Proceedings of the National Academy of Sciences of the United States of America
PulseAugur coverage of Proceedings of the National Academy of Sciences of the United States of America — every cluster mentioning Proceedings of the National Academy of Sciences of the United States of America across labs, papers, and developer communities, ranked by signal.
8 day(s) with sentiment data
PNAS publishing research on AI vulnerabilities and limitations
The Proceedings of the National Academy of Sciences (PNAS) has recently published two distinct studies highlighting limitations and potential vulnerabilities in AI systems. One study focuses on the readability and sourcing of AI-generated articles, while another details how human persuasion tactics can trick AI models into objectionable requests. This suggests PNAS is becoming a venue for critical research into AI's current state.
AI-generated content quality may lag behind human-edited sources
Given the PNAS study on Grokipedia, it's plausible that AI-generated encyclopedic content will continue to be characterized by lower readability and fewer citations compared to human-edited sources like Wikipedia in the near future. Further research may be needed to track the evolution of AI content generation quality.
AI models may require new defenses against social engineering tactics
The PNAS finding that human persuasion tactics can trick AI models indicates a potential avenue for misuse. It is likely that developers will need to implement new security measures or fine-tuning techniques to specifically counter these 'parahuman' compliance vulnerabilities in future AI deployments.
-
Humans can be trained to spot AI-generated faces
Researchers at the Australian National University have developed a training method to help humans better identify AI-generated faces. The study, published in PNAS, highlights that while AI-generated faces are increasing…
-
Neurobiologist Chih-Ying Su moves from US to China for research role
Neurobiologist Chih-Ying Su has relocated from the US to China, joining the SMART research institute in Shenzhen. Su, a former faculty vice dean, cited the institute's advanced hardware and academic atmosphere, as well …
-
Saber-toothed cat Smilodon had surprisingly weak bite, study finds
New biomechanical modeling of the saber-toothed cat, Smilodon fatalis, reveals its bite force was significantly weaker than previously assumed, comparable to a large house cat rather than a lion. Researchers found that …
-
Evolutionary biologist explains why evolution appears static despite constant change
Evolution, the continuous process of change in species, appears static to humans due to its operation on geological timescales, far exceeding human lifespans. While bacteria visibly evolve rapidly due to their short gen…
-
The philtrum: An ancient evolutionary scar with repurposed functions
The philtrum, the groove above the human lip, is a vestigial structure resulting from the fusion of facial tissues during embryonic development. This developmental process is ancient and conserved across vertebrates, wi…
-
Psychologist reveals two key habits for long-term wealth accumulation
Accumulating wealth over a lifetime is less about knowing financial mechanics and more about psychological orientations toward time and decision-making. Individuals who build wealth tend to default to preferring future …
-
US-China AI hotline proposed to foster collaboration amid mistrust
A proposal for a US-China AI hotline aims to foster collaboration and mitigate geopolitical risks, drawing parallels to historical Cold War crisis communication. The author suggests that a controlled environment for joi…
-
China study reveals common threat to subsea cables
A new study, led by Tsinghua University and published in the Proceedings of the National Academy of Sciences, has developed a framework to better predict and manage self-accelerating turbidity currents. These powerful u…
-
LLMs map scientific literature, revealing hidden cross-topic connections
Researchers have developed a new framework using large language models (LLMs) to map scientific literature and identify cross-topic connections. This method was tested on a corpus of engineering articles from the Procee…
-
Psychologist outlines 5 habits for lasting happiness
Psychological science indicates that lasting happiness is not merely a fleeting mood but an outcome of structured life choices and environment. Key habits for well-being include investing deeply in relationships, as hig…
-
NOVA framework discovers interpretable driver behavior models
Researchers have developed NOVA, a symbolic regression framework designed to uncover interpretable models of driver behavior from trajectory data. Applied to millions of driving observations, NOVA identified a robust tw…
-
Psychologist links open-minded thinking to higher intelligence
Research indicates that actively open-minded thinking, the tendency to seek challenging evidence and revise beliefs, is a strong predictor of higher intelligence. This habit compounds existing cognitive abilities by enc…
-
Immortal jellyfish reverses aging by resetting its life cycle
The immortal jellyfish, Turritopsis dohrnii, is the only known animal capable of reversing its biological aging process. When faced with stress, this small hydrozoan can revert from its adult medusa stage to a polyp sta…
-
Baby Boomers' economic impact shifts as retirement creates labor and housing challenges
Baby Boomers, a large demographic cohort, have significantly influenced the U.S. economy for decades by distorting labor markets and housing prices. Their prolonged presence in the workforce suppressed wages for younger…
-
AI-generated Grokipedia articles are longer, less readable than Wikipedia
A recent study indicates that AI-generated articles on Grokipedia are less readable and cite fewer sources than human-edited Wikipedia articles. The research, published in the Proceedings of the National Academy of Scie…
-
Human persuasion tactics trick AI models into objectionable requests
A new paper published in PNAS reveals that traditional human persuasion tactics can influence AI models, a phenomenon termed "parahuman" compliance. Researchers found that techniques like flattery and appeals to authori…
-
MatSource Chairman Li Hao Featured in Science for AI in Materials
Professor Li Hao, chairman of MatSource, was featured in a Science magazine report highlighting his work in AI for Science. The report details three key projects from Li's team that integrate AI with material science, f…
-
Study: Solar panels on corn ethanol land could boost US energy output
A new study published in PNAS suggests transitioning corn-for-ethanol farmland to solar energy production could significantly boost the US's energy output while reducing ecological pressures. Researchers found that conv…
-
Google tests LLMs as expert research partners in physics
Google Research has published a study in the Proceedings of the National Academy of Sciences evaluating the effectiveness of Large Language Models (LLMs) in answering expert-level questions within specialized scientific…