AI is a trending term being used a lot at the moment in reference to a wide range of Machine Learning programs. These programs do not actually possess Artificial Intelligence as it has been imagined in science fiction movies, but rather;
Artificial Intelligence is a field, which combines computer science and robust datasets, to enable problem solving. (IBM, 2024)
There are various types, or sub-fields of AI, but ultimately, they are computer algorithms which make predications or classifications based on the data put into them. (IBM, 2024)
Within the libraries community of academics, researchers and professionals, attention has grown around Large Language Models (LLMs) such as Generative AI Tool, ChatGPT. With free access given by ChatGPT developer, OpenAI, individuals and organisations are experimenting with different applications for the programs, to improve efficiencies and solve other business questions and problems across the world, and in this field alone hundreds of applications have appeared in just the last few years that use this technology.
AI tools are indeed being effectively utilised to improve efficiency; particularly where large data is concerned. AI is at its best when quickly processing great volumes of data that would take a human interpreter a great deal of time to do. However, AI does not escape the fact that outputs can only be as good as the input data.
There have already been numerous case studies that show how biased data can be, towards a particular ethnic group or gender, simply by the volume of that data in the sample being unbalanced. If a data sample is made up of 80% men and only 20% women, on the subject of heart disease for example, the AI outputs are going to be more focused on treating men. (Runciman, 2023)
The information being used might not necessarily be current. With the sheer breadth of information available on the internet, these tools will always, based on text recognition, use the best match, regardless of the date of publication or any consideration for the quality of the source. This effects some tools more than others, depending on the data being used by the AI in question. Suppliers are starting to focus in more on only feeding their AI tools more curated data to use.
There is additionally the known problem of AI “hallucinations” with these Generative AI tools – whereby research results are entirely created by the AI to complete an output where it could find no data to create that from. (STM, 2023)
With these factors in mind, it is possible to use AI to great effect, the key being to understand the limitations of these tools in their current early states of development. AI is being used effectively for information retrieval, summarising collections of information, and more. Amazon have deployed a new AI tool to summarise user reviews on products, and provide a short paragraph that contains the highest recurring points from potentially thousands of user reviews for the benefit of the consumer.
In a library setting specifically, AI is being used to enhance the capabilities of search engines, or to summarise collections of articles or information around a given topic. In this scenario however, the matter of copyright becomes pertinent, in that particularly with free AI tools, the information provided to them will be retained, primarily as a means to improve (teach) the AI tools. A user would effectively be uploading articles to the internet that they do not have the copyright license for. The Copyright License Agency is actively working to address the rising trend of AI applications in the UK, and should you wish to read more on the subject, or stay abreast of these developments, see the following article for an entry point to AI and Copyright: Principles for Copyright & Generative AI | Policy Framework | CLA)
AI tools are still very new, but there are already thousands of tools available and relevant to research, library and information alone, and people are already using them to great effect. While the NHS Wales e-Library do not endorse the use of these AI tools, if you are using them, there are some important things to consider.
The NHS Wales e-Library cannot term any AI tool as “safe”. Below is a list of AI tools that we have encountered, and some points about our experience with them. This is NOT a list of recommendations and is intended purely for information purposes.
To note, some of these tools are free to use, others have elements or versions that you can subscribe to for enhanced functionality. As this list is not intended to endorse any of these AI tools, but rather be a simple list of the tools we have encountered, differing functionalities of subscribed products are not detailed here.
The NHS Wales e-Library is responsible for providing electronic resources for information and evidence to support the delivery of Health and Social Care in Wales. To this end, there will undoubtedly come a time when we are making use of AI tools, much as we do Evidence Summaries and Databases. The e-Library team pay keen attention to technological developments in the research, library and information field, and there will likely come a time when there are AI tools on the market that we can assure our users are safe to use, and effective in the enhancement of the provision of healthcare. Until such a time, we will endeavour to update and maintain this page with any new regulations, policies and guidance relevant to AI in our field.
AI and Copyright: Principles for Copyright & Generative AI | Policy Framework | CLA)
NHS Digital published guidance: AI and Digital Regulations Service for health and social care - AI regulation service - NHS (innovation.nhs.uk)
Leeds_AI_in_libraries_digital_futures_report_2023.pdf
Copyright and AI – Written evidence submitted by Dr Hayleigh Bosher
|
Algorithm |
A process or set of rules to be followed in calculations or other problem-solving operations, especially by a computer |
|
Artificial Intelligence (AI) |
The theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages |
|
Generative AI |
A type of AI that can create new content and ideas, including conversations, stories, images, videos and music. (AWS, 2024) |
|
Large Language Model |
A machine learning model that aims to predict and generate plausible language at a larger scale (Auto-complete is an example of a Language Model) (Google, 2024) |
|
Machine Learning |
The use and development of computer systems that are able to learn and adapt without following explicit instructions, by using algorithms and statistical models to analyse and draw inferences from patterns in data |
AWS. (2024, February 8). What is generative AI? (AWS) Retrieved from AWS: https://aws.amazon.com/what-is/generative-ai/
Google. (2024, February 8). Introduction to Large Language Models. (Google) Retrieved from Machine Learning: https://developers.google.com/machine-learning/resources/intro-llms
IBM. (2024, February 2). What is artificial intelligence (AI)? Retrieved from IBM: https://www.ibm.com/topics/artificial-intelligence
Runciman, B. (2023). Artifical Intelligence: Where Do You Fit In? ITNOW, 65(4), 4 - 7. doi:https://doi.org/10.1093/itnow/bwad109
STM. (2023). Generative AI in Scholarly Communications. STM. Retrieved 12 18, 2023, from https://www.stm-assoc.org/wp-content/uploads/STM-GENERATIVE-AI-PAPER-2023.pdf