A recent study published by Oxford University Press critically examines the biases inherent in generative artificial intelligence (AI), particularly regarding Australian themes. While big tech promotes generative AI as groundbreaking, the research highlights alarming portrayals of Australia and its people. Using prompts in five popular image-generating AIs, the study revealed outputs steeped in outdated stereotypes, often depicting a homogenized, white-dominated version of Australian life. Notably, images of families showcased white, heteronormative ideals, while depictions of Aboriginal Australians reverted to harmful, regressive tropes. The AI tools often defaulted to outdated cultural representations, reinforcing negative racial stereotypes and ignoring Indigenous Data Sovereignty. Even recent updates to AI models have shown minimal improvement, perpetuating these biases. This reveals a critical need for ethical oversight in AI training data and a reevaluation of how these technologies shape societal perceptions. Given their omnipresence in everyday applications, the implications of biased AI outputs are profound and require immediate attention.
Source link
Share
Read more