Careers may end up being determined by who has the best artificial intelligence tools rather than who has the best research as academia enters an AI arms race, a conference has been warned.
Concerns have grown about the impact of AI on the status quo in R&D as the technology’s abilities and its adoption surge. Academics are increasingly using tools such as ChatGPT to help write articles, while funders and publishers have also employed the technology.
“We have to be alert to this idea that we may be entering the race to come up with the most dominant generative AI or proposal software and the harm that that can potentially do,” Dave Lewis, head of AI at Trinity College Dublin’s computer science school, told research managers at a conference in Brussels.
Proposal-writing tools
Lewis believes the use of AI could particularly affect the way researchers submit proposals for grants and how these are sifted and selected.
Some researchers and companies have already created their own generative AI tools for grant proposals, which will “incentivise everyone else to do the same”, he warned at the conference held by the European Association of Research Managers and Administrators.
“If [researcher-developed tools] indicate success in getting funding, everybody is going to start doing that. Do we get into a situation where we are not competing to do the best research but we are competing to do the best generative AI for writing proposals? We don’t want to be going into that situation,” said Lewis.
He warned that a race for the best proposal-writing AI tool could entail certain groups “capturing the market and pulling the rug out from everybody else who can’t afford and doesn’t have the resources to get involved in that.”
AI tools for proposal writing also throw up issues such as who owns proposals and how they are treated in terms of copyright, Lewis said.
Urgent action needed
There is a “reciprocity” that researchers get when they collaborate with others to write proposals, which could be lost if there is a boom in developing generative AI writing tools, Lewis warned. He said that researchers may be less willing to partner up to write proposals if they think their counterparts will just use the proposal to train their generative AI.
Lewis said that the threat of a generative AI race in academia could call into question the usefulness of traditional proposals in allocating money for research. The research community “urgently” needs to look at the process of how it awards funding, he added.
One solution might be different mechanisms in funding allocation, such as more use of interviews and a lottery-style selection process whereby proposals are selected at random provided they reach a quality threshold.