A Sony patent application describes a system that could use machine learning to change how computer-controlled characters behave based on a player's encounters with them. The filing was published on October 1, 2026, after being filed on March 31, 2025.
As Wccftech reports, the system would collect player and game data from at least one encounter between a player and a non-player character. That data would be used as input for a machine learning model, which could generate character content for the game to apply.
The application describes character content broadly. It could cover behavior, dialogue, appearance, or other character attributes. In practice, an enemy might respond differently to a player's tactics in a later encounter. The filing does not specify a particular model architecture or commit to one kind of character change.
The proposal could affect games built to accept AI-generated character updates, but the patent does not establish that Sony plans to use the system in a commercial game. The FromSoftware-style boss comparison in Wccftech's headline describes a possible implication: if enemies adapt to a player's tactics, familiar encounters might become less predictable. That remains a hypothetical outcome, not a confirmed feature.
Comments
0No comments yet. Be the first to comment.