Insilico Medicine has completed IND-enabling studies for ISM8969, an orally available NLRP3 inhibitor targeting inflammation and neurodegenerative disorders. The drug candidate demonstrated favorable PD/PK profiles and dose-dependent efficacy in motor functions in animal models.
Researchers trained GAMES to generate accurate SMILES combinations, enhancing computer-aided drug design and accelerating discovery. The model combines LoRA and QLoRA techniques to reduce hardware and energy needed.
Researchers at King's College London found that AI chatbots can be exploited to reveal up to 12.5 times more personal information from users. The study used three types of malicious AIs to encourage disclosure of private info, with reciprocal strategies proving the most effective.
A study of 1,100 professionals reveals that AI tools can make managers' emails more professional but also undermine trust when used for routine communication tasks. Employees perceive AI-generated content as less sincere and question their leadership abilities.
Researchers found that large language models like ChatGPT outperform humans in identifying neuromyths, but struggle to critically correct them. By prompting AI to explicitly correct false assumptions, error rates can be significantly reduced.
A large-scale study found that Green AI significantly improves operational and environmental performance in Pakistani SMEs, but only when leaders invest, and institutions are in place. Adoption works under the right conditions, with perceived ease of use and usefulness driving behavior.
A study by researchers at UC San Diego and University of Chicago found that most visual artists lack the technical know-how to block AI crawlers, despite wanting to protect their creative work. The team developed a tool called Glaze, but more effective solutions are needed to prevent AI data harvesting.
A new framework leverages automated 3D object segmentation and mesh reconstruction to simulate dynamic and realistic construction sites. The approach captures real-world scenes, extracts key objects, and reconstructs them into editable 3D meshes, enabling applications such as virtual safety walkthroughs and AI-driven layout planning.
SourceELSP·JournalSmart Construction·TypeComputational simulation/modeling·DateAug 4, 2025
The article provides reporting recommendations for studies evaluating chatbot health advice, highlighting the importance of standardization in interpretability. Key findings include the need for clear guidance on chatbot performance metrics and data quality control.
Researchers found that nonequilibrium thermodynamics explains why optimal transport theory is optimal in diffusion models, leading to robust data generation. The discovery offers a novel thermodynamic approach to machine learning research, particularly in image generation.
The journal is seeking high-quality submissions that explore the potential of voice and speech analysis in diverse medical applications. Submissions focus on acoustic and voice analysis techniques, voice or vocal phenotyping, and vocal biomarkers in health and medicine.
DPM-Solver++ is a fast training-free solver for guided sampling of diffusion probabilistic models. It addresses instability issues with high-order solvers and achieves improved sample quality in just 15 steps, making it a significant advancement in the field.
A team of computer scientists created 2,300 original sudoku puzzles and asked AI tools like OpenAI's ChatGPT to solve them. The results showed that while some AI models could solve easy sudokus, most struggled to provide accurate explanations, raising questions about the trustworthiness of AI-generated information.
The team aims to deliver AI power directly to devices, improving resilience and speed in constrained environments. By processing data step-by-step across a network of devices, they can create a safe and adaptable system that can withstand attacks and extreme conditions.
Researchers create AI model called UNITE that detects fakes by analyzing faces and backgrounds, making it one of the first tools capable of identifying synthetic or doctored videos. The system can flag a range of forgeries, from simple facial swaps to complex fully synthetic videos.
The Insilico Medicine recent summer updates webinar introduced new features and functional updates to the company's generative AI platform, including Generative Biologics, Chemistry42, PandaOmics, DORA, and PreciousGPT. These updates aim to accelerate drug discovery and advance life sciences research.
Researchers found that large language models are unaware of their own mistakes, unlike humans who can adjust expectations. The models' lack of self-awareness has implications for the use of AI chatbots in daily life, particularly when it comes to tasks with low stakes.
A University of Kansas study found that people rate corporate crises messages written by humans as more credible and trustworthy, regardless of the approach taken. However, the approach itself didn't vary between participants who read human or AI-written content.
Researchers used generative AI models like ChatGPT and DALL-E to analyze men's fashion trends and create realistic fashion collection images. The study highlights the importance of expertly worded prompts for accurate fashion design implementation, enabling efficient and creative use of generative AI in fashion.
Researchers from the Weizmann Institute of Science and Intel Labs have developed new algorithms that allow AI developers to combine the power of different AI models, speeding up large language model performance by 1.5-2.8 times. This enables faster collaboration between models, reducing processing power costs and improving overall effi...
