Articles tagged with Generative Ai
Evo, a generative AI model, uses patterns in microbial genomes to write new genetic code, expanding the length of sequences models can process and improving resolution. Researchers use Evo to understand microbial and viral genomes, fashion new proteins, and reprogram microbes for remarkable tasks.
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.
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.
Researchers found that teenagers use generative artificial intelligence for emotional support, social interactions, and academic purposes. However, they also reported significant misconceptions about generative AI among parents and children. Parents had little to no understanding of their children's use of GAI, and they viewed AI as a ...
Most medical large language models offer little better than guesses when it comes to interpreting medical information. ChatGPT-4 achieved the best performance, but its accuracy remained insufficient for some medical code questions. The study highlights the need for broader clinical language in models and caution in their widespread use.
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.
Insilico Medicine has developed an innovative platform using generative AI to discover novel molecules. The company has nominated over 20 preclinical candidates and received IND clearance for 10 molecules.
Recent Nobel Prizes in physics and chemistry have recognized the convergence of AI with physics and chemistry, emphasizing the need for interdisciplinary research. Researchers advocate for nurturing AI-enabled polymaths to bridge the gap between theoretical advancements and practical applications.
Alex Zhavoronkov, Insilico Medicine's founder and CEO, was named Highly Cited Researchers of 2024. He has published over 40 peer-reviewed research papers, with a total H-index of 64, and led the development of Insilico's lead drug pipeline, ISM001-055.
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.
A study found that ChatGPT provides higher fatality numbers when asked in Arabic compared to questions in Hebrew, highlighting the impact of user language on information dissemination. The researchers believe this has profound social implications, as it can shape perceptions of conflict and fuel biases.
Insilico Medicine has been selected by Fortune as one of the top AI innovators, leveraging its generative AI platform to develop novel drugs and treatments. The company's lead drug pipeline, ISM001-055, has shown promising results in Phase IIa trials, offering new hope for patients with idiopathic pulmonary fibrosis.
A novel no-code prototype, Auto-DSM, generates design structure matrices (DSMs) using large language models. It improves productivity and accuracy compared to traditional methods, with speeds of up to four minutes per DSM generation.
Experts discuss the potential of AI to enhance healthcare, including its use in diagnosing diseases and educating patients. However, concerns about the accuracy of chatbot responses and the need for verified sources highlight the perils of relying on AI in medicine.
The GenAI4ED project aims to revolutionize secondary education by developing a digital platform that assesses and selects GenAI-based educational software. The platform will provide personalized recommendations based on predefined criteria, taking into account both technical effectiveness and educational impact.
New algorithm CALDERA enables leaner LLMs with improved performance, reducing costs, energy consumption and privacy risks. This allows organizations to adapt models to their specific needs without sharing sensitive data, enhancing user privacy.
Researchers developed BiomedParse, an AI medical image analysis model that works across nine types of medical images to predict systemic diseases. This enables non-specialists to analyze and identify subtle variations in medical images, leading to improved diagnosis.
Researchers found that membership inference attacks on large language models (LLMs) are not effective in measuring information exposure risks. The common method used to test LLM leaks suffers from ambiguity due to the fluidity of language, making it difficult to define a representative set of non-member candidates.
Pharma.AI Week will showcase the latest advancements in Insilico's generative AI platform, including PandaOmics and Science42, with expert speakers and hands-on demos. The event aims to empower researchers and scientists with tools for faster and more accurate discoveries.
The Healthcare AI Challenge enables healthcare professionals to evaluate and compare AI solutions for their effectiveness in various clinical situations. By crowdsourcing input from experts, the challenge aims to create continuous and reliable expert evaluations of AI technologies in medicine.
The Phase IIa trial of ISM001-055 showed dose-dependent improvement in forced vital capacity (FVC) and percent predicted FVC, suggesting potential to slow or reverse disease progression. The treatment also demonstrated a favorable pharmacokinetics profile with minimal adverse events.
Insilico Medicine CEO Alex Zhavoronkov to discuss AI business potential and economic growth at Fortune Global Forum 2024. The company has developed a generative AI-powered platform that utilizes deep learning techniques for novel target discovery and molecular structure generation.
