Source: Harvard Law Today Harvard Law Today details a new initiative by the Harvard Library aimed at improving the quality and diversity of data used to train artificial intelligence models. Titled "Food for AI Thought," the project seeks to address the "garbage in, garbage out" problem, where biased or low-quality training data leads to flawed and discriminatory AI outputs. The initiative recognizes the crucial role that libraries, as curated repositories of high-quality information, can play in shaping a more responsible and equitable AI ecosystem. The project will explore how library collections can be leveraged to provide AI developers with better "digital diets" for their models. The initiative was discussed at a recent event hosted by the Harvard Law School Library, featuring experts in law, library science, and technology. Panelists highlighted the risks associated with AI models trained on vast, uncurated datasets from the internet, which often reflect and amplify societal biases. The Harvard Library's project aims to create pathways for AI systems to access its extensive and diverse collections, thereby promoting the development of AI that is more accurate, fair, and beneficial to society. This effort represents a proactive step by a major academic institution to influence the foundational layers of AI development.