Artificial intelligence (AI)-based inclusive prompt recommendations and filtering
An inclusive prompt recommendation system for generative AI utilizes an inclusive prompt recommendation model to provide recommendations of inclusive language to include in a prompt in order to promote inclusivity and diversity of generated content. The inclusive prompt recommendation model is trained to analyze input text to identify situations, such as gaming, storytelling, social media, projects or presentations for work/school, and like, where the user's intent is to generate an image or description of a person. The model is trained to identify patterns associated with ways users have historically incorporated inclusive terminology intext. The system can include an ethical filtering mechanism for ensuring that prompt recommendations do not have language that directly or indirectly promotes bias and/or stereotypes.
1 . A data processing device comprising:
a processor; and
a memory in communication with the processor, the memory comprising executable instructions that, when executed by the processor alone or in combination with other processors, cause the data processing device to perform functions of:
receiving an input text for a content request prompt from a client application via a network, the input text describing content to be generated by a content generating service based on the input text;
before sending the content request prompt to the content generating service, delivering the input text to an inclusive prompt recommendation model as the input text is being received, the inclusive prompt recommendation model being trained to process the input text to:
determine an intent of the content request prompt via the inclusive prompt recommendation model; and
generate at least one inclusive prompt recommendation based at least in part on the determined the intent, the inclusive prompt recommendation including text recommending at least one visible human trait, characteristic, or condition to include in the content to be generated;
returning the inclusive prompt recommendation to the client application;
receiving a completed content request prompt from the client application that includes the input text and the inclusive prompt recommendation;
delivering the completed content request prompt to a content generating model of the content generating service, the content generating model being a model trained to generate output content based on the prompt; and
returning the output content to the client application.
2 . The data processing device of claim 1 , wherein the at least one visible human trait, characteristic, and/or condition included in the inclusive prompt recommendation includes at least one of vitiligo, misaligned eyes, cleft lip/palate, heterochromia, and alopecia.
3 . The data processing device of claim 1 , wherein the intent of the content request prompt includes at least one of avatar creation, character creation, education content creation, and social media post creation.
4 . The data processing device of claim 1 , wherein the inclusive prompt recommendation model comprises a Large Language Model (LLM) which tokenizes the input text and processes the tokenized input text to generate the inclusive prompt recommendation.
5 . The data processing device of claim 1 , further comprising:
performing an ethical filter operation on inclusive prompt recommendations generated by the inclusive prompt recommendation model to detect and remove language in the inclusive prompt recommendations that does not satisfy predetermined ethical standards.
6 . The data processing device of claim 1 , further comprising:
collecting user interaction data and feedback data pertaining to use of the inclusive prompt recommendation model;
generating updated training data for the inclusive prompt recommendation model to reinforce, adjust, and/or update the inclusive prompt recommendation model; and
training the inclusive prompt recommendation model with the updated training data.
7 . The data processing device of claim 1 , wherein the inclusive prompt recommendation model is trained to recognize patterns in prompts associated with inclusion of the at least one visible human trait, characteristic, and/or condition.
8 . A data processing device comprising:
a processor; and
a memory in communication with the processor, the memory comprising executable instructions that, when executed by the processor alone or in combination with other processors, cause the data processing device to perform functions of:
receiving input text for a content request prompt for a content generating service via a user interface of a client application, the input text describing content to be generated by the content generating service based on the input text;
before sending the content request prompt to the content generating service, delivering the input text to an inclusive prompt recommendation model of a local prompt recommendation system, the inclusive prompt recommendation model being trained to process the input text to:
determine an intent of the content request prompt via the inclusive prompt recommendation model;
generate an inclusive prompt recommendation based at least in part on the determined intent, the inclusive prompt recommendation including text recommending at least one visible human trait, characteristic, or condition to include in the content to be generated; and
return the inclusive prompt recommendation to the client application;
causing the inclusive prompt recommendation to be displayed in the user interface of the client application;
receiving user input indicating that the inclusive prompt recommendation has been selected for inclusion in the content request prompt for the content generating service;
in response to receiving the user input indicating that the inclusive prompt recommendation has been selected, integrating the inclusive prompt recommendation into the content request prompt for the content generating service;
detecting a sequence termination command;
in response to detecting the sequence termination command, generating a completed content request prompt that includes at least the input text and the inclusive prompt recommendation;
delivering the completed content request prompt to the content generating service, the content generating service being configured to generate output content based on the completed content request prompt;
receiving the output content from the content generating service; and
causing the output content to be displayed in the user interface of the client application.
