“Too many things are still uncertain to know how it will change things in teaching and research.”
Dr Noah Buckley’s comment summarises the general attitude among academics towards the rapidly advancing impact of artificial intelligence (AI). Uncertainty about how powerful AI tools will become, and disagreements about whether we are over or underestimating their potential, impede consensus on how they may come to change the research and education landscape.
Despite this ambiguity, many academics have already begun to embrace the power of machine learning (ML) and generative software to speed up research, while others seek to pioneer its use in their own fields.
One such area is that of quantum mechanics, in which Professor Stefano Sanvito says the potential of ML is promising.
“There is quite a vibrant activity in that space and the expectation is that ML will speed up otherwise time-consuming computational tasks,” says Sanvito, Professor of Condensed Matter Theory in the School of Physics. “We are actively working in the field and progress is fast.”
Sanvito is keen to distinguish this kind of use of ML from “quantum AI”, the direct implementation of ML and AI on quantum computers, about which he is more sceptical.
Nonetheless, he expects that the use of ML to solve quantum mechanical problems will be integrated into existing processes to optimise tasks and accelerate the pace of research usually done with conventional computing software.
In other fields, large-scale integration appears less imminent. “I don’t see AI changing how political science is done very drastically, at least in the short term,” says Buckley, an assistant professor in that department: “We will still be collecting and analysing data, the same as before.”
Buckley particularly emphasises the role human beings will continue to play in research: “Original, interesting ideas for research designs or new studies are still going to come from people – those with real, on-the-ground experience in the social and political world.”
Buckley echoes Sanvito’s view that AI tools will speed up research by automating existing tasks, but does not view this as any profound change.
“AI will make some tasks easier to do, which is nice, but those tasks aren’t really what research is all about.”
Professor Jane Ohlmeyer is likewise optimistic about the potential for generative software to accelerate historical research, and plans to employ ChatGPT in a research project recently awarded €2.5 million by the European Research Council.
VOICES aims to investigate women’s experiences of social upheaval and extreme trauma in early modern Ireland, an undertaking requiring the examination of vast tracts of documents.
“ChatGPT can provide remarkably accurate summaries of seventeenth century documents, such as wills, inquisitions, depositions, and allows us, with remarkable speed and accuracy, to interrogate big data and to extract the women and their lived experiences.”
While Ohlmeyer’s project, as Buckley forecasts, remains led by human ideas and research designs, theuse of LLM software is likely to quickly catch on in humanities and social science fields.
Sanvito highlights that perhaps the greatest application of AI tools will be for the purpose of materials discovery, the process of discovering new compounds and superalloys for use in technological advancements. ML techniques have already been used to accelerate the synthesis of new materials, including in work led by Sanvito himself.
He highlights that LLMs can similarly be used to extract information from scientific literature at ultra-high speed – allowing researchers to “read” up to 250,000 papers in a week – a function which can also be applied to speeding up research in the soft sciences and humanities.
The potential for AI tools to massively speed up the pace of research emerges as its most attractive feature across different disciplines. While there are concerns around ethics and integrity, the general consensus among researchers is summed up by professor of biochemistry Luke O’Neill: “AI is to be fully embraced by academics – it’s yet another tool to enhance teaching and learning.” In a world facing increased risk of pandemics and colossal environmental challenges, the potential to accelerate scientific discovery is a welcome boost to an ever more critical sector.
