In plain words: Scientists asked a chatbot to invent scientific hypotheses the way human researchers do. It made many mistakes, but it wove together large amounts of research into interesting, testable ideas that could guide future experiments.
Abstract
We investigate whether large language models can perform the creative hypothesis generation that human researchers regularly do. While the error rate is high, generative AI seems to be able to effectively structure vast amounts of scientific knowledge and provide interesting and testable hypotheses. The future scientific enterprise may include synergistic efforts with a swarm of "hypothesis machines", challenged by automated experimentation and adversarial peer reviews.
Yang Jeong Park, Daniel Kaplan, Zhichu Ren, Chia-Wei Hsu, Changhao Li, Haowei Xu, Sipei Li, Ju Li
arXiv:2304.12208 · cs.CL · submitted Mar 30, 2023
abstract · pdf
