Articles tagged with Artificial Intelligence
The AI model can accurately identify disease-causing mutations in human genes and is capable of designing new genomes that are as long as the genomes of simple bacteria. Evo 2 has been trained on over 100,000 species across the entire tree of life and can process genetic sequences of up to 1 million nucleotides at once.
A European consortium is developing AI-powered therapy assistants to help doctors adhere to medical guidelines. The GUIDE-AI project aims to reduce the complexity of guideline navigation, particularly for chronic diseases such as heart failure and asthma.
A new study explores the market effects of unleashing generative AI on creative endeavors, finding that low-quality 'AI slop' harms consumers and professionals. High-quality AI, however, can enhance professional work while providing consumers with better content options.
A recent study found that AI programs are up to 97% accurate at detecting pictures of deepfake faces, but performed at chance levels when it comes to detecting deepfake videos. Humans correctly identified real and fake videos about two-thirds of the time, picking up on subtle inconsistencies in movement and facial expressions. The stud...
A study at George Mason University found that nursing students in an AI-enhanced health informatics course showed statistically significant learning gains compared to traditional students. The course used AI-generated videos and clinically contextualized assignments to improve student engagement and outcomes.
Research participants reading AI-generated summaries of historical events showed more liberal opinions than those reading Wikipedia summaries, indicating a subtle influence of AI biases on public understanding of the past. The study highlights the need for LLMs to be designed with bias-reducing features to provide unbiased information.
Researchers at Weill Cornell Medicine will develop an AI system to identify patients with early signs of treatment-resistant prostate cancer. The project aims to improve clinical trials by predicting treatment outcomes and avoiding ineffective treatments for patients with specific tumor subtypes.
A new study proposes a framework for connecting AI risk prediction systems with scheduling platforms to detect emerging risks and adjust project plans before delays occur. This approach aims to close the gap between early warning and actionable response, enabling more resilient project delivery.
Researchers from the University of Exeter argue that AI models are not as smart as claimed, citing a potential 'transferability crisis'. They found that generalisability is difficult to predict and benchmark metrics often overstate model performance. The study highlights the need for rapid testing in real-world applications.
Researchers developed a digital safeguard, My Music My Choice, to protect songs from generative AI cloning. The tool adds imperceptible changes to a song's waveform, making it difficult for AI models to replicate.
A new study by Cleveland Clinic and Dyania Health demonstrates the potential of large language models in accurately identifying patients who could benefit from clinical research. The AI system reviewed 1,476 patients and identified 46 as potential matches, resulting in a more diverse patient population.
The study reveals how AI is fundamentally shifting the paradigm for discovering high-performance batteries and efficient electrocatalysts. AI-driven generative models and Large Language Models are enabling scientists to predict exact chemical structures and navigate vast chemical spaces at unprecedented speeds.
New research warns that popular AI systems for predicting cancer biology may rely on visual shortcuts, raising concerns about their reliability for real-world patient care. Despite achieving high accuracy, these models often fail when conditions change or confounding factors are controlled.
Researchers at the University of Bath developed an AI tool that predicts building emissions from simple text descriptions, providing architects with real-time feedback on sustainability. The tool uses machine learning and natural language processing to estimate embodied carbon and suggests improvements to environmental conditions.
The journal argues that AI can process vast datasets to detect patterns and predict climate impacts, improving pollution tracking, climate modeling, and agricultural planning. AI-driven modeling can also help policymakers evaluate tradeoffs and anticipate unintended consequences, supporting more informed decision-making.
Recent advances in machine learning enable the non-targeted analysis of thousands of chemical features in a single environmental sample. Machine learning models can predict tandem mass spectra from known molecular structures and infer molecular formulas, significantly narrowing down candidate structures.
The study identifies three diverging global AI pathways, each shaped by distinct policy priorities, innovation models, and governance philosophies. The US prioritizes innovation speed, China emphasizes deployment scale, and the EU focuses on governance and societal safeguards.
This study reveals that a femtosecond laser can induce a rise in electronic temperature, transiently blocking optical absorption and enabling multicolor modulation from a single material platform. The discovery opens a new pathway toward ultrafast, broadband, and energy-efficient photonic devices.
AI research at Georgia Tech suggests that simple safety measures are not enough to address the technology's moral implications. The study proposes a middle ground between safe and autonomous AI by advocating for 'end-constrained ethical AI.' This approach prioritizes fairness, honesty, and transparency in AI decision-making.
