A.I. for UVic Librarians -- Internal

Artificial Intelligence as a foundational concept and a field of research, as well as many of the technologies summarized under the AI term, have a long-standing history. They have been existing for decades, with some of the ideas and visions behind AI being even older, as is evident across early scholarly endeavors and popular culture. Mention of Machine Learning at UVic dates back to 1968, the first course on Artificial Intelligence was taught as early as 1976.

To this day, some of the general conceptualizations of AI that can be found in the research literature have not changed much in their approach since they first emerged. Some foundational AI research and seminal publications may still hold relevance today due to the steady nature of certain AI conceptualizations.

This continuity can provide librarians with a stable foundation when they need to navigate the vast landscape of AI research, and librarians can use heritage sources with a long half-life as reference points when curating AI-related resources.

While AI and its encompassed technologies are not new and have been around for decades, it is the latest iteration, Generative AI, that allows users for the first time to interact directly with standalone AI tools. This novel way of access and the unprecendented capacities of these tools make generative AI one of the most rapdily growing commercial markets, which contributes to the incredible pace of generative AI development. The already unique, ever-evolving attributes of generative AI challenge our traditional understanding of information technology.

Librarians must not only discern whether generative AI falls under the umbrella of traditional information literacy, or if it stands as its own entity, due to its novel qualities and the challenges they present for the information sphere (for example for → Fake News). Understanding said novelty of generative AI, librarians should also attempt to make professional deductions that allow them to be empathetic to their patrons' experiences with and reactions to AI and the challenges they might face in engaging with it.

Even though standalone generative AI tools represent a new technological dimension within the larger AI complex, it is important to recognize that even precursors to such generative AI have been ingrained in our everyday professional and personal lives for years. This illustrates and underscores the persistence of AI in the technological landscape. AI is not a passing trend, but an enduring development that will have a lasting impact on the future of various domains and applications.

Librarians should engage with AI concepts in a sustainable way to ensure that their knowledge and resources remain current and relevant over the long term.

Tools like ChatGPT and Bing Chat boast impressive capabilities. Their vast flexibility suits numerous applications. Their accessibility, ability to interact in natural language across various languages, and their increasing level of wide use and adoption significantly amplify their impact.

Recognizing these tools' capacities and anticipating their growth, in respect to capabilities, multi-modality, and level of adoption is vital to understand their impact, not least on librarian work.

Although some of the capabilities of current generative AI tools are unprecedented, they also have their own unique weaknesses and flaws. Not only are they prone to bias and hallucination with corresponding ethical complications, there are also contextual and semantic problems, as well as very specific weaknesses with particular kinds of tasks and topics. For example, caution should be exercised in mathematical calculations, scripting or improving computer code, and analyzing long text documents. Comprehending these issues is all the more important because it can be assumed that the prominence of these weaknesses will be eclipsed by the popularity of the tools due to their strengths.

Librarians need to become aware of the full range of inherent weaknesses of generative AI tools. Their limitations beyond the well-known flaws with ethical implications, such as biases and hallucinations, may include shortcomings at even a pure utilitarian level. Librarians should guide users to approach AI results with caution and emphasize their limitations in addition to their capabilities.

The rate of evolution in the features, performance, and functionalities of these AI tools is unprecedented in tech history. This swift advancement is driven by economic competition and the inherent self-evolving nature of AI.

For AI work in libraries, it is crucial to acknowledge this speed and reflect on its implications, for example what it means for maintaing their knowledge and their resources on AI.

The many technologies subsumed under the term AI, particularly Generative AI, consist of extraordinarily complex concepts and mechanisms that sometimes elude full comprehension even by experts.

To effectively introduce these in library settings and, where necessary, to break them down appropriately for addressing specific user groups with particular information needs, requires continuous learning, updated terminologies, and adaptability.

AI is one of the most inter- and transdisciplinary fields of research, leading to ambigious terminology and fuzzy concepts, with inconsistencies between disciplines and their existing research literature, as well as across publications resulting from extramural and applied research, for example those from the major players in AI development such as IBM, NVidia, Google

Librarians, given AI's interdisciplinary scope, must navigate and bridge diverse terminologies and concepts, offering accurate guidance and promote unified understanding across disciplines.

The introduction of AI presents manyfold ethical and safety considerations, among them inherent and prolonged bias, hallucination, and data privacy.

Librarians must approach these issues with caution, a deep understanding, and a balanced perspective, especially considering the technology's impact on academic settings.