Tech Firms From Dell to HP Warn of Memory Chip Squeeze From AI

Dell Technologies Inc. and HP Inc. have issued warnings about potential memory-chip supply shortages in the upcoming year, attributing this to the surging demand driven by the expansion of artificial intelligence infrastructure. This situation reflects the increasing reliance on AI technologies across various sectors, which is expected to strain supply chains.

A recent study has introduced a framework that enhances the efficiency of large language models (LLMs) by combining fine-tuning and rectification techniques. This approach optimally allocates limited labeled samples to improve LLM predictions and correct biases in outputs, addressing challenges in market research and social science applications.

OpenCML has introduced an innovative end-to-end framework for open-world machine learning, enabling systems to learn unknown classes incrementally. This approach addresses the limitations of traditional closed-world models, allowing for continual learning and adaptation to new data. The model has shown superior performance compared to existing methods in open-world learning scenarios.

Dell Technologies Inc. has raised its annual projections for shipments of artificial intelligence servers, indicating a strong and sustained demand in the data center market. This adjustment reflects the company's confidence in the growth of AI technologies and their applications in various sectors.

At Microsoft Ignite 2025, Microsoft Corp. unveiled its AI roadmap, showcasing a shift from simple AI copilots to fully integrated agents across data infrastructures, alongside partners like Dell Technologies Inc. and Oracle Corp. The event featured over 70 product announcements, including Agent 365, a new control plane for AI management.

A new study has introduced a model that combines deep learning systems with metaheuristic approaches, specifically targeting the detection of anxiety disorders. This model utilizes advanced artificial intelligence techniques to analyze multimodal datasets, including physiological and behavioral signals, thereby moving beyond traditional subjective assessments that can vary significantly between evaluators.

A new method called FedRef has been proposed for federated learning, focusing on communication-efficient Bayesian fine-tuning using a reference model. This approach aims to mitigate issues such as catastrophic forgetting, which can degrade model performance due to data and system heterogeneity among clients. By integrating a proximal term, the method enhances model performance while preserving user data privacy.

A novel approach has been introduced to enhance the performance of Spiking Neural Networks (SNN) by utilizing bit plane representation and color models. This method aims to improve accuracy without increasing the model size, addressing a significant challenge in the practical adoption of SNNs in computational neuroscience and artificial intelligence.