Over the past five years, the FIU Libraries have developed and implemented various machine learning and AI technologies with the goal of improving discovery and access to materials as well as provide new methods for analysis of content. A sampling of these projects include the development of resource recommendation functionality using machine learning, which is embedded into our digital library system; the use of Microsoft's Cognitive Services AI for transcription of audio files as well as translation of text in over 60 languages; the evaluation of serval AI systems and training data sets for facial recognition in archive photographs; and text analysis within a large corpus of dLOC newspapers. This presentation will provide an overview of these projects, the limitations of these technologies, the inherent issues regarding the biases baked into the decision making in algorithms and of training data sets, as well as the potential privacy issues these projects may raise. The presentation will also cover the ways the FIU Libraries have worked to mitigate these issues, in addition to the benefits these technologies have brought in terms of patron experience, technical infrastructure, and professional development.
