AI Achieves High Accuracy in Early Dementia Detection Using EEG Signals
Researchers at Örebro University have developed two novel AI systems capable of detecting dementia by analyzing EEG (brain-wave) data. These systems can effectively differentiate between healthy individuals and those with dementia, including Alzheimer’s disease and frontotemporal dementia. One model, which utilizes a combination of temporal convolutional and LSTM networks, has demonstrated an accuracy of over 80%. A second model, designed with a focus on privacy and efficiency through federated learning, has reportedly achieved an impressive accuracy of over 97%.
Sources:
Schaeffler and NTU Singapore Launch Robotics Lab to Advance Humanoid Technology
Motion technology company Schaeffler has partnered with Singapore’s Nanyang Technological University (NTU) to open a new robotics and artificial intelligence laboratory. The 900-square-meter facility is part of Schaeffler’s global SHARE (Schaeffler Hub for Advanced Research) program and aims to bolster its R&D in Asia, with a specific focus on the rapidly advancing field of humanoid robotics. The collaboration will concentrate on humanoid robotics, collaborative robots, and automated mobile platforms, combining NTU’s research expertise with Schaeffler’s industrial knowledge. Schaeffler views humanoid robotics as a significant growth area and intends to be a key component supplier to humanoid manufacturers, in addition to deploying these robots in its own production facilities.
Sources:
Black Forest Labs Releases FLUX.2 for Photorealistic AI Image Generation
Black Forest Labs has released the FLUX.2 family of state-of-the-art image generation models. These new models are capable of producing photorealistic images with up to 4-megapixel resolution, featuring realistic lighting and physics to reduce the artificial look often associated with AI-generated visuals. Key new features include direct pose control, allowing users to specify a subject’s pose, and a multi-reference capability that can generate numerous similar image variations from up to six reference images. The models are computationally intensive, with a 32-billion-parameter model requiring 90GB of VRAM. To improve accessibility, NVIDIA collaborated with Black Forest Labs to quantize the models to FP8, which reduces VRAM requirements by 40% and increases performance by the same margin. These optimized models are available directly within the ComfyUI application.
Fraunhofer Deploys Secure In-House Generative AI Platform “FhGenie”
The Fraunhofer-Gesellschaft has successfully deployed FhGenie, an in-house generative AI platform designed to provide employees with powerful AI tools while ensuring high standards of data protection and technological sovereignty. Since its introduction in 2023, the secure, ChatGPT-like service has been adopted by nearly 12,000 employees, who generate over a million prompts monthly. The platform is GDPR-compliant and gives the organization full control by utilizing its own language models and software. FhGenie integrates more than 20 large language models, including GPT-5, Claude Sonnet 4.5, and Teuken 7B, a model trained in all 24 official EU languages. Recent updates have expanded its capabilities to include image generation with DALL-E 3, web and intranet search, and integration with internal systems.
Sources:
The Evolving Role of Enterprise Architecture in the Age of AI
In an era driven by artificial intelligence, the function of enterprise architecture (EA) is shifting from a rigid control mechanism to a strategic enabler of business agility and innovation. Traditionally, EA provided a framework to standardize IT landscapes and manage complexity. However, the dynamic nature of AI, with self-retraining models and evolving data pipelines, renders traditional EA frameworks like TOGAF and Zachman less effective. The article argues that architecture must be embedded throughout the entire development lifecycle, rather than being a final checkpoint. This evolution requires leaders to view EA as a strategic operating model that fosters speed, safety, and responsibility in AI implementation.
Sources:
Aramco Deploys Saudi Arabia’s First Industrial Quantum Computer
Aramco, in collaboration with Pasqal, has deployed Saudi Arabia’s first quantum computer, which is also the first in the Middle East dedicated to industrial applications. The neutral-atom quantum computer, located at Aramco’s data center in Dhahran, is capable of controlling 200 qubits in programmable two-dimensional arrays. This initiative is aimed at accelerating the development of quantum applications in the energy, materials, and industrial sectors within the region. The partnership also includes training programs and joint research opportunities to foster a local quantum ecosystem and develop technical talent.
Sources:
IonQ and Heven AeroTech Partner to Develop Quantum-Enabled Drones
IonQ has announced a strategic partnership and investment in Heven AeroTech to integrate quantum technologies into advanced hydrogen-powered unmanned aerial systems (UAS). The collaboration aims to incorporate IonQ’s quantum computing, networking, sensing, and security technologies into Heven AeroTech’s autonomous drone platform. This integration is expected to enhance mission resilience, stealth, and operational performance, particularly in GPS-denied environments. The partnership will focus on creating ultra-secure communication links between drones and using quantum computing to optimize fleet routing and analyze imagery in real-time.
Sources:
KServe Joins CNCF to Standardize AI Model Serving on Kubernetes
The Cloud Native Computing Foundation (CNCF) has accepted the open-source project KServe as an incubating project. KServe provides a standardized platform for deploying, serving, and scaling both predictive and large language models on Kubernetes. Originally developed by Google, IBM, Bloomberg, Nvidia, and Seldon Technologies, KServe aims to simplify the complexity of modern AI workloads by offering a unified solution for model serving. Its features include multi-framework support, intelligent routing, autoscaling, and GPU acceleration. By joining the CNCF, KServe’s development will be driven by community needs, with the goal of becoming a fully abstracted inference platform that allows users to focus on models while KServe manages orchestration and scaling.
Sources:
How to Ship Syslog Data to Elasticsearch Using Fluent Bit
A new guide details how to set up Fluent Bit as a centralized server to collect and forward Syslog data to Elasticsearch for analysis. Fluent Bit, a lightweight and open-source log processor and forwarder, can be configured to accept logs over UDP. This is particularly useful as Syslog remains a standard for network devices and legacy systems, even with the rise of structured JSON logging. The process involves configuring Fluent Bit to listen for Syslog messages and then ship them to an Elasticsearch instance. This approach allows for the centralization of logs from various sources, making them easily searchable and analyzable in Kibana. Using Fluent Bit decouples the logging process from the application, provides buffering and retries, and can route logs to multiple destinations.