Articles tagged with Artificial Intelligence
A new technique identifies and removes specific points in a training dataset that contribute most to a model's failures on minority subgroups. This approach maintains the overall accuracy of the model while improving its performance regarding underrepresented groups.
Researchers at Drexel University developed an AI tool using large language models to identify and suggest alternative words that stigmatize people with substance use disorder. This framework aims to educate users and preserve supportive spaces in online forums.
The study found that integrating AI into sales teams requires sales managers to combine traditional interpersonal communication competence with new AI-related skills. Effective communication involves motivation, knowledge, empathy, and adaptability, as well as crafting effective prompts to AI and maintaining a polite tone.
A groundbreaking AI model called NitroFusion creates images in seconds using modest hardware, eliminating the need for large computing resources. The open-source technology enables creative professionals and individuals to produce high-quality images affordably.
Researchers have developed an AI framework called XLuminA that autonomously discovers new experimental designs in microscopy. The framework performs optimizations 10,000 times faster than well-established methods, opening the path for exploring completely new territories in microscopy.
A new system, EXPLINGO, enables AI models to generate readable narratives explaining their predictions, helping users make better decisions. The system, developed by MIT researchers, uses large language models to transform complex explanations into plain language and automatically evaluate their quality.
Researchers trained an AI system to predict future warming based on climate model simulations and historical temperature data. The study suggests that even in the best-case scenario, global warming will likely exceed 2 degrees Celsius, with severe impacts expected in regions such as South Asia and Central Europe.
Researchers from Bar-Ilan University discover that classifying objects together through Multi-Label Classification can yield better results than detecting individual objects. This new method allows networks to learn correlations between object combinations, making them more recognizable in real-life applications such as autonomous vehi...
A study of 9,000 US-based participants reveals that they prioritize efficiency over fairness when choosing between human and algorithmic decision-makers. Notably, Republicans prefer humans more than Democrats, yet most claim fairness is a top priority.
A new shape-changing device called Shape has been developed to help people with visual impairment perform a location task as effectively as sighted people. The study found that participants with visual impairment located targets significantly faster using Shape than with vibration technology, and preferred it over current methods.
A recent study by ESMT Berlin scholars presents a comprehensive framework detailing the impact of AI on open innovation. The framework highlights three key ways in which AI is transforming open innovation practices: enhancing existing methods through efficiency and scalability, enabling new forms of collaboration and business models, a...
Cell culture medium optimization is crucial for effective growth of cells, as the medium determines the culture results. Researchers developed a machine learning model that optimizes the composition of the medium based on training data, achieving higher cell concentrations than commercially available media.
Scripps Research scientists created an AI model called MovieNet that processes videos by mirroring brain processes, recognizing visual scenes and distinguishing subtle changes. This breakthrough has the potential to transform fields like medical diagnostics, autonomous driving and more.
The new AI framework, DIMON, solves ubiquitous math problems known as partial differential equations in nearly all scientific and engineering research. It enables faster prediction of solutions for complex engineering designs, such as crash testing, orthopedics research, and bridge stress resistance.
Two studies found that readers perceive a lower level of credibility in news produced with AI assistance, regardless of whether they understand the extent of AI's involvement. Readers attribute human contributions more value than AI-driven content.
Researchers at Tokyo University of Science have developed a new method called black-box forgetting, which enables selective removal of unnecessary information from large pre-trained AI models. This approach enhances model efficiency and improves privacy by reducing computational resources and information leakage.
Professor Gerhard Neumann at KIT has received a EUR 2.4 million ERC Consolidator Grant to improve robots' ability to perform complex tasks in real environments. His SMARTI³ project aims to develop intuitive interfaces and scalable capabilities for complex manipulation tasks.
A recent study by the Society for Risk Analysis found that employees who visit adult or gambling sites are more likely to encounter seven types of malware, including coinminers and ransomware. Visiting these types of websites increases the risk of infection by two-fold.
Researchers created a measurement scale to assess robot human likeness, revealing four key qualities: appearance, emotional capacity, social intelligence, and self-understanding. To seem lifelike, robots must exhibit these traits, with self-understanding being the most challenging aspect to simulate.
Researchers developed a deep-learning model to predict organoid development at an early stage, outperforming human experts in accuracy and speed. The model classifies bright-field images of organoids into three quality categories, indicating their potential for regenerative medicine applications.