Despite their ability to provide turn-by-turn driving directions in New York City with near-perfect accuracy, generative AI models do not form a coherent internal map of the city. The researchers developed new metrics to test a transformer's world model and found that only one generated a coherent world model for Othello moves.
Insilico Medicine has entered into a revolving loan facility of up to US$100 million with HSBC, enabling the company's global expansion and AI-driven innovation in biotechnology. The credit line will support Insilico's proprietary novel drug discovery pipeline and its end-to-end diversified AI platform, Pharma.AI.
A new generative AI tool called Theseus is being developed to reduce time and labor in computational modeling, suggesting and running simulations to accelerate research and discovery. The AI assistant aims to enhance the work of computational scientists by providing scientific intuition, checking code for bugs and making suggestions.
Researchers found significant bias in AI-powered hiring systems, with LLMs favoring white-associated names 85% of the time, female-associated names only 11% of the time, and never Black male-associated names. The study highlights the need for intersectional analysis to ensure fairness in AI systems.
A new study from Virginia Tech shows large language models can assess human-made environments using street-view images, similar to traditional methods. LLM-based performance offers a more accessible tool for users in small to medium-sized cities, making it easier to manage smart urban infrastructure.
The collaboration utilized Insilico's generative biology AI platform, Pharma.AI, to identify a novel lead for treating oncology diseases. The joint R&D team addressed the druggability challenge of highly novel targets using novel scaffolds generated by Chemistry42.
Researchers found that AI doctors who recall patients' social information, with privacy control offered, increase patient satisfaction. Human doctors don't require social or medical information to establish a close relationship with patients.
A new study by University of Technology Sydney researchers challenges the Infinite Monkey Theorem, finding it unlikely a monkey can type out Shakespeare's works before the universe ends. With around a 5% chance, even an increased number of monkeys would not produce the Bard's entire works.
A study of 50 U.S.-licensed physicians found that GPT-4 did not significantly improve clinical reasoning compared to conventional resources. The integration of GPT-4 as a diagnostic aid alongside clinicians showed promising results but required further exploration to understand its potential benefits.
Researchers at UC Berkeley found that humans create more diverse and nuanced stories than generative AI tools like ChatGPT. While AI shows improvement, it still struggles with complexity and clichés.
MIT researchers developed a versatile technique that combines diverse data from various sources into a shared language for generative AI models. This approach outperformed traditional techniques by 20% in simulation and real-world experiments.
A new study by Penn State researchers suggests that making AI training data diversity information available can shape users' expectations of algorithmic fairness and trust. Displaying racial diversity cues in AI interfaces can enhance users' perceptions of algorithmic fairness and trust, according to the study's findings.
A recent study reveals that when people are told a story was written by AI, they rate it poorly, suggesting a distrust of AI-generated art. Researchers believe this could hinder the use of AI in public health campaigns to encourage healthy behaviors.
Researchers at University of Zurich used AI to analyze antibiotic resistance using GPT-4 model, creating EUCAST-GPT-expert system for accurate interpretation of antimicrobial resistance mechanisms. The AI system performed well in detecting certain types of resistance but had limitations, while human experts were more accurate but slower.
A new study reveals that AI-driven chatbots may perpetuate racial and ethnic biases in pain assessment, leading to further inequalities in healthcare. Researchers found that Black patients were consistently underassessed for their pain compared to white patients, regardless of whether the rater was human or AI.
A study from Lehigh University and Seattle University found that making patients aware of biases in human healthcare decisions increases receptiveness to AI recommendations. By highlighting the limitations of human judgment, healthcare providers can create a more balanced relationship between patients and emerging technologies.
Researchers found that AI-powered chatbot answers on top 50 most prescribed drugs in US were difficult to read and lacked information, posing potential harm to patients. Only 54% of answers aligned with scientific consensus, highlighting the need for caution when relying on these tools.
A study from the University of Kansas found that parents trust AI-generated text, like ChatGPT, more than healthcare professionals' content. Parents rated ChatGPT as more trustworthy, accurate, and reliable in many cases, even when there were significant differences in ratings.
Researchers developed AI chatbot UroBot to answer complex urology questions with high accuracy, exceeding human urologists. The model justifies its answers based on European Society of Urology guidelines and has been tested on 200 specialist questions, achieving an accuracy rate of 88.4%.
