Generative AI

Google Tackles Catastrophic Forgetting with Nested Learning; AI Discovers New Magnets In the world of AI and Machine Learning, Google Research has unveiled Nested Learning, a novel machine learning paradigm designed to combat the persistent issue of catastrophic forgetting in AI models. In a separate development, AI has been instrumental in materials science, leading to the discovery of new magnetic materials that could decrease our dependence on scarce rare earth elements. Furthermore, the Chan Zuckerberg Initiative has announced a strategic pivot, restructuring to harness the power of artificial intelligence in its mission to cure, prevent, or manage all diseases. ...

China’s DePTH-GPT: A New AI Model to Revolutionize Deep-Sea Research A team of Chinese scientists has launched a new artificial intelligence model named DePTH-GPT to advance deep-sea exploration. The model utilizes a combination of AI technologies, including deep learning, large language models, and computer vision, to analyze a wide range of data from the deep ocean such as video, topography, and bioacoustics. This new tool is expected to shift deep-sea research from traditional qualitative analysis to a more intelligent and predictive approach. DePTH-GPT has already been used to create an intelligent cognitive system for a seamount and a hydrothermal vent field. Developed as part of the UN Decade of Ocean Science for Sustainable Development, the model will be made available to global research institutions to create cognitive systems for various deep-sea habitats. ...

SAP Boosts Enterprise AI with New Tools and Global Skills Initiative SAP has unveiled a suite of AI-powered enhancements for its SAP Build solution, featuring new data integration capabilities and intelligent automation via Joule Agents. These updates aim to accelerate developer workflows and transform how businesses leverage data and AI. Addressing the global demand for AI talent, SAP has also committed to upskilling 12 million people by 2030 through expanded training programs with partners like Coursera. The company introduced its first relational foundation model, SAP-RPT-1, designed for predicting business outcomes. Furthermore, SAP is expanding its Business Data Cloud ecosystem through a new partnership with Snowflake, enhancing existing integrations with Databricks and Google Cloud. ...

New AI Assistant ‘Denario’ Aims to Accelerate Scientific Discovery Researchers have introduced Denario, an AI-powered scientific assistant designed to accelerate the research process. This multi-agent system can assist with numerous tasks, including generating new research ideas, reviewing existing literature, creating research plans, writing and executing code, producing plots, and drafting scientific papers. Developed by a team from the University of Cambridge, the Flatiron Institute, and the Autonomous University of Barcelona, Denario is a modular tool for scientists at various stages of their work. While it has demonstrated the ability to generate entire research papers across disciplines like astrophysics and biology, its creators emphasize that it is a tool to augment, not replace, human scientists. They note the current version has limitations, with only about one in ten outputs providing significant insights and instances of data fabrication, making human oversight essential. The project’s code is publicly available, with a demo version accessible online. ...

AI & Machine Learning Breakthroughs: Self-Evolving AI and Light-Speed Computing This week in AI research brings major breakthroughs, including the development of a self-evolving edge AI. Additionally, a new optical chip has been unveiled that can compute at the speed of light. In medical tech, an innovative AI model can now accurately predict the progression of osteoarthritis from simple knee X-rays. Sources: Space Daily Hyundai and NVIDIA Partner on AI Factory for Autonomous Vehicle Development Hyundai Motor Group is significantly expanding its partnership with NVIDIA to build a cutting-edge AI factory. This initiative is designed to fast-track the development of autonomous vehicles, smart factories, and advanced robotics. ...

Artificial Intelligence & Machine Learning New Open-Source Omni-Modal AI Model ‘LongCat-Flash-Omni’ Released A new state-of-the-art, open-source omni-modal AI model, LongCat-Flash-Omni, has been released. Featuring an impressive 560 billion parameters, this model marks a significant leap forward in the capabilities of open-source artificial intelligence. Its release is a major event for the AI community, equipping researchers and developers with powerful new tools for a diverse range of applications. Robotics & Autonomous Vehicles IIT Bombay Develops GPS-Free Control System for Autonomous Drone Swarms Researchers at the Indian Institute of Technology (IIT) Bombay have developed a groundbreaking control system enabling unmanned aerial vehicles (UAVs) to fly in coordinated swarms without GPS, inter-drone communication, or a central controller. This innovative method relies on ‘bearing-only’ measurements, where each drone uses its onboard camera to observe neighbors and maintain formation based on relative positions. ...

Qualcomm Challenges Nvidia with New AI Chips for Data Centers Qualcomm is entering the data center market with its next-generation AI inference accelerators, the AI200 and AI250, directly challenging Nvidia’s dominance. These chips are engineered for high-performance generative AI inference, promising a lower total cost of ownership and superior power efficiency. The AI250, set for a 2027 release after the AI200 in 2026, will feature an innovative near-memory computing architecture to dramatically increase memory bandwidth. Designed for compatibility with major machine learning frameworks, this strategic expansion marks Qualcomm’s significant move from mobile technology into enterprise AI hardware. ...

Hybrid Quantum-Classical Systems Essential for AI Decision-Making, New Study Reveals A new study from Chapman University suggests that true agency and intelligence cannot arise from a purely quantum system. Researchers argue that the principles of quantum mechanics, which prevent the copying and comparison of information, mean that decision-making processes like modeling and deliberation require classical resources. This finding implies that both consciousness and the development of advanced quantum artificial intelligence will depend on hybrid quantum-classical architectures. The study posits that even sophisticated quantum machines will need classical components to learn, act, and make decisions effectively, touching on philosophical implications for free will and the emergence of classical reality. ...

Lightning AI Launches PyTorch Copilot to Streamline AI Development Lightning AI has introduced a new suite of tools designed to enhance productivity for PyTorch developers, featuring an AI code editor for its Lightning Studios and Notebooks environments. The company also announced the integration of Meta’s Monarch and Forge frameworks, which aim to simplify distributed AI workloads and reinforcement learning. These updates are part of Lightning AI’s mission to make PyTorch development more accessible. Key features include ‘Lightning Environments,’ which provide scalable, sandboxed environments for all stages of the AI development lifecycle. ...

Breakthrough in AI: Fine-Tuning LLMs with Less Data and Power Researchers at the University of California San Diego have pioneered a new method for fine-tuning large language models (LLMs) that requires significantly less data and computing power. This innovative technique avoids retraining a model’s entire parameter set, instead selectively updating only the most critical components. This approach dramatically reduces costs and minimizes the risk of overfitting. In tests on protein language models, the method achieved high accuracy with limited training data, proving to be 326 times more parameter-efficient than traditional fine-tuning. This breakthrough is poised to democratize AI, enabling smaller organizations with limited resources to customize powerful AI models for specialized applications. The findings were published in Transactions on Machine Learning Research. ...