Addressing Bias and Underrepresentation in AI: Challenges and Solutions.
David Restrepo
Foundation models have emerged as transformative forces within artificial intelligence, heralding new capabilities in numerous sectors, including healthcare. Yet, the dialogue about their risks and limitations, and their potential to widen social disparities remains insufficiently explored. In the seminar, we will progressively delve into how transformer models have reshaped our understanding of AI interaction in medicine in areas such as natural language processing and computer vision. This discussion will serve as a foundation to address the significant challenges of fairness , bias, and computation that are inherent in these models. The discourse will then shift to explore initiates to increase the representation in datasets and AI usability. Through specific examples, we will illustrate how foundation models and LLMs can be leveraged to bridge language barriers and improve healthcare delivery.