LLMs now write lots of science. Good

LLMs could also export, through their words, the cultural environment in which they were trained. (Image: Pixabay)
LLMs could also export, through their words, the cultural environment in which they were trained. (Image: Pixabay)

Summary

  • Easier and more lucid writing will make science faster and better

Many people are busily experimenting with chatbots in the hope that generative artificial intelligence (AI) can improve their daily lives. Scientists, brainy as they are, are several steps ahead. As we report, 10% or more of abstracts for papers in scientific journals now appear to be written at least in part by large language models. In fields such as computer science that figure rises to 20%. Among Chinese computer scientists, it is a third.

Some see this enthusiastic adoption as a mistake. They fear that vast quantities of poor-quality papers will introduce biases, boost plagiarism and jam the machinery of scientific publication. Some journals, including the Science family, are imposing onerous disclosure requirements on the use of llms. Such attempts are futile and misguided. llms cannot easily be policed. Even if they could be, many scientists find that their use brings real benefits.

...
View Full Image
...

Research scientists are not just devoted to laboratory work or thinking big thoughts. They face great demands on their time, from writing papers and teaching to filling out endless grant applications. LLMs help by speeding up the writing of papers, thereby freeing up time for scientists to develop new ideas, collaborate or check for mistakes in their work.

The technology can also help level a playing-field that is tilted towards native English speakers, because many of the prestigious journals are in their tongue. LLMs can help those who do not speak the language well to translate and edit their text. Thanks to LLMs, scientists everywhere should be able to disseminate their findings more easily, and be judged by the brilliance of their ideas and ingeniousness of their research, rather than their skill in avoiding dangling modifiers.

As with any technology, there are worries. Because llms make it easier to produce professional-sounding text, they will make it easier to generate bogus scientific papers. Science received 10,444 submissions last year, of which 83% were rejected before peer review. Some of these are bound to have been ai-generated fantasies.

LLMs could also export, through their words, the cultural environment in which they were trained. Their lack of imagination may spur inadvertent plagiarism, in which they directly copy past work by humans. “Hallucinations" that are obviously wrong to experts, but very believable to everyone else, could also make their way into the text. And most worrying of all, writing can be an integral part of the research process, by helping researchers clarify and formulate their own ideas. An excessive reliance on llms could therefore make science poorer.

Trying to restrict the use of LLMs is not the way to deal with these problems. In the future they are rapidly going to become more prevalent and more powerful. They are already embedded in word processors and other software, and will soon be as common as spell-checkers. Researchers tell surveys that they see the benefits of generative AI not just for writing papers but for coding and doing administrative tasks. And crucially, their use cannot easily be detected. Although journals can impose all the burdensome disclosure requirements they like, it would not help, because they cannot tell when their rules have been broken. Journals such as Science should abandon detailed disclosures for the use of llms as a writing tool, beyond a simple acknowledgment.

Science already has many defences against fabrication and plagiarism. In a world where the cost of producing words falls to nothing, these must become stronger still. Peer review, for instance, will become even more important in a gen-ai world. It must be beefed up accordingly, perhaps by paying reviewers for the time they sacrifice to scrutinise papers. There should also be more incentives for researchers to replicate experiments. Hiring and promotion committees at universities should ensure that scientists are rewarded based on the quality of their work and the quantityof new insights they generate. Curb the potential for misuse, and scientists have plenty to gain from their llm amanuenses.

© 2024, The Economist Newspaper Limited. All rights reserved. From The Economist, published under licence. The original content can be found on www.economist.com

Catch all the Technology News and Updates on Live Mint. Download The Mint News App to get Daily Market Updates & Live Business News.
more

topics

MINT SPECIALS