In a groundbreaking move, Microsoft Research Asia’s prestigious StarTrack Scholars has officially taken flight, extending a global invitation to brilliant young minds for an immersive three-month research visit. Picture this: collaboration with elite researchers, a deep dive into the Microsoft Research environment, and a valuable opportunity to transform academic brilliance into real-world impact.
The future of interdisciplinary research between artificial intelligence (AI) and brain science is exceptionally promising, offering potential breakthroughs that could revolutionize multiple fields. The fusion of these disciplines may ultimately unlock new horizons in human knowledge and capability, driving progress in healthcare, technology, and beyond.
If you are an aspiring young researcher with a zeal for exploring the intersection of AI and brain science, we invite you to apply to the Microsoft Research Asia StarTrack Scholars Program. Applications are now open for the 2025 program. Apply now and become a part of this transformative journey. For more details and to submit your registration, visit our official website: Microsoft Research Asia StarTrack Scholars Program – Microsoft Research
Create a Synergistic Relationship between AI and the Brain
The brain is one of the most complex objects in the world. Although our research on the brain has been ongoing for thousands of years, there are still many mysteries about the human brain.
The team hopes to conduct interdisciplinary research and by using artificial intelligence technology to help neuroscientists better understand the brain. This understanding not only aids in exploring the mechanisms of brain diseases and promoting brain health, but also provides inspiration from the brain to design smarter artificial intelligence.
To create a synergistic relationship between AI and the brain, Dongsheng Li and his colleagues at Microsoft Research Asia – Shanghai emphasize the need to integrate expertise in both AI and brain science. This integration is essential for bridging the gap and uncovering new opportunities. Their research focuses on brain-inspired AI, brain-computer interfaces, and AI for brain health, all of which hold significant importance for society and humanity.
Brain-inspired AI: Applying the Efficient Mechanisms of the Brain to AI
As AI research and technology continue to advance, it is crucial to consider the energy and infrastructure resources needed to manage large datasets, perform complex computations, and handle open-ended tasks. The human brain serves as an exemplary model of efficiency, adeptly managing intricate tasks with minimal resources. Inspired by this, the team aims to understand the brain’s efficient processes and replicate them in AI.
In collaboration with the partners, the team are exploring three research directions to foster more sustainable AI. First, leveraging the energy-efficient spiking neurons in the brain could make the computational mechanisms in artificial neural networks up to three orders of magnitude more efficient. Second, designing new neural network architectures that mimic the brain’s learning and computational methods could enhance learning efficiency. Third, embodied AI, when interacting with the real world, can draw from the human brain’s strategies to operate efficiently and effectively in open-ended environments and goals.
Brain-computer Interface: Promote EEG Decoding of Brain Signals
Understanding how the brain works is crucial for addressing the fundamental scientific question of the origin of intelligence, as well as developing next-generation brain-computer interface (BCI). Electroencephalogram (EEG) signals are among the most popular tools for studying the brain with non-invasive electrodes because of its convenience and reasonable quality for decoding brain states including both what people sense (perception) and what people want (control).
However, decoding brain signals with non-invasive EEG is a rather challenging task because of the lack of data and neuroscience guarantees. To address these challenges, the first promising research direction is to build the foundation models for understanding EEG signals, e.g., self-supervised learning on EEG signals or multi-modal learning between EEG and human language, by leveraging large-scale unlabeled EEG data. The other promising direction is to combine neuroscience knowledge in machine learning algorithm design, e.g., designing more bio-plausible decoding algorithms or BCI paradigms.
AI for Brain Health: Advancing the Understanding of Brain Diseases
AI can help conquer brain disorders in several ways, including diagnostics, mechanism understanding, and treatment.
In diagnostics, machine learning algorithms can analyze complex medical data, such as EEG signals, genetic information, and MRI scans, with remarkable accuracy and speed, enabling earlier and more precise identification of brain disorders.
For mechanism understanding, AI can sift through digests vast amounts of research data to uncover patterns and insights that may not be immediately apparent to human researchers, thereby advancing our knowledge of the underlying causes and progression of neurological diseases.
In treatment, AI-driven tools can personalize therapy plans by predicting individual responses to different treatments, optimizing drug dosages, and even assisting in the development of new medications. By integrating AI into these aspects, we can significantly improve outcomes for patients with brain disorders, making diagnosis quicker, understanding more comprehensive, and treatment more effective.
Cross-Disciplinary Collaboration: Looking for Talents across Various Fields
Collaborations between AI researchers and brain science or neurology experts offer tremendous potential, but they also come with significant challenges.
Data quality and availability is also a huge challenge. High-quality, standardized, and sufficiently large datasets are crucial for training AI models, but such datasets are often difficult to obtain in brain science and neurology due to ethical, privacy, and logistical constraints.
Brain data is highly complex and variable. AI models need to be able to handle the intricacies of neural data, such as high dimensionality, noise, and non-linear relationships. Also, brain science often involves multimodal data, including imaging, electrophysiological recordings, and behavioral data. Integrating these diverse data types into a coherent AI model is challenging.
AI models, especially deep learning models, are often seen as “black boxes.” Ensuring that these models provide interpretable and actionable insights for neuroscientists and clinicians is a significant challenge.
In addressing these unresolved and challenging issues, Microsoft Research Asia StarTrack Scholars advocates an open attitude, encouraging dialogue and joint experimentation with researchers from various disciplines to discover viable solutions. Now visit our official website to know more: Microsoft Research Asia StarTrack Scholars Program – Microsoft Research
Reference:
[1] (opens in new tab)https://www.msra.cn/zh-cn/news/executivebylines/dongsheng-ai-neuroscience (opens in new tab)
[3] Yansen Wang, Xinyang Jiang, Kan Ren, Caihua Shan, Xufang Luo, Dongqi Han, