Meet academics who reject the use of generative AI

Daniel Crowley is tired of people telling him to use generative artificial intelligence (genAI). A marine zoologist at Bangor University in the UK, she says she’s about the only PhD student in her year who doesn’t use it. She has seen colleagues use genAI tools to get the tone of their coding and emails just right. At one point, she was even encouraged by an instructor to use it to create a poster for a conference.

She says her colleagues were often surprised to hear that she had never tried it and suggested she use it for things like coding. “A lot of people said, ‘Yeah, but you have to use that,'” she recalls. But Crowley has her own reasons. She has concerns about copyright ethics, a lack of transparency from companies about how data is used, the environmental impact of AI tools, and the accuracy of what genAI models spit out.

She also believes that using the tool is counterproductive to studying. “Coding is a skill I want to learn and develop because it’s not something I’m the most confident in,” she says. She would rather learn from her mistakes and try it herself.

GenAI has been a hot topic in recent years as technology companies race to release the best models to the public. Researchers use these tools for tasks such as: write a paperpeer review and coding. It can save time, mental energy, and sometimes money. However, Crowley and others who are intentionally abstinent often find themselves criticized by their peers.

“A lot of people say, ‘This is the future, everyone is using it,'” she says. Not drinking, she continued, “feels like you’re going to a function and saying you don’t drink.”

It’s efficient, but at what cost?

According to nature A survey of nearly 5,000 researchers published last May found scientists divided on the ethics of using AI in academia. Over 90% of respondents felt it was acceptable to use AI to edit or translate their own text, but fewer respondents were open to the idea of ​​using AI to directly generate text. Additionally, fewer respondents said they had actually used AI tools at work. About a quarter of respondents had used them to edit a paper, but only 8% had used them to translate, summarize, or write a first draft.

More recently, a survey of 3,234 researchers published last November by academic publisher Elsevier found that 58% of researchers use AI in their research, up from 37% the previous year. When asked about how researchers use or want to use AI tools, 61% said they would use them to search for new research, 51% to gather and summarize literature, and 41% to prepare grant applications. Those surveyed were generally positive about technology’s potential to improve efficiency.

Hugh Possingham, a mathematician and conservation scientist at the University of Queensland in Brisbane, Australia, is another researcher not using AI. He has made a conscious effort to avoid any kind of genAI, instead pledging on LinkedIn to rely on “nature’s stupidity.”

“I’ve never used them at all,” he says. Even though AI is embedded in many everyday functions, he has never clicked a button to generate or summarize text when writing an email, for example.

He specifically complains about the mistakes he finds in AI-generated texts. AI sometimes hallucinates and provides false or misleading information with confidence. “I read a master’s thesis where the person who was quoted had died 10 years before the paper was published, which is a remarkable gesture,” he says.

Why do universities need to fundamentally rethink exams in the AI ​​era?

Audrey Moore, a chemist at McGill University in Montreal, Canada, has also seen AI make mistakes in her field. She has seen various representations of chemicals produced by AI go wrong. She first noticed this in a chemistry magazine that featured “pointless” molecules, but it became enough of a trend in presentations and other magazines that she and her colleagues wrote commentary articles.¹ We are calling on the chemistry community to ban the use of genAI for certain tasks.

“It’s like asking a 3-year-old to draw a picture of a chemical,” Moores said. And the AI ​​model “has never gone through a chemistry course like a human would,” she added. Even if you’re given the task of drawing a simple molecule like caffeine, you’re likely to fail. (natureLike other publishers, we have guidelines that prohibit the use of AI-generated images. )

And cynics say that validating the information generated by AI often defeats the purpose of using tools for efficiency. Tanisha Jowsey, a social scientist at Bond University in Robina, Australia, said that as the faculty’s designated “AI champion,” she is tasked with evaluating models, figuring out what they do best, and suggesting how faculty can use them. But ironically, checking them creates even more work, she emphasizes.

She says that 95% of the time, “it’s faster to do it yourself than to have a tool do it for you and then check to see if it’s done correctly.” She also finds it to be an ineffective tool for qualitative research. This is a view she expressed in an explanatory article she co-authored² Posted on preprint platforms SSRN.

Other disadvantages

Another major concern, researchers say, is the impact on the ecosystem. The data centers that support the genAI system use large amounts of energy and water. Research published in pattern estimates that in 2025, the carbon footprint of AI systems worldwide could be between 32.6 million and 79.7 million tons of carbon dioxide, and the water emissions between 312.5 billion and 764.6 billion liters.³. To put that in perspective, this is equivalent to the carbon footprint of the entire city of New York, the study says.

Potentially harmful environmental effects are one of the many reasons Crowley refrains from using genAI tools. “Especially if my project is tackling climate change, I didn’t feel it was appropriate to use this tool that basically does the same thing as other tools but uses more energy to do so,” she says.

Other scientists also cited ethical concerns. Sustainability scientist Juan Rocha believes that using AI tools could help large private companies train their algorithms to become even better, potentially replacing human workers in the long run. “You are being used by the AI, but you are not using the AI,” says Rocha, who works at the Stockholm Resilience Center in Sweden. “By giving away freedom, we are making the workforce obsolete in the future and making university work less important.”

Michaela Sokoloff, a psycholinguist at the Massachusetts Institute of Technology in Cambridge, also has concerns about how AI collects information. “The main reason I am against generative AI is that it is trained on the work of authors who have not given their consent,” which she considers plagiarism, she says. “This is just stealing the work of writers and artists.”

write a complaint

AI-generated text is an area that Elizabeth Wolkovich also sees as problematic. A conservation scientist at the University of British Columbia in Vancouver, Canada, she says she’s tired of reading student papers generated by AI. Therefore, she has decided to no longer chair dissertation defense committees or serve on graduate student committees where students are using AI in their writing. And her lab can only use genAI in certain cases, like spell checking.

For her, outsourcing her writing to genAI is a way to outsource the opportunity to develop new thinking. “I try to train students how to communicate their research, especially as climate change ecologists,” she says. “You’re coming to grad school to develop your skills, and you’re working with me to learn it from me. I don’t know if generative AI knows how to do that well.”