The University of Tennessee Institute of Agriculture has won a four-year grant to create hands-on curriculum about AI-related technologies for future farmers and leaders. Selected students will test the curriculum in drones, robotics, and other smart agriculture technologies, gaining skills in coding, drone-work, and robotics.
JMIR Perioperative Medicine is launching a new theme issue exploring burnout among perioperative clinicians, including anesthesiologists and advanced practice providers. The call for papers aims to examine the critical issue of burnout within perioperative medicine.
A new study improves breast cancer risk prediction by analyzing up to three years of previous mammograms, identifying individuals at high risk 2.3 times more accurately than the standard method. The AI method considers subtle changes in mammogram images and holds up well across diverse settings.
A self-pay, AI-enhanced breast cancer screening program was offered to patients across 10 clinical practices, resulting in a 43% higher cancer detection rate among enrolled women. The AI-powered program attributed 21% of the increase in cancer detection, with higher-risk patients choosing to enroll more frequently.
A team of UNC-Chapel Hill researchers will develop an oncology health learning system using AI-ready data to generate real-time identification of new cancer cases and support personalized treatments. The project aims to improve cancer diagnosis, care equity, and quality by creating accessible tools for health providers.
Researchers are developing a system to instantly conceal and anonymize voices through computer-generated speech. The technology will allow users to control factors like age, gender, and dialect in real-time conversations.
A new study found that AI-fact checking can lead to increased belief in false headlines, while decreasing faith in true ones. Human-generated fact checks, however, enhanced discernment of true headlines.
Researchers are using AI to find undocumented orphaned wells across the US, which can leak toxic substances into water sources and contribute to climate change. By analyzing historical topographic maps and modern sensors, they aim to locate and assess these forgotten relics.
The LiITA project creates an interoperable Knowledge Base for Italian language resources, enabling dialogue between resources and artificial intelligence model development. The project aims to foster innovation in linguistics and counteract declining interest in humanities studies.
The DIME project has achieved a critical breakthrough in overcoming technical obstacles to widespread adoption of edge AI. It has developed strategies to optimize performance without compromising decision quality and an innovative online learning algorithm to verify AI model accuracy in real time.
Researchers develop a simple fix to an existing technique, enabling the generation of sharp, high-quality 3D shapes that rival top model-generated 2D images. The new approach improves upon previous methods by avoiding costly retraining and complex postprocessing.
Researchers at the University of Bonn are developing an AI foundation model to improve radiology and detect Parkinson's disease earlier. The 'Re-Start PD' project aims to identify the condition before symptoms appear, with potential applications in diagnosis, treatment, and prevention.
Researchers developed a platform to help AI learn complex tasks through nuance and real-time instruction, achieving up to a 30% increase in success rates. The GUIDE framework allows humans to provide ongoing, nuanced feedback, fostering incremental improvements and deeper understanding.
The book examines AI's current advances, hurdles, and potential, emphasizing the need for science to maintain core norms and values. Experts advocate for human accountability and responsibility when using AI in research, highlighting the importance of transparent disclosure and attribution.
A study of 1,600 patient records found that large language model-generated notes were superior to physician-written summaries in automated evaluation methods. However, they were marginally inferior in usefulness and safety using a novel evaluation framework.
A $13 million initiative led by PNNL has created an AI-ready searchable database of 28,212 documents, reducing the time to conduct environmental reviews from 2-4 years to minutes. The PolicyAI toolkit is currently beta testing with federal NEPA reviewers, aiming to deliver time and cost savings.
The MINE-THE-GAP project aims to bridge the gap in mining's environmental and social impacts by providing reliable mining indicators using advanced satellite data and artificial intelligence. The project will provide mine-level indicators, including on land use, waste generation, and extracted minerals.
A new AI model, Spherical DYffusion, can predict climate patterns over 100 years in just 25 hours, making it 25 times faster than the state of the art. The model uses generative AI and physics data to achieve this breakthrough, reducing the need for supercomputers.
Scientists at MIT developed a fully integrated photonic processor that can perform all key computations of a deep neural network optically on the chip. The device completed machine-learning classification tasks in under half a nanosecond while achieving over 92% accuracy, similar to traditional hardware.
Large language models like GPT-4 can analyze gene sets with high accuracy, identifying common functions and providing detailed narratives to support their findings. This breakthrough has the potential to revolutionize functional genomics research by synthesizing complex information and generating new testable hypotheses.
Experts from UTHealth Houston and Baylor College of Medicine developed a pragmatic approach to monitor and manage AI systems in healthcare organizations. The guidance emphasizes the need for robust governance systems, testing processes, and transparency with patients to ensure safe AI adoption.
