One of the great hopes for AI within the academy is that it will enable or enhance research—accelerating science and the production of knowledge. AI can be used to interrogate huge sets of complex data, generate classification schemes, derive rules for taking in new information, generate predictions, make adaptations, and either make decisions or assist in decision-making (Siemens et al., 2022). Right now, a common metaphor for the relation between researchers and AI is to use AI as a co-pilot – and indeed, Copilot became the brand name of Microsoft’s “everyday AI companion.”
There are a variety of ways in which AI can advance research or will be able to do so soon.
- Analyzing large-scale datasets. For example, AI can analyze vast amounts of Geospatial data. Another example is that AI trained on 23,000 samples of Alzheimer’s patients allowed scientists to understand when a cell would be predicted to die and are closing in on predictive models that may enable early intervention (Frueh, Sara, 2023).
- Based on ingesting massive amounts of data, AI can make accurate predictive models, such as predicting how particular proteins fold from chains of amino acids into 3D shapes. AlphaFold by DeepMind has been incredibly successful thus far. This will enable rapid analysis of protein folds in dangerous pathogens and then creation of models of pharmacological compounds to block them (Service, 2020).
- Based on the shape that you need a molecule or protein to be to fulfill a particular purpose, AI can design it (Brumfiel, 2023).
- Hypothesis generation. AI can crunch scholarly literature for an entire field and then generate new research hypotheses that humans would evaluate to see if they were worth testing.
- AI has been used to analyze the scholarly record for an entire field searching for a solution to a problem and recovered a long-forgotten technique overlooked by researchers.
- AI can help design and execute experiments.
- AI can run massive simulations.
- Deciphering damaged texts. With one program, damaged ancient texts could be restored with 62% accuracy, whereas historians alone could restore with 25% accuracy. Historians with AI could restore with 72% accuracy (Recker, Jane, 2022).
- Data-wrangling and cleaning.
- Transcription, including research interview transcription.
- Language translation enabling access to research by scholars in languages that the researcher cannot understand.
- AI can be used to summarize information (Roberts, Molly, 2023).
From the perspective of education, AI-enabled applications for personalized learning could be promising. AI might also help students brainstorm ideas, edit, summarize and explore information.
In addition to determining the library’s role in educating or supporting students and faculty in these pursuits, some implications of these uses for traditional library functions are:
- A likely increase in researchers’ desire to mine massive amounts of data from our collections, such as for the purposes listed in numbers four and five above.
- The ability to use AI to translate sources will likely make collecting non-English language scholarship a higher priority. Hopefully it will make this process easier by enabling librarians to translate items titles and description and publisher-catalogs from languages they do not know.
The ability of AI to decipher damaged texts or difficult to read handwriting could make archival material more accessible.