The two-year study aims to explore biases in AI systems and develop a 'human-in-the-loop' framework for quality data discovery. It will investigate how humans can be involved as labelers, prompters, and validators to improve data sets and user interfaces.
A new study from UCSF has found that ChatGPT overprescribes in emergency care situations, providing unnecessary x-rays and antibiotics. The AI model was less accurate than resident physicians, with a tendency to err on the side of caution. Researchers emphasize the need for better frameworks to evaluate clinical information before AI c...
Recent studies have found that large AI language models are becoming increasingly unreliable, failing to match human expectations on task difficulty. The models excel in complex tasks but struggle with simple ones, leading to a mismatch between human perception and model performance.
The CityUHK team develops a new federated learning technology integrated with research knowledge graphs and large language models to digitally transform STI services. This will create a comprehensive, domain-specific knowledge base and provide intelligent assistant services for researchers.
Researchers found that AI-generated essays are most similar to those written by male students from higher socioeconomic backgrounds and private schools. The writing styles also tend to be less varied than human-written essays, with AI favoring longer words and affiliations.
A study by Osaka Metropolitan University found that ChatGPT's diagnostic performance for brain tumors was comparable to that of neuroradiologists, with an accuracy rate of 73%. The model's performance varied depending on the type of clinical report written, with higher accuracy when using reports from neuroradiologist writers.
A recent study published in PNAS Nexus found that widespread adoption of large language models like ChatGPT led to a significant decline in user activity on Stack Overflow. The study highlights the impact of ChatGPT on public knowledge sharing and its implications for AI's future.
A new study reveals that AI chatbots like ChatGPT can offer accurate advice on low back pain, including recommended treatments and exercise. However, the systems had limitations, particularly with complex responses and risk factors.
The article discusses the potential benefits and risks of large language models (LLMs) on collective intelligence and proposes recommendations for action. LLMs can increase accessibility and accelerate idea generation, but also pose risks such as undermining motivation to contribute to collective knowledge commons.
A new AI model called Crystalyze can analyze X-ray crystallography data to determine the structure of powdered crystals. The model was trained on a database of over 150,000 materials and successfully predicted structures for over 100 previously unsolved patterns.
A survey of UK general practitioners reveals that 20% of doctors use generative AI tools like ChatGPT in their practice. The study highlights the potential benefits of AI in reducing administrative burdens and supporting clinical decision-making, but also raises concerns about errors, biases, and patient privacy.
A recent review suggests that artificial intelligence can learn by thinking, similar to humans, through processes such as explanation, simulation, analogy, and reasoning. This finding has implications for understanding the similarities and differences between human and artificial cognition, and could lead to improvements in AI systems.
A fifth of UK GPs are using generative AI in clinical practice, with most using tools for administrative tasks or to support clinical reasoning. However, experts caution against the limitations and risks of these tools, including inaccuracies, biases, and potential harm to patient privacy.
A Cornell University-led research team found that using a technique called 'cultural prompting' reduces bias in large language models like ChatGPT for over 100 countries. The approach involves specifying a cultural identity directly in prompts, adapting to different cultural norms and values.
A recent study found that ChatGPT exhibits cultural values similar to those in English-speaking and Protestant European countries. The model's responses consistently reflected a focus on self-expression values, such as environmental protection and tolerance for diversity.
A new study from Mass General Brigham researchers found that large language models demonstrated no racial or gender discrimination in opioid treatment recommendations. The results suggest that these AI models have the potential to reduce bias and improve health equity in pain management, which is a critical area where disparities exist.
Researchers at Purdue University found that autonomous vehicles can interpret and respond to commands from passengers using large language models like ChatGPT. This technology allows the vehicle to personalize its driving to a passenger's satisfaction and take into consideration traffic rules, road conditions, and weather.
A study by OHSU researchers found that generative AI models like ChatGPT outperform human students in knowledge assessments for a biomedical and health informatics course. The results raise concerns about cheating but also highlight the need to assess student learning beyond basic knowledge.
Rice University researchers developed ElasticDiffusion, a method that separates local and global signals to create non-square aspect ratio images without visual imperfections. The new approach can improve consistency and realism in AI-generated images, but still requires significant computational power.