Data and #AI #CILIPConf23

I'm attaneding the CILIP 2023 conference in Birmingham, UK, together with colleague Dr John Israilidis and five of our current Masters students. I'm spending a lot of tme on our Information School exhibition stand, but also taking in some sessions for some liveblogging. The first was the session on Data and AI. My colleague Dr Andrew Cox was the first speaker. His talk included looking at some of the ethical issues, and identifying the role of information professionals. He said there were some things we should worry about - such as the power of big tech companies, information overload, biased information. However there were also reasons to be optimistic - for example applying our exisiting professional skills in understanding of data and searching. Cox highlighted the Library strategy and AI post https://nationalcentreforai.jiscinvolve.org/wp/2023/06/05/library-strategy-and-artificial-intelligence/ which reports on an event organised by the University of Sheffield Information School, and links to a paper you can comment on: Draft for comment: Developing a library strategic response to Artificial Intelligence: Working paper

Then (via a video) Aaron Tay talked about AI and search. He thought that generative AI would improve relevancy, generation of direct answers, and extraction of information. He pointed that algorithms were already used extensively to drive search engine results (as this was increasingly the case, then search results would become less predictable). He said that "ChatGPT alone can't cut it" because it isn't possible to check the source information, and also the training data isn't fully up to date. A solution is "Retriever augemented generation" - retrieving good quality information and feeding the AI the sources, for it to generate an answer. Tay said that Elicit is an academic site that uses this technique. A 2nd combination is ChatGPT and the Scholar.io plugin. A disturbing result from one piece of research he mentioned was that people seemed to trust resultant pieces of work that were not really accurate (e.g. the references inserted dinto the text didn't always support the information they claimed to support). 

The last panellist was Sam Thomas talking on AI in a healthcare context. He pointed out that clinicians were already using AI extensively in clinical practice (e.g. diagnosis) and administration. He thought that it was important to engage the library team in discussing ideas, engaging positively in finding out about AI and how it could be used, identifying training and development opportunities, and considering the ethical implications of using AI.Thomas thought it was important to "focus on your context" - seek out information that can help you understand how AI could be used by stakeholders (what needs might it fulfill, how and why could they use it). He stressed that it should be embedded into what you were already doing (e.g.in teaching literature searching, evidence use). Thomas saw potential for time saving - literature searching (generating search terms; simple explanations etc), writing support, ideas generation, getting initial insights into data that you need to analyse. His final messages were: engage your team; focus on your users; use AI yourself.

Photo by Sheila Webber: fruit shots at the conference!