Researchers argue that AI's limitations and alignment with human values are key to preventing potential threats. They suggest sector-specific policies rather than universal regulations.
A research team from HKUST developed GrainBot, an AI-enabled toolkit that automates the extraction and quantification of multiple microstructural features from microscopy images. This provides a systematic method for converting complex image information into quantitative data, accelerating materials discovery and development.
Researchers used machine learning techniques to compress a large model of the visual cortex, creating smaller versions that predict neural responses with high accuracy. The compact models revealed specific computational patterns in how neurons detect important features, offering insights into how visual information is processed.
The International AI Art Biennale brings together artists, experts, and researchers to explore the role of AI in culture, art, and science. The event features exhibitions, performances, and a scientific conference that focuses on creative dialogue and education with AI.
A new study finds that fewer than 250 fossils are required to train an image-based AI algorithm, a significant improvement on previously thought numbers. The discovery could greatly speed up the identification process in vertebrate paleontology, where most fossils are fragmented and difficult to analyze.
A global consortium created an exam with 2,500 questions spanning multiple subjects to assess AI capabilities. Current AI models consistently fail the exam, highlighting gaps in their understanding. The project aims to provide a long-term benchmark for evaluating advanced AI systems and demonstrate the importance of human expertise
A new study by Virginia Tech and University of Vermont researchers identifies the design features that influence whether trusted agricultural advisors choose AI tools. The research suggests that cost, data ownership, and human-centered design are pivotal to selection.
A multicenter Japanese study developed and validated a tool to measure attitudes toward artificial intelligence among medical students and residents. The J-ATTARI-12 scale assesses AI anxiety and aversion versus AI optimism and acceptance, with moderate positive correlations with related constructs.
Researchers at Linköping University have developed an AI model that can accurately determine the time of death from blood samples, providing crucial information in murder investigations and police work. The method uses metabolites to analyze changes in small molecules over time, outperforming current methods.
Researchers found that shrinking ferroelectric tunnel junctions significantly boosts their performance, producing larger resistance contrasts between 'ON' and 'OFF' states. This enables efficient and reliable memory technologies for emerging applications in AI, edge computing, and IoT.
Researchers warn that AI's friction removal could eliminate desirable difficulties, weakening learning and motivation. Effortful engagement strengthens learning and memory, while loneliness serves as a signal to form human connections.
A recent study reveals that creators who disclose using AI face negative judgments from others, despite their reputation. This gap between technological capability and social perception highlights the need for creators to manage perceptions surrounding how their work was created.
Researchers have developed a definitive guide to safely navigating health information on AI-powered chatbots, addressing the risk of 'hallucinated' or factually incorrect advice. The Health Chatbot Users’ Guide focuses on harm reduction and maximises benefits to users, offering a pragmatic and neutral approach.
CERCat research projects combine physics-based and AI-based models to predict extreme weather hazards, assess vulnerability, and predict regional climate impacts. The partnership enables the development of machine learning systems for real-time damage assessments and multi-scale hazard modeling.
Researchers at Linköping University have developed an AI-boosted electronic nose that can detect ovarian cancer from blood plasma samples with high accuracy. The method uses machine learning to identify patterns specific to the disease, making it a promising tool for early detection and improved survival rates.
Researchers found that farming practices, unsafe sanitation, and domestic animals linked to infection risk. Household-level interventions are essential for accelerating elimination, especially when community-wide strategies fail near elimination.
Researchers developed open-source software to measure AI model energy use, revealing trends in how design affects power requirements. The tool can automate the search for efficient parameters, helping lower energy costs and environmental impact.
Researchers at Memorial Sloan Kettering Cancer Center investigate ferroptosis as a potential target for cancer treatment. They also develop an AI model to analyze country-specific factors driving global cancer outcomes. Additionally, the institution introduces an AI-based incident analysis approach to improve patient safety.
Researchers at UTIA and UT Knoxville create an automated sensor network to monitor compost piles, reducing labor costs and improving regulatory compliance. The new system uses battery-free sensors and machine learning algorithms to analyze temperature and moisture variations, enabling data-driven decision making.
Collaborative teams of AI-powered robots successfully navigated and extinguished simulated and hybrid simulation-physical fires, demonstrating a 99.67% success rate in real-world deployment potential.
Epic Games has acquired Meshcapade, a Max Planck startup that develops solutions for creating and animating digital humans. The technology, based on the SMPL body model, enables realistic human movement and expression in 3D.
