As the Information Science department on campus, the NMSU Library is naturally involved with any information processing technology like AI, especially generative AI. Generative AI is a transformative technology that can serve as an advanced user interface, provide a useful mental prothesis, direct autonomous action, and work with inputs like data and text in ways that more primitive natural language processing cannot. However, it should be remembered that generative AI is a technology with many risks and pitfalls. Hallucinations and spurious performances of expertise by AI models can lead to negative and even terrible consequences. Most people are unaware of the limitations of the technology and the models themselves tend to elide their positionality, making them appear neutral despite taking on the positionality derived from their enormous training corpus of largely dominant discourse. The large commercial (especially the most cutting edge "frontier") models running out of data centers use vast resources and have significant environmental impact. 

What AI Projects is the Library Systems Department Working On?

Recently, Systems completed its first run on the over 17,000 scanned pages (many handwritten) in the Amador Family collection, using an in-house created automated process with an edge model (a 31 billion parameter Gemma 4 model from Google) to generate transcriptions which were then translated into English and Spanish. Currently, the department is running a transcription operation on the Archives and Special Collections Oral History audio files using another proprietary automated process that leverages the Whisper and WhisperX AI models. Spearheaded by Critical Systems Analyst Nirali Patel, work is also being done now to test and fine-tune another model to provide image descriptions for more local collections. All of these projects are running on local hardware to minimize environmental concerns and provide full control over the models. All AI-derived metadata will be labeled as such in NMSU Library platforms, so end users will be aware of its limitations and derived positionality.

Is the Library Doing Any Other AI Work?

Yes! The NMSU Library is deeply involved with AI. Several library departments are using AI in their work and the Library teaches classes on AI literacy. Associate Professor Jason Bengtson is the Head of Systems and he continues to do both practical and research work on AI and other IS topics. See how he explored metadata refinement with ChatGPT or, with his colleagues Assistant Professor Nathan Camp (first author) and Professor John Sandstrom, interrogated the creation of counterfeit citations by language models as a threat to knowledge creation. Bengtson is actively exploring other projects as time allows, including model training experiments and applications of humanities theoretical models to AI interaction and behavior.