This editorial introduces persistence landscapes as a mathematical method to identify and correct biases in medical imaging. Persistence landscapes offer a way to reduce random noise while preserving important details, making it easier for clinicians to focus on meaningful image parts.
Researchers developed an AI tool called BrainBench to test large language models' ability to predict neuroscience study outcomes. The results showed that LLMs surpassed human experts with an average accuracy of 81%, highlighting their potential as powerful tools for accelerating research.
The Birmingham Blade is the world's first geographically tailored urban wind turbine designed by AI, addressing the challenge of harnessing clean wind energy in low-wind areas. Developed by EvoPhase and Kwik Fab, it boasts an optimal design for curved blades that spin around a central point, significantly enhancing efficiency.
Researchers introduce a new tool called 'persistence images' that uncovers hidden biases in medical imaging data, advancing fairness in healthcare AI. The technology helps detect bias, filter out noise, and improve overall accuracy, leading to more reliable diagnoses and better patient outcomes.
A novel physical reservoir computing device uses a dye-sensitized solar cell to mimic human synaptic elements, enabling efficient time-series data processing and low power consumption. The device achieved high computational performance in tasks such as human motion classification with over 90% accuracy.
The HumanTech Summit 2024 conference brings together experts in technological and social innovation, discussing key topics such as data security and interpersonal relationships. With a proven legacy of impact, the event has attracted over 1,400 participants and featured renowned keynote speakers from top universities.
Researchers in Lithuania have developed Forest 4.0, a smart forest data processing model that enables real-time monitoring of forest conditions, sustainable resource accounting, and transparent governance. The system uses IoT sensors and AI algorithms to detect tree diseases, illegal logging, and predict changes in forest ecosystems.
Researchers developed an artificial intelligence model that selects the best phage cocktail for a given patient based on their genome. The model was tested on a new collection of E. coli strains responsible for pneumonia and showed high success rates.
A new AI tool generates realistic satellite images of future flooding, which can help communities visualize and prepare for approaching storms. The method combines a generative artificial intelligence model with a physics-based flood model, producing more accurate and realistic images than an AI-only approach.
The National Center for Supercomputing Applications (NCSA) has received the Readers' Choice Award: Best HPC Collaboration and Editors' Choice: Best Use of HPC in Physical Sciences. This is the 14th consecutive year NCSA has been honored with an HPCwire award.
Researchers at MIT have introduced a new algorithm that strategically selects the best tasks for training an AI agent, resulting in improved performance and reduced training costs. The technique outperforms existing methods by five to 50 times, making it more efficient and effective.
Researchers at Mass General Brigham developed an AI tool called EVOLVEpro that can engineer proteins to be more stable, precise, and efficient. The tool has shown promise in improving medications used for treating autoimmune diseases, genetic diseases, and cancer.
Large language models show promise in identifying housing instability from clinical notes, but AI interpretation can be incorrect and human review is still necessary. De-identification methods also require refinement to preserve privacy without losing important details.
Researchers have developed a new method for designing large artificial proteins with high accuracy, utilizing AI-based software Alphafold2. The approach combines accurate structure prediction with optimization techniques, allowing for the creation of proteins with tailored properties, such as precise binding and stability.
A study found that police radio transmissions disproportionately involve Black suspects, posing a higher privacy risk for Black people. The researchers analyzed 24 hours of human-transcribed communications and found that Black males were mentioned most often, even in majority white zones.
Insilico Medicine has received FDA clearance for ISM5939, a potential best-in-class oral small molecule inhibitor targeting ENPP1 for the treatment of solid tumors. ISM5939 demonstrated robust anti-tumor efficacy in preclinical studies and showed favorable safety profiles.
Researchers discovered that seven out of eight AI models generated significant amounts of biased text in interactions, particularly when discussing caste. The study's findings suggest a pressing need for policy regulations to ensure these models are safe for everyone, especially in sensitive tasks like hiring.
A global team, including Lehigh University researcher Xuanhong Cheng, is exploring molecular- and cellular-level changes in muscle tissue that could lead to better diagnostic tools and therapeutic options for CFS and long COVID. The team aims to develop noninvasive diagnostic tools using electrical signatures.
Researchers found that none of the four AI image generators accurately represented gender, racial, or disability diversity among real chemists. The study highlights concerns about the accuracy of AI-generated content and its potential to perpetuate stereotypes.