Researchers developed a new method called Learn-to-Steer, which analyzes internal attention patterns of image-generation models to guide their placement according to user instructions. The approach improved accuracy in understanding spatial relationships by up to 61% in existing trained models.
Researchers at the University of East London identified a critical weakness in Industrial Internet of Things networks that can destabilize factories and safety-critical infrastructure. Attackers can exploit subtle 'delay attacks' to quietly disrupt coordination, making timing just as important as information in highly automated systems.
A new method to steer AI output uncovers vulnerabilities and potential improvements. The researchers identified 512 concepts within five classes and improved LLM performance on narrow tasks.
Researchers developed MULTI-evolve, a framework for efficient protein evolution that applies machine learning models to predict beneficial mutations and their combinations. The approach identified function-enhancing mutations and tested their pairwise combinations, demonstrating improved efficiency in protein engineering.
The partnership aims to accelerate advanced nuclear reactor deployment and reduce costs through AI-driven design, licensing, and operation. By leveraging AI, Prometheus Grand Challenge seeks to deliver nuclear energy at a faster schedule, with at least 2x acceleration and greater than 50% operational cost reductions.
A team from MIT and UC San Diego has developed a new method to uncover hidden biases, moods, and abstract concepts in large language models (LLMs). The approach identifies these connections within the model and allows for manipulation of the concept in generated answers.
A hierarchical 3D LiDAR localization method improves robot positioning in large outdoor spaces even after seasonal changes. The method integrates deep learning techniques to extract discriminative local features from 3D point clouds, making it robust to environmental variability.
A Mount Sinai-led study developed an AI tool that analyzes ECGs to identify patients at risk for harmful heart changes, potentially reducing the need for expensive and time-consuming cardiac MRI scans. The tool could make lifelong heart monitoring more accessible and efficient for people born with congenital heart disease.
Researchers propose a new framework to detect chronic diseases earlier by tracking individual biological changes. Advances in technology and AI enable the analysis of complex patterns from wearable devices and biological samples.
Researchers argue that mandated reminders about a chatbot's non-human nature may exacerbate mental distress and form stronger attachments. The warnings come after recent deaths by suicide linked to chatbots like ChatGPT and Character.AI, highlighting the need for careful crafting and timing of these reminders.
Researchers developed a free-to-use software tool, PSBench, to verify the accuracy of artificial intelligence-based protein structure predictions. The database includes 1.4 million annotated protein models, verified by experts, and provides reliable information for building more accurate AI systems.
Researchers developed machine learning models that accurately forecast pollen counts for both grass and birch tree pollen, enabling early warning systems. This breakthrough could lead to more effective hayfever treatments by allowing individuals to take preventative measures before symptoms appear.
Researchers found that personalization features in LLMs can increase the likelihood of sycophancy, a phenomenon where models mirror users' viewpoints. The study suggests that context and user profiles play a significant role in shaping this behavior.
Researchers found that dosed nonlinearity improves model performance in various tasks, especially with limited data. Nonlinear units function like flexible switches, adapting linear processing modes based on context.
Researchers from Texas A&M University are using black soldier fly larvae to recycle organic waste and produce protein fertilizer. The project aims to create autonomous systems for rehabilitating extreme environments, reducing human exposure to hazardous conditions.
Researchers found Chinese chatbots more likely to refuse or provide inaccurate responses to sensitive questions about politics, while non-Chinese models offered longer, more accurate answers. The study suggests the source of this issue cannot be explained solely by training data or model development choices.
Researchers at UCSF and Wayne State University found that generative AI tools can perform orders of magnitude faster than human teams in analyzing health data. Junior researchers paired with AI generated viable prediction models in minutes, outperforming experienced programmers in hours or days.
MIT researchers used a large language model to optimize the genetic sequences of proteins manufactured by yeast, reducing production costs. The new model predicted which codons would work best for manufacturing six different proteins, including human growth hormone and a monoclonal antibody, with successful results.
Research reveals major gaps in studying the relationship between AI and journalism, including environmental impact and Global South deployment. Most studies are conducted from Europe, leaving scarce attention to the Middle East, Latin America, and sub-Saharan Africa.
Despite AI tool limitations, generative AI projects continue due to various justifications that make them hard to stop. The study highlights nine conceptions of the common good that drive innovation momentum. Effective boundary work and reframing failure as a learning process also sustain innovation.