US 12,260,482 B2
Controlling safety settings for behavior characteristics of artificial intelligence characters
Ilya Gelfenbeyn, Palo Alto, CA (US); Mikhail Ermolenko, Mountain View, CA (US); Kylan Gibbs, San Francisco, CA (US); Aleksey Tikhonov, Berlin (DE); and Khaled Jedoui, Mountain View, CA (US)
Assigned to Theai, Inc., Mountain View, CA (US)
Filed by Theai, Inc., Mountain View, CA (US)
Filed on Dec. 28, 2023, as Appl. No. 18/399,592.
Claims priority of provisional application 63/436,109, filed on Dec. 30, 2022.
Prior Publication US 2024/0221264 A1, Jul. 4, 2024
Int. Cl. G06T 13/40 (2011.01); G06F 40/35 (2020.01); G06T 19/00 (2011.01)
CPC G06T 13/40 (2013.01) [G06T 19/00 (2013.01); G06F 40/35 (2020.01)] 20 Claims
OG exemplary drawing
 
1. A method for controlling safety settings for behavior characteristics of an Artificial Intelligence (AI) character model, the method comprising:
receiving, by at least one processor, context associated with the AI character model, the receiving the context includes:
receiving a plurality of streams of inputs during an interaction of a user with an AI character in a virtual environment associated with the AI character, the AI character being generated by the AI character model, the plurality of streams including a first stream of a user input received from the user during the interaction, a second stream of environment parameters associated with the virtual environment, and a third stream of events occurring in the virtual environment during the interaction; and
passing the plurality of streams of inputs through a first plurality of heuristic machine learning models to produce outputs of the first plurality of heuristic machine learning models;
determining, by the at least one processor, based on the outputs of the first plurality of heuristic machine learning models, a level of safety of a content associated with the AI character model;
based on the level of safety, adjusting, by the at least one processor, the safety settings associated with the content;
feeding, by the at least one processor, the outputs of the first plurality of heuristic machine learning models and the safety settings associated with the content to a second plurality of primary machine learning models; and
generating, by the at least one processor and via the second plurality of primary machine learning models, the content based on the outputs of the first plurality of heuristic machine learning models and the safety settings, wherein the content includes the AI character and the virtual environment associated with the AI character, the AI character interacting with the user in the virtual environment, wherein:
the adjusting the safety settings includes:
changing an appearance of the AI character; and
removing one or more objects from being displayed in the virtual environment associated with the AI character; and
the determination of the level of safety of the content generated by the AI character model includes:
determining, by the at least one processor and based on the context, an indication that the user has played a specific game; and
selecting, by the at least one processor and based on the indication that the user has played the specific game, the level of safety from a set of levels of safety.