The textbook version of the “Out of Africa” hypothesis holds that the first human species to leave the continent around 1.8 million years ago was Homo erectus. But in recent years, a debate has emerged suggesting it wasn’t a single species, but several. New research published in the journal PLOS One now hopes to settle the matter once and for all.
A research team at the Facility for Rare Isotope Beams (FRIB) is the first ever to observe a beta-delayed neutron emission from fluorine-25, a rare, unstable nuclide. Using the FRIB Decay Station Initiator (FDSi), the team found contradictions in prior experimental findings. The results led to a new line of inquiry into how particles in exotic, unstable isotopes remain bound under extreme conditions. Led by Robert Grzywacz, professor of physics at the University of Tennessee, Knoxville (UTK), the team included Jack Peltier, undergraduate student at UTK, Zhengyu Xu, postdoctoral researcher at UTK, Sean Liddick, professor of chemistry at FRIB and interim chairperson of MSU’s Department of Chemistry, and Rebeka Lubna, scientist at FRIB.
The team published its results in Physics Letters B.
“The different results on decay lifetime we obtained for fluorine-25 were similar to previously measured decay of oxygen-24. And while we are not entirely certain why we found this difference between previously published results, we have conducted numerous checks on our results and are confident in our findings,” Grzywacz said.
The simulation hypothesis—the idea that our universe might be an artificial construct running on some advanced alien computer—has long captured the public imagination. Yet most arguments about it rest on intuition rather than clear definitions, and few attempts have been made to formally spell out what “simulation” even means.
A new paper by SFI Professor David Wolpert aims to change that. In Journal of Physics: Complexity, Wolpert introduces the first mathematically precise framework for what it would mean for one universe to simulate another—and shows that several longstanding claims about simulations break down once the concept is defined rigorously.
His results point to a far stranger landscape than previous arguments suggest, including the possibility that a universe capable of simulating another could itself be perfectly reproduced inside that very simulation.
AI has successfully been applied in many areas of science, advancing technologies like weather prediction and protein folding. However, there have been limitations for the world of scientific discovery involving more curiosity-driven research. But that may soon change, thanks to Kolmogorov-Arnold networks (KANs).
A recent study, published in the journal Physical Review X, details how this new kind of neural network architecture might help scientists discover and understand the physical world in a way that other AI can’t.
Despite its tropical climate and floodplain location, Bangladesh—one of the world’s most densely populated nations—seasonally does not have enough freshwater, especially in coastal areas. Shallow groundwater is often saline, a problem that may be exacerbated by rising sea levels.
Rainfall is highly seasonal and stored rainwater often runs out by the end of the dry season. And contamination by naturally occurring arsenic deposits and other pollutants farther inland further depletes supplies of potable water, which can run desperately short during annual dry seasons. According to the UN’s Sustainable Development Goals, 41% of Bangladeshis do not have consistent access to safe water.
Hoping to ease the crisis, researchers from Lamont-Doherty Earth Observatory, which is part of the Columbia Climate School, led an exploration for new freshwater sources along the Pusur River in the Ganges-Brahmaputra Delta. They recently published their results in the journal Nature Communications.
Scientists in China have unveiled a new AI chip called LightGen that is 100 times faster and 100 times more energy efficient than NVIDIA chips, the leading supplier of AI chips worldwide. Instead of using electricity to move information, this new optical chip relies on light to perform complex generative tasks.
Traditional general AI models, such as ChatGPT and Stable Diffusion, run on everyday silicon chips and require massive amounts of computing power and electricity, which can generate significant heat. For particularly complex tasks, these chips can struggle with the workload, resulting in slow processing times.
“Feral” gossip spread via AI bots is likely to become more frequent and pervasive, causing reputational damage and shame, humiliation, anxiety, and distress, researchers have warned.
Chatbots like ChatGPT, Claude, and Gemini don’t just make things up—they generate and spread gossip, complete with negative evaluations and juicy rumors that can cause real-world harm, according to new analysis by philosophers Joel Krueger and Lucy Osler from the University of Exeter.
The research is published in the journal Ethics and Information Technology.
Whether you’re a scientist brainstorming research ideas or a CEO hoping to automate a task in human resources or finance, you’ll find that artificial intelligence (AI) tools are becoming the assistants you didn’t know you needed. In particular, many professionals are tapping into the talents of semi-autonomous software systems called AI agents, which can call on AI at specific points to solve problems and complete tasks.
AI agents are particularly effective when they use large language models (LLMs) because those systems are powerful, efficient, and adaptable. One way to program such technology is by describing in code what you want your system to do (the “workflow”), including when it should use an LLM. If you were a software company trying to revamp your old codebase to use a more modern programming language for better optimizations and safety, you might build a system that uses an LLM to translate the codebase one file at a time, testing each file as you go.
But what happens when LLMs make mistakes? You’ll want the agent to backtrack to make another attempt, incorporating lessons it learned from previous mistakes.
For people, matching what they see on the ground to a map is second nature. For computers, it has been a major challenge. A Cornell research team has introduced a new method that helps machines make these connections—an advance that could improve robotics, navigation systems, and 3D modeling.
The work, presented at the 2025 Conference on Neural Information Processing Systems and published on the arXiv preprint server, tackles a major weakness in today’s computer vision tools. Current systems perform well when comparing similar images, but they falter when the views differ dramatically, such as linking a street-level photo to a simple map or architectural drawing.
The new approach teaches machines to find pixel-level matches between a photo and a floor plan, even when the two look completely different. Kuan Wei Huang, a doctoral student in computer science, is the first author; the co-authors are Noah Snavely, a professor at Cornell Tech; Bharath Hariharan, an associate professor at the Cornell Ann S. Bowers College of Computing and Information Science; and undergraduate Brandon Li, a computer science student.
Aviation’s climate impact is partly due to contrails—condensation that a plane streaks across the sky when it flies through icy and humid layers of the atmosphere. Contrails trap heat that radiates from the planet’s surface, and while the magnitude of this impact is uncertain, several studies suggest contrails may be responsible for about half of aviation’s climate impact.
Pilots could conceivably reduce their planes’ climate impact by avoiding contrail-prone regions, similarly to making altitude adjustments to avoid turbulence. But to do so requires knowing where in the sky contrails are likely to form.
To make these predictions, scientists are studying images of contrails that have formed in the past. Images taken by geostationary satellites are one of the main tools scientists use to develop contrail identification and avoidance systems.