This study investigates methods to steer Arabic Large Language Models (LLMs) towards generating specific dialects, addressing the challenge of data scarcity for dialectal Arabic. Researchers identified sparse neuron populations encoding dialect-specific features and developed a vector-steering approach using dialect-specific activation directions. These inference-time methods allow for controlling dialectal output by amplifying or suppressing neuron activity or injecting specific vectors. Why it matters: This research offers a principled, interpretability-grounded framework to improve dialectal accuracy in Arabic LLMs without fine-tuning, crucial for enhancing their utility in the diverse Arabic-speaking world.
MBZUAI researchers are developing spiking neural networks (SNNs) to emulate the energy efficiency of the human brain. Traditional deep learning models like those powering ChatGPT consume significant energy, with a single query using 3.96 watts. SNNs aim to mimic biological neurons more closely to reduce energy consumption, as the human brain uses only a fraction of the energy compared to these models. Why it matters: This research could lead to more sustainable and energy-efficient AI technologies, addressing a major challenge in deploying large-scale AI systems.
Scientists at King Abdullah University of Science and Technology (KAUST) have discovered that astrocytes, a type of non-neuronal cell, actively influence communication between neurons and processes linked to learning and memory. Their study, published in The Journal of Physiology, revealed that lactate not only provides energy to neurons but also triggers a signaling pathway within them, challenging the traditional view of astrocytes as passive support cells. This research, conducted largely at KAUST with international collaboration from Italy and Switzerland, provides new insights into the molecular mechanisms supporting brain cell communication and synaptic plasticity. Why it matters: These findings are significant for understanding fundamental brain science and could open new avenues for therapeutic strategies targeting neurological disorders and memory loss.
KAUST is highlighting women in STEM, including Professor Leena Ibrahim, Ph.D. student Amani Al-Amodi, and postdoc Dr. Wejdan Alghamdi. Leena Ibrahim's research focuses on understanding how sensory perception is established across development, studying the role of inhibitory neurons in the cortex. She aims to uncover how disruptions in sensory processing contribute to neurodevelopmental disorders like autism. Why it matters: Showcasing women's contributions can inspire future generations of female researchers in the Kingdom and beyond.
KAUST professor Pierre Magistretti has been elected to the Norwegian Academy of Science and Letters. His election recognizes his contributions to neuroscience, specifically his work on lactate's role in brain function. Magistretti's research focuses on the lactate shuttle system and how neurons and glial cells cooperate to meet energy demands. Why it matters: This honor highlights KAUST's contribution to international neuroscience and can foster further collaboration in the field.
KAUST's Vice President of Research, Professor Pierre Magistretti, presented the 2023 Nansen Neuroscience Lecture in Oslo. The lecture highlighted his discovery of the astrocyte-neuron lactate shuttle and its role in brain physiology and pathology. His research explains how astrocytes feed neurons lactate as a primary energy source, impacting memory and cognition. Why it matters: The lecture recognizes KAUST's contribution to neuroscience and highlights the importance of understanding brain energy consumption for treating neurological disorders.
KAUST and EPFL Blue Brain Project researchers propose a new theory about a 'secret language' used by cells for internal communication regarding the external world. Using a computational model, they suggest that metabolic pathways can code details about neuromodulators that stimulate energy consumption. The model focuses on astrocytes and their cooperation with neurons in fueling the brain. Why it matters: This suggests a new avenue for understanding information processing in the brain and how cells contribute to the energy efficiency of brains compared to computers.
KAUST researchers collaborated with the Blue Brain Project to study astrocytes, brain cells crucial for memory and learning. Dr. Corrado Calì produced 3D models of astrocytes using serial block-face electron microscopy to understand their structure. The study, published in Progress in Neurobiology, reveals how lactate transfer from astrocytes to neurons contributes to brain energy usage. Why it matters: Understanding astrocyte function could lead to new drugs for treating conditions like stroke and Alzheimer's disease by improving brain cell function.