Generative Artificial Intelligence: A Double-Edged Sword

Despite the advancements that have come out of the artificial intelligence (AI) boom and subsequent generative AI race, there are a significant number of problems under the surface. With the rise of AI generators in social media and pop culture, these problems amplify as the demand for AI tools skyrockets. These solutions regularly exhibit harmful biases and stereotypes deeply embedded in the data developers use to train AI models. Perhaps exacerbating the immediacy of the issue, generative AI models can create highly realistic imagery, empowering users with malicious intent. This paper explores the pervasive issue of bias in generative AI and its implications. It also examines how these models, often trained on biased datasets, can unintentionally amplify, replicate, and reinforce harmful biases and stereotypes. Additionally, it delves into the objectification of individuals, particularly women, and in a novel study, discusses how a popular TikTok filter that uses generative AI exhibits gender and racial biases while excessively sexualizing its users. While research on AI bias is forthcoming, existing literature referenced throughout this paper underscores its presence across various systems. This paper demonstrates the continued propagation of AI bias as these systems expand and iterate, often with inadequate mitigation efforts. It advocates for ethical considerations, accountability, and transparency in generative AI design and data selection. The Author proposes a novel framework called the Stereotypes, Objectification, Racism, and Datasets (SORD) Framework. The SORD Framework is proposed as a lens to examine the multifaceted dimensions of bias in generative AI. By critically analyzing these components, this paper aims to shed light on the ethical implications of biased generative AI, stimulate thoughtful discussions regarding responsible AI development, and provoke the lengthy process of removing bias from the systems and training datasets to eliminate AI bias at the source. Furthermore, it calls for collaboration among developers, researchers, and engineers to advance research, mitigate biases, and promote algorithmic fairness using the SORD Framework. Future work includes refining the SORD framework and assessing its effectiveness. Lastly, the Author underscores the pivotal role of engineering education in nurturing ethical AI practices, empowering future engineers to address societal challenges stemming from biased datasets and AI systems.