A newly-granted Disney patent hopes to address the issue of AI characters “hallucinating” by giving them memories to pull from.
Memories for AI Characters
Titled Memories for Virtual Characters, US-12718481-B2 was granted on August 25, 2026, after Disney filed it on May 22, 2024. Here’s the abstract:
One embodiment of the present invention sets forth a technique for synthesizing an interaction with a virtual character. The technique includes matching a first message from a user to a first set of memories associated with the virtual character and determining at least a portion of the first set of memories based on a plurality of factors associated with the first set of memories. The technique also includes inputting a first prompt that includes (i) one or more instructions associated with the virtual character, (ii) the at least a portion of the first set of memories, and (iii) the first message into a machine learning model. The technique further includes generating, via execution of the machine learning model based on the first prompt, a first response by the virtual character to the first message, and causing the first response to be outputted to the user.
The technology could be implemented with chatbots, interactive robots, video games, virtual reality, and more.
As the patent notes, these kinds of interactions have traditionally been driven by scripted dialogue trees and/or dialogue flows, limiting what the character can say to a predetermined set of conversational topics with no improvisation.
Large language models (LLMs) and similar generative models are starting to be incorporated into virtual character interactions, allowing more natural conversions. However, because they are trained on large datasets and lack grounding in a consistent character background, they can “hallucinate” incorrect or out-of-character dialogue.
Disney references Retrieval-Augmented Generation (RAG), which supplements a prompt with relevant external information, but is more focused on factual knowledge rather than experiences and personality traits.
Disney’s new technique would match a message from a user to a set of memories associated with the virtual character and then generate an appropriate response.
The memories would be drawn from “knowledge sources” like articles, books, scripts, character descriptions, storyboards, graphic novels, websites, and past chats. There is also a formula for the AI character to naturally “forget” memories over time depending on how trusted the source is and how often the memory is accessed.
This technique would theoretically avoid hallucinations and inconsistencies while being more “dynamic, spontaneous, and natural” than scripted dialogue.
Other recent Disney patents address ride preferences and aversions and projection mapping on smoke.
What do you think of this technology? Let us know on social media.
For the latest Disney Parks news and info, follow WDW News Today on Twitter, Facebook, and Instagram.
The post Disney Patent For Giving AI Virtual Characters & Robots Memories, Preventing ‘Hallucinations’ appeared first on WDW News Today.
