GoKawiilNews
The preprint repository arXiv has announced a one-year posting ban for researchers whose submissions are found to contain references hallucinated by artificial intelligence. Even after this penalty period, affected researchers can’t post to arXiv unless their manuscript has already been accepted at a “reputable peer-reviewed venue”, according to computer scientist Thomas Dietterich, chair of arXiv’s computer science section. Some researchers have praised the server for taking a stand; others suggest it doesn’t go far enough to tackle ‘AI slop’ in preprints.
Nature | 6 min read
News
Two new systems use teams of AI agents to develop hypotheses, propose experiments and analyse data in a fraction of the time it would take humans alone. The approaches still rely on human input at various stages, but when asked to identify existing drugs that might be repurposed for different conditions, they arrived at plausible answers in a matter of hours. “The goal is to give scientists superpowers,” says Google DeepMind researcher Vivek Natarajan, who helped to develop one of the systems.
Nature | 5 min read
Reference: Nature paper 1 & paper 2
Opinion
“Much of the enthusiasm for AI tools … comes from their promise to offload work,” write anthropologist Lisa Messeri and psychologist M. J. Crockett. “But many ‘low-skilled’ tasks have conventionally been important starting points for trainee scientists.” Cleaning raw data reveals its flaws; reviewing the literature gives a holistic view; and crucial know‑how is transmitted through practical experience. Scientists should have field-wide conversations now about AI’s influence on the expertise of future colleagues, the authors argue.
Nature | 11 min read
... continue reading