9 . The data processing device of claim 8 , wherein the at least one visible human trait, characteristic, and/or condition included in the inclusive prompt recommendation includes at least one of vitiligo, misaligned eyes, cleft lip/palate, heterochromia, and alopecia.
10 . The data processing device of claim 8 , wherein the intent of the content request prompt includes at least one of avatar creation, character creation, education content creation, and social media post creation.
11 . The data processing device of claim 8 , wherein the inclusive prompt recommendation model comprises a Large Language Model (LLM) which tokenizes the input text and processes the tokenized input text to generate the inclusive prompt recommendation.
12 . The data processing device of claim 8 , further comprising:
performing an ethical filter operation on inclusive prompt recommendations generated by the inclusive prompt recommendation model to detect and remove language in the inclusive prompt recommendations that does not satisfy predetermined ethical standards.
13 . The data processing device of claim 8 , further comprising:
collecting user interaction data and feedback data pertaining to use of the inclusive prompt recommendation model;
generating updated training data for the inclusive prompt recommendation model to reinforce, adjust, and/or update the inclusive prompt recommendation model; and
training the inclusive prompt recommendation model with the updated training data.
14 . The data processing device of claim 8 , wherein the inclusive prompt recommendation model is trained to recognize patterns in prompts associated with inclusion of the at least one visible human trait, characteristic, and/or condition.
15 . A method for a content generating system, the method comprising:
receiving input text for a content request prompt from a client application, the input text describing content to be generated by a content generating system for the content request;
before sending the content request prompt to the content generating system, delivering the input text to an inclusive prompt recommendation model as the text is being received, the inclusive prompt recommendation model being trained to process the input text to:
determine an intent/context of the content request; and
generate at least one inclusive prompt recommendation based at least in part on the determined intent/context, the inclusive prompt recommendation including text recommending at least one visible human trait, characteristic, and/or condition to include in the content to be generated;
returning the inclusive prompt recommendation to the client application;
receiving a completed content request prompt from the client application that includes the input text and the inclusive prompt recommendation;
delivering the completed content request prompt to a content generating model of the content generating system, the content generating model being trained to generate output content based on the prompt;
returning the output content to the client application.
16 . The method of claim 15 , wherein the at least one visible human trait, characteristic, and/or condition included in the inclusive prompt recommendation includes at least one of vitiligo, misaligned eyes, cleft lip/palate, heterochromia, and alopecia.
17 . The method of claim 15 , wherein the intent/context of the content request prompt includes at least one of avatar creation, character creation, education content creation, and social media post creation.
18 . The method of claim 15 , further comprising:
performing an ethical filter operation on inclusive prompt recommendations generated by the inclusive prompt recommendation model to detect and remove language in the inclusive prompt recommendations that does not satisfy predetermined ethical standards.
19 . The method of claim 15 , further comprising:
collecting user interaction data and feedback data pertaining to use of the inclusive prompt recommendation model;
generating updated training data for the inclusive prompt recommendation model to reinforce, adjust, and/or update the inclusive prompt recommendation model; and
training the inclusive prompt recommendation model with the updated training data.
20 . The method of claim 15 , wherein the inclusive prompt recommendation model is trained to recognize patterns in prompts associated with inclusion of the at least one visible human trait, characteristic, and/or condition.