The Hong Kong Polytechnic University has secured funding from the Research Grants Council to develop a cost-effective and sustainable collaborative Generative AI (Co-GenAI) model. The project aims to democratize AI development by reducing centralized computational demand and enabling broader participation in GenAI research and deployment.
Researchers used AI model ChatGPT-4 to extract clinical variables from MRI and CT scans of nearly 1,000 adults with pancreatic cysts. The accuracy rate ranged from 97% for solid components to 99% for calcific lesions, comparable to human performance.
A recent survey found that nearly two-thirds of U.S. adults consider AI-generated summaries to be somewhat or very reliable, while another 45% say they sometimes give them the answer they need. However, half of respondents are not comfortable with healthcare providers using AI tools.
Researchers found that LLMs are better at representing non-sensorimotor concepts but struggle with sensory and motor concepts. Incorporating sensory input improves LLM performance and leads to more human-like representation, consistent with previous human studies on multimodal learning.
A new report from the University of Cambridge's Minderoo Centre for Technology and Democracy highlights the alarming trajectory of energy use by big tech companies, driven by AI growth. The report predicts that data centres could consume up to 8% of global greenhouse gas emissions by 2040, far exceeding current emissions from air travel.
Australian scientists used AI to generate a protein that can kill antibiotic-resistant bacteria like E. coli, a major step forward in combating superbugs. The new approach enables efficient de novo design of proteins with specific characteristics and functions.
A recent study published in the Association for Computing Machinery Conference on Fairness, Accountability, and Transparency found that AI chatbots are not safe replacements for therapists. The researchers evaluated these chatbots against clinical standards for therapists and concluded that they don
Insilico Medicine's latest Pharma.ai updates showcase advancements in Biology42 and Chemistry42, enhancing antibody design and omics data analysis. The company's AI software suite has demonstrated significant efficiency improvements in practical applications, setting a benchmark for AI-driven drug research.
A new study reveals that AI systems transition from relying on word positions to meaning-based understanding as they receive enough data for training. The transition occurs abruptly, similar to a phase transition in physical systems, and is driven by the amount of data available.
SourceSissa Medialab·JournalJournal of Statistical Mechanics Theory and Experiment·TypeData/statistical analysis·DateJul 7, 2025
A new technique called WeGeFT improves large language model performance in tasks such as commonsense reasoning and code generation. By fine-tuning key parameters, researchers reduce the need for significant computational power, advancing the field of artificial intelligence.
At the upcoming AI for Good Global Summit, Alex Zhavoronkov PhD will discuss the future potential of generative AI. Insilico Medicine's founder and CEO has invented critical technologies in generative AI and reinforcement learning, and his keynote will explore applications in drug discovery and longevity.
A new study published in the Journal of Experimental Psychology found that interacting with robots through social games makes them seem more human-like. The researchers used a box-shaped robot called Cozmo and found that participants who played games with it considered it more human-like, whereas those who interacted mechanically did not.
The College of Engineering at Texas A&M is developing a suite of university-wide resources to integrate generative AI into course material, research, and outreach. The initiative aims to make generative AI a core part of the academic toolkit accessible to faculty across disciplines.
A new study in Nature Communications found that AI models exhibit a geometric property called convexity, which helps humans form and share concepts. Convexity is also linked to the performance of AI models on specific tasks.
The Hebrew University of Jerusalem is part of a €9.3 million, 36-month Horizon Europe project to develop trustworthy GenAI systems. The consortium aims to integrate neural and symbolic methods for human-centric design in complex systems.
Researchers propose a paradigm shift toward an educational model built on human interdependence, urging a re-examination of success and purpose. The new approach emphasizes uniquely human capacities, creativity, emotional intelligence, and collaboration with AI.
Paulina DeVito, a Ph.D. candidate in computer science at Florida Atlantic University, has been awarded the NSF Graduate Research Fellowship to pursue research on developing large language model-based approaches for analyzing public discourse on social media. Her work aims to understand how people discuss emerging technologies and infor...
Researchers developed a puzzle game called AI Puzzlers that teaches kids to spot errors in AI responses and think creatively. By comparing their solutions to those from various AI chatbots, kids learn to use AI as a tool that needs guidance rather than an answer machine.
The AI for Good Global Summit 2025 will showcase AI innovations delivering better healthcare and education, reducing disaster risks, ensuring water and food security, and bolstering economic resilience. The event, organized by the International Telecommunication Union (ITU), features talks from AI leaders and 100+ demos.
