image. The game terrain system can automatically detect and apply feature and object masks based on the identified features and/or objects from the image segmentation. The game terrain system can place the features and/or objects in corresponding masks in the three dimensional space according to the application of terrain population rules.

Abstract: Embodiments of systems presented herein may identify users to play a multiplayer video game together using a mapping system and machine learning algorithms to create sets of matchmaking plans for the multiplayer video game that increases player or user retention. Embodiments of systems presented herein can determine the predicted churn rate, or conversely retention rate, of a user waiting to play a video game if the user is matched with one or more additional users in a multiplayer instance of the video game.

Abstract: Embodiments of the systems and methods disclosed herein provide a request distribution system in which a request for resources may be executed by a plurality of workers. Upon receiving a request for resources from a user computing system, the request distribution system may select a subset of workers from the plurality of workers to execute the request within a time limit. Once the workers generate a plurality of outputs, each output associated with a quality level, the request distribution system may transmit the output associated with the highest quality level to the user computing system.

Abstract: System and methods for utilizing a video game console to monitor the player's video game, detect when a particular gameplay situation occurs during the player's video game experience, and collect game state data corresponding to how the player reacts to the particular gameplay situation or an effect of the reaction. In some cases, the video game console can receive an exploratory rule set to apply during the particular gameplay situation. In some cases, the video game console can trigger the particular gameplay situation. A system can receive the game state data from many video game consoles and train a rule set based on the game state data. Advantageously, the system can save computational resources by utilizing the players' video game experience to train the rule set.

Abstract: Systems and methods are disclosed for training a machine learning model to control an in-game character or other entity in a video game in a manner that aims to imitate how a particular player would control the character or entity. A generic behavior model that is trained without respect to the particular player may be obtained and then customized based on observed gameplay of the particular player. The customization training process may include freezing at least a subset of layers or levels in the generic model, then generating one or more additional layers or levels that are trained using gameplay data for the particular player.

Abstract: Sequence predictors may be used to predict one or more entries in a musical sequence. The predicted entries in the musical sequence enable a virtual musician to continue playing a musical score based on the predicted entries when the occurrence of latency causes a first computing system hosting a first virtual musician to not receive entries or timing information for entries being performed in the musical sequence by a second computing system hosting a second virtual musician. The sequence predictors may be generated using a machine learning model generation system that uses historical performances of musical scores to generate the sequence predictor. Alternatively, or in addition, earlier portions of a musical score may be used to train the model generation system to obtain a prediction model that can predict later portions of the musical score.

Abstract: Systems and methods are disclosed for converting a player-controlled character or virtual entity in a video game to at least temporarily be under emulated control when certain criteria is met, such as when the player's device has lost its network connection to a game server. The character or virtual entity may continue to behave in the game in a manner that emulates or mimics play of the actual player until the end of the game session or until the underlying connection problem or other issue is resolved, such that other players participating in the game session have the same or similar gameplay experience as they would have had if the disconnected player had continued to play.

Abstract: Systems and methods are disclosed for enabling a player of a video game to designate custom voice utterances to control an in-game character. One or more machine learning models may learn in-game character actions associated with each of a number of player-defined utterances based on player demonstration of desired character actions. During execution of an instance of a video game, current game state information may be provided to the one or more trained machine learning models based on an indication that a given utterance was spoken by the player. A system may then cause one or more in-game actions to be performed by a non-player character in the instance of the video game based on output of the one or more machine learning models.

Abstract: Sequence predictors may be used to predict one or more entries in a musical sequence. The predicted entries in the musical sequence enable a virtual musician to continue playing a musical score based on the predicted entries when the occurrence of latency causes a first computing system hosting a first virtual musician to not receive entries or timing information for entries being performed in the musical sequence by a second computing system hosting a second virtual musician. The sequence predictors may be generated using a machine learning model generation system that uses historical performances of musical scores to generate the sequence predictor. Alternatively, or in addition, earlier portions of a musical score may be used to train the model generation system to obtain a prediction model that can predict later portions of the musical score.

Abstract: Methods for providing contextual video recommendations within a video game are provided. In one aspect, a method includes executing an application that uses a rendering engine. The method also includes determining that a video recommendation threshold has been met. The method also includes providing a current contextual state of the application to a server such that the server selects a video from a plurality of videos based on the provided current contextual state and an index, wherein the index includes output from a vision model applied on the plurality of videos, and wherein the vision model is trained on footage generated by the rendering engine. The method also includes receiving a reference to the selected video from the server. The method also includes providing for display, via the reference, the selected video within a user interface of the application. Systems and machine-readable media are also provided.

Abstract: A system can manage distribution of computing jobs among a plurality of third-party network or cloud computing providers to maximize utilization of available computing resources purchased or otherwise obtained by an entity. The system can determine a dependency relationship between jobs and distribute the jobs among the network computing providers based at least in part on the dependency relationship between the jobs. Moreover, the system can use machine learning algorithms to generate one or more prediction algorithms to predict future computing resource usage demands for performing a set of scheduled and unscheduled jobs.

Abstract: Systems presented herein may automatically and dynamically modify a video game being played by a user based at least in part on a determined or predicted emotional state of a user. Using one or more machine learning algorithms, a parameter function can be generated that uses sensory and/or biometric data obtained by monitoring a user playing a video game. Based on the sensory and/or biometric data, an emotional state of the user can be predicted. For example, it can be determined whether a user is likely to be bored, happy, or frightened while playing the video game. Based at least in part on the determination of the user's emotional state, the video game can be modified to improve positive feelings and reduce negative feelings occurring in response to the video game.

Abstract: Embodiments of the systems and methods described herein provide a virtual object aging system. The virtual object aging system can utilize artificial intelligence to modify virtual objects within a video game to age and/or deteriorate for a certain time period. The virtual object aging system can be used to determine erosion, melting ice, and/or other environmental effects on virtual objects within the game. The virtual object aging system can apply aging, rust, weathering, and/or other effects that cause persistent change to object meshes and textures.

Abstract: Embodiments of systems presented herein may perform automatic granular difficulty adjustment. In some embodiments, the difficulty adjustment is undetectable by a user. Further, embodiments of systems disclosed herein can review historical user activity data with respect to one or more video games to generate a game retention prediction model that predicts an indication of an expected duration of game play. The game retention prediction model may be applied to a user's activity data to determine an indication of the user's expected duration of game play. Based on the determined expected duration of game play, the difficulty level of the video game may be automatically adjusted.