Shulou(Shulou.com)11/24 Report--

CTOnews.com, January 14 (Xinhua)-- Scientific American (Scientific American), an American popular science magazine, published an article today, saying that scientific papers written by ChatGPT are so highly academic that senior scientists can't tell them apart.

Image source Nature

Artificial intelligence (AI) chatbots are already capable of writing convincing fake research papers that scientists can't tell without careful identification, according to a preprinted paper published on the bioRxiv server in late December.

"I am very worried about this," said Sandra Wachter, who studies technology and regulation at the University of Oxford in the UK. "if even experts can't tell whether the paper is true or false, it will affect the cornerstone of our research."

CTOnews.com small class: chat robot ChatGPT will create realistic and intelligent-sounding text based on user prompts. It is a "large language model", a neural network-based system that learns to perform tasks by digesting a large number of existing human-generated texts. OpenAI, a software company based in San Francisco, California, released the tool on November 30 and is available for free.

A team led by Catherine Gao of Northwestern University in Chicago, Illinois, used ChatGPT to generate abstracts of artificial research papers to test whether scientists could find them. The researchers asked chatbots to write 50 medical research abstracts based on selections published in the Journal of the American Medical Association (JAMA), the New England Journal of Medicine (The New England Journal of Medicine), the British Medical Journal (The BMJ), the Lancet (The Lancet) and Natural Medicine (Nature Medicine).

They then compared these abstracts with the original abstracts using plagiarism detectors and AI output detectors, and asked a team of medical researchers to find fabricated abstracts.

The summary generated by ChatGPT successfully passed the plagiarism checker (plagiarism checker): the median score for originality was 100%, indicating that no plagiarism was detected.

The AI output detector found 66% of the generated summaries, while manual audits correctly identified only 68% of the generated summaries and 86% of the real summaries. Experts mistakenly identified 32% of the generated summaries as real summaries and 14% of the real summaries as generated.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.