Elements of responsible AI for the library community

Beth Rudden delivered an inspiring keynote talk about moving responsible AI forward in our communities at Internet Librarian Connect (ILC), the virtual conference held 16-19 October, 2023. She concentrated on describing the AI revolution, giving a practical guide to AI experimentation, and stressing the need for librarians to be courageous stewards.

Rudden is CEO and Chairwoman of Bast AI and co-author (with Phaedra Boinodiris) of AI for the Rest of Us. In her ILC talk, Rudden started with some comments on the AI revolution, noting that the promise of AI is hyper-personalization. Information, she said, combines data and context, while knowledge is all about information coupled with relationships. Librarians have great power in using AI to help patrons. Just imagine what it would be like to provide the correct information every time, to tell the right story in the right language to every patron walking through the library doors. That is definitely using our power for good.

Diversity is integral to the effective use of AI in libraries. Using the analogy of the “wisdom of the crowd”, Rudden explained that the larger the diversity of a group, the smaller the error rate. Essentially, this boils down to needing diversity to achieve accuracy.

AI Experimentation

Gaining proficiency as well as becoming comfortable with the technology requires experimenting with AI technologies. She asked the same question of four different AI chatbots and compared the answers. ChatGPT was verbose and she wasn’t overly pleased with its reply. Pi.ai gave a more personalized response. She relies on Perplexity.ai to provide not only answers but non-hallucinatory citations. Finally, her company built its own GPT, called GenieCat.

Myths abound regarding AI. It’s important to recognize that AI is simply one model. Actually, there are numerous models for information professionals to play with. Developing AI is not 100% coding. Training data and experimentation are important as well. We hear so much about Large Language Models, which leads us to believe that AI requires large volumes of data and large computing power. Remember that quantity does not equal quality and that small data sets can generate wonderful results.

Transparency is another myth Rudden debunked. If a developer says that AI is a black box machine that can’t be rendered transparent and that what’s going on with the models is hidden, it’s a sign of laziness. Additionally, she is absolutely certain that AI will not take all our jobs.

Ethics is important to moving AI forward in a responsible fashion. She presented three fundamental aspects of ethical AI:

  • The purpose of AI is to augment human intelligence.
  • Data and insights belong to their creator.
  • New technology must be transparent and explainable

Rudden strongly believes that “We were built for unity, but we need courageous stewards”. Librarians should model ethical principles. After all, people notice what you do more than what you say. All AI systems should be explainable and transparent. In line with this, we should always demand lineage and provenance of algorithms. Think about how AI augments people. Ask who was the person the AI was created for? Who was the user? What was the intent of the AI system?

Her main takeaways from her talk were:

  • You belong in AI.
  • Experiment with existing models now.
  • Demand proof, ask the questions.

Internet Librarian Connect recorded talks will be available to subscribers until 31 December 2023 on the Pheedloop platform.