Libraries in the Artificial Intelligence Era
by Mary Aycock
Did you know that AI advances have outpaced even Moore’s law, the principle that the speed and capability of computers can double every few years? (https://spectrum.ieee.org/ai-training-mlperf) Trying to keep track of all the advances practically requires an AI assistant! The stakes for machine learning seem particularly high, not only in the tremendous potential for social benefit, but the potential harm.
For instance, the AI in our phone has increased the efficiency and convenience factor until modern day phones bear little resemblance to the flip phones that appeared in the 90’s. However, the encroaching algorithms track our behavior and movements, all in the service of profiling and monetizing our behavior for corporate profit, despite the fact that we didn’t consciously opt into such services. In authoritarian countries, the phone becomes a partner of the state, tracking and crushing dissent with all the tools at its disposal.
Similarly, AI could expand into the library sphere in either an equitable manner or an inequitable manner. Facial recognition could be a convenience for identification or a tracking mechanism. Trained chatbots can answer questions immediately, but may also fail to meet other users’ needs. Natural language processing can discover patterns in digitized material and metadata and even expose past prejudices. But if used without consideration, NLP can perpetuate bias from previous decades that lurks in historical documents.
The current library vendor ecosystem is dominated by a few, resource-wealthy companies. One of the vendors at the top of the pyramid is in actuality an analytics company. Who will safeguard our users’ data against monetization when there are few alternatives for access? Can libraries speak to power? Many libraries are promoting open scholarship as one alternative.
Additionally, will libraries also be willing to invest in the expertise to understand the AI behind the curtain? Not only to prevent exploitation of users but to explore the possibilities of using machine learning for the benefit of their community? Even if they are willing, can they marshal the necessary resources to do so? It is highly unlikely that libraries can afford to hire AI experts: will they be willing to invest the time and resources to grow their own experts to foster innovation?
If not, corporate interests may fill that vacuum, harvesting data in the interests of profit rather than privacy. As Jason Giffrey said in the introduction of Artificial Intelligence and Machine Learning in Libraries: (https://journals.ala.org/index.php/ltr/issue/view/709)
‘And, like all emerging technologies, if we don’t understand it, don’t experiment with it, and don’t build some of our own tools, we will be beholden to the commercial entities that trade our failures for our money. ”
However, I’m inspired by the growing community of researchers and activists raising awareness about biased algorithms. I’m also encouraged by the organizers, presenters, and fellows of this institute, not only in terms of their technical knowledge, but their willingness to learn about and experiment with this advancing field.
Libraries have served their users for over a century as the ground shifts through successive eras, such as industrialization, the shift to online, and finally this rapid advancement in AI. Both our institutions and ourselves have a role to play with AI, as a trusted organization. Will we meet the challenge?