By Laila Casiano | Staff Writer and Rebecca Robinson | News Editor

Since the rise of artificial intelligence in the media, people have witnessed AI-generative tools tend to reflect racial bias in its preference toward whiteness and its depictions of people of color.  

The Washington Post’s article, “This is how AI image generators see the world,” discusses AI’s creation and imitation of human thinking. It uses AI rendered images to show how the “photos” reflect racial biases and amplifies stereotypes.

Is AI itself biased or merely reflecting human bias and sociology? After all, AI is based on algorithms trained on data provided by humans. 

LaVonda Evans, an administrative specialist and temporary instructor for sociology, said these cliches and stereotypes are not AI-generated, but that AI learns, trains and creates from pre-existing societal concepts made by human-based constructs through metadata information.

“From a sociological standpoint, I would argue that these images are not necessarily racially charged or lies, but rather, they are identifying factors that are attributable to human difference based on region, status and class,” Evans said. 

AI has heavily leaned on stereotypes when rendering images such as leaders being depicted as men, beauty being eurocentric, Asian women being sexualized and other racial stereotypes.

Jacob Canton, Ph.D., an associate professor of philosophy, said the errors in AI stem from pre-existing content made by humans and the way that content is tagged for AI usage. 

“Some of these don’t have any feedback,” Canton said. “There’s no kind of human correction and I really think a source of error for these kinds of problems is, ‘how are these systems released to the public?’”

Regarding making better accurate depictions of people from different racial and cultural backgrounds, Canton said the way people describe things matters. 

Katherine Baker, Ph.D., associate professor of art history, said because AI often draws from art history, AI image-making reflects bias.

“The history of art is filled with problematic and untrue representations of individuals and so that bias is getting put directly into an output,” Baker said. 

Baker said there can be biased representation in art itself and through these artworks being digitized, it is easier for AI to replicate those biased representations. 

“The machine itself doesn’t know how to push back against stuff,” Baker said. “If we continue to have a society that has inherent biases, how are the machines gonna get trained out of it?”

AI’s relationship with humanity is complicated and molded by how society responds to racial bias itself, yet it has room to change alongside society. 

“Humanity (specifically humans) can be artificial (in their own intelligence) but artificial intelligence can never be human,” Evans said. “Regardless of the amount of data training, in-depth algorithms and continuous feeding of information, AI is art and will always imitate life.”

Categories: Arts & Entertainment