GoKawiilWhat researchers think about the validity, and risk, of AI doomsday scenarios. Plus, a social network for agentic AI bots and how personalized CRISPR therapies could reach more people who need them.
Hello Nature readers, would you like to get this Briefing in your inbox free every day? Sign up here.
Researchers caught an African leopard on camera eating bats from a cave in Uganda. It might be the first confirmation that leopards eat live bats.Credit: Bosco Atukwatse/VSPT Kyambura Lion Project
Researchers who set up camera traps in a Ugandan national park have captured a host of animals feasting on Egyptian fruit bats (Rousettus aegyptiacus) on film. The footage offers real-time insight into how viruses might spread: the bats are known carriers of the Marburg virus, which can cause a fatal haemorrhagic fever in people. Scientists know that viruses can be passed from bats to intermediate animals and then on to humans, but this is the first time that potential intermediates have been caught on camera, the team says.
Nature | 5 min read
Reference: Current Biology paper
The Reddit-style social network Agent4Science is offering artificial intelligence agents a space to share, debate and discuss research papers — without human interruption. The site hosts different subgroups for agents to chew over AI-generated research papers on topics such as AI safety, prompts and deep learning. Human researchers can observe the chatter, but can’t contribute. The site is an experiment to have AI agents “freely discuss science and see where that will lead us”, says AI researcher Chenhao Tan, co-creator of the site.
Nature | 5 min read
Children in Tanzania that lived in specially designed houses were less likely to get malaria, diarrhoea and respiratory infections than were those living in conventional homes during a 3-year trial. The two-storey ‘Star Home’ puts bedrooms on the upper floor, where mosquitoes are less likely to venture, and the latrine outside to help reduce the spread of diarrhoeal diseases. Star Homes cost around US$8,800 to build, which is prohibitively expensive for many of the people it might help the most. But the design is a proof of concept to demonstrate that research can inform architecture that improves health outcomes, says health researcher and study co-author Lorenz von Seidlein.
Science | 6 min read
... continue reading