Researchers discovered AI art protection tools have critical weaknesses that cannot reliably stop AI models from training on artists' work. LightShed, a new method, can detect and remove distortions, stripping away protections and rendering images usable again for generative AI model training.
Computer scientist Zeynep Akata has developed innovative methods to combine visual, linguistic, and conceptual elements in AI to increase user trust. Her research on explainable AI aims to make image classification decisions more transparent.
City of Hope has launched a novel generative AI platform called HopeLLM, which leverages multiple commercial and open-source AI models. The platform enables doctors to spend more time interacting with patients, reducing the administrative burden of new patient onboarding and providing personalized treatment options.
The collaboration aims to bridge academia and industry through joint research projects, academic exchange, and hands-on training for students and faculty. The partnership also opens career opportunities for UAEU graduates at Insilico, fostering local talent and supporting the expansion of the UAE's biotechnology sector.
The study analyzed 50 countries' national AI strategies, finding that only 13 gave high priority to training the current workforce and improving AI education. Common themes included establishing AI-focused programs in universities and on-the-job training, with a focus on human soft skills such as creativity and collaboration.
Researchers at KAIST have developed a technology to enhance creative generation of AI generative models like Stable Diffusion, generating novel and useful images. The algorithm amplifies internal feature maps to boost creativity without new training, outperforming existing methods in novelty and utility.
Cassie, a digital-human assistant developed by Texas A&M University, is transforming the way patients interact with healthcare providers. With facial recognition and emotional intelligence, Cassie offers a two-way interaction that feels like a conversation.
Researchers at UTA are developing an AI-powered system to promote healthy behaviors in cancer survivors, which can significantly improve quality of life and reduce risk of death.
Global education leaders call for collaboration, ethics, and human-centered teaching as they confront the benefits and challenges of AI in education. The discussion emphasized the need for responsible use policies, equitable access to AI tools, and preserving uniquely human qualities in education.
Researchers used nonsense words to test ChatGPT's language processing capabilities, finding it excelled at discovering relationships and providing definitions for extinct words. However, the AI also generated incorrect or made-up answers in some cases, highlighting its limitations.
Insilico Medicine has completed the first-in-patient dosing of ISM3412, a novel MAT2A inhibitor with potential as a best-in-class treatment for locally advanced and metastatic solid tumors. The Phase 1 study aims to evaluate safety, tolerability, and preliminary anti-tumor efficacy of ISM3412.
Nach01, a large language model-based foundation model, leverages structural and spatial data for diverse chemical tasks. The model is now available on AWS Marketplace, streamlining access for researchers and pharma companies to build and apply large-scale generative models.
Researchers developed a new computational tool using generative AI to identify key gene combinations underlying complex illnesses. The method amplifies limited gene expression data, enabling researchers to resolve patterns of gene activity that cause complex traits.
The virtual teaching assistant (VTA) provides personalized feedback to individual students even in large-scale classes. The system, which automatically vectorizes a large volume of course materials and uses them as the basis for answering students' questions, has been shown to significantly reduce the burden on TAs.
Professor Shalom Lappin argues that tech companies' dominance in AI development poses a threat to public benefit and environment. He calls for comprehensive international regulation, intellectual property reform, and preparation for workforce disruption to address these challenges.
The ELSI University Summit in Japan brought together experts to discuss challenges and prospects of responsible AI development. Key findings include the need for ethics-driven approaches, collaboration between academia and industry, and the importance of setting evaluation indicators for Responsible AI.
A field experiment by Tulane University researchers found that generative AI tools enhance employees' creativity when used by those who can manage their own thinking. Employees using ChatGPT performed better in generating novel and useful ideas.
A new study explores how generative AI models with different dimensional outputs support combinational creativity. Expert designers applied each AI type to six combinational tasks, revealing distinct divisions of labor: LLMs excel in early idea generation, while image and 3D models shine in visualization and prototyping.
Ashish Sharma's dissertation developed fundamental advances in natural language processing to positively impact mental health. His AI-supported tools have been used by over 160,000 people, with more than fifty percent of users reporting a household income of less than $40,000 per year.
The new minor will equip students with cutting-edge AI and machine learning skills essential for modern business leadership. It combines hands-on experience with practical applications of artificial intelligence in business settings.
The Phase IIa trial of Rentosertib showed encouraging clinical data, with patients receiving the treatment experiencing significant improvements in lung function compared to placebo. The study also validated the biological mechanism of TNIK inhibition using exploratory biomarkers analyses.