Professional Documents
Culture Documents
I
n December, computational biolo- a mistake in a reference to an equation. The trial convincingly fluent text, whether asked to
gists Casey Greene and Milton Pividori didn’t always run smoothly, but the final manu- produce prose, poetry, computer code or — as
embarked on an unusual experiment: scripts were easier to read — and the fees were in the scientists’ case — to edit research papers.
they asked an assistant who was not a modest, at less than US$0.50 per document. The most famous of these tools, also known
scientist to help them improve three of This assistant, as Greene and Pividori as large language models, or LLMs, is ChatGPT,
their research papers. Their assiduous reported in a preprint1 on 23 January, is not a version of GPT-3 that shot to fame after its
aide suggested revisions to sections of a person but an artificial-intelligence (AI) release in November last year because it was
documents in seconds; each manuscript algorithm called GPT-3, first released in made free and easily accessible. Other gener-
took about five minutes to review. In one 2020. It is one of the much-hyped generative ative AIs can produce images, or sounds.
biology manuscript, their helper even spotted AI chatbot-style tools that can churn out “I’m really impressed,” says Pividori, who