How should AI depict marginalized communities? Technologists look to a more inclusive future
What is the focus of William Agnew's research?
William Agnew's research focuses on ensuring that the outputs of large language models are representative of the communities they reference. As a Carnegie Bosch postdoctoral fellow and organizer of Queer in AI, he works on analyzing the integrity of training datasets for large language models, identifying and overcoming biases across various mediums, and empowering marginalized communities in discussions around privacy and security in AI implementation.
What is the role of Queer in AI within the AI Safety Institute Consortium?
Queer in AI is a member of the National Institute of Standards and Technology's AI Safety Institute Consortium, which aims to advance the trustworthiness and safety of AI systems. Queer in AI's role includes working to ensure that AI outputs are representative of marginalized communities and helping AI developers identify and overcome biases across various mediums, with the goal of making technology more equitable in its application.
How do Agnew's audits address biases in AI datasets?
Agnew's audits address biases in AI datasets by analyzing the integrity of training datasets for large language models, identifying and overcoming biases across mediums like text, images, voice, and music. The goal is to ensure datasets are representative, inclusive, and free from toxic stereotypes, protecting communities' rights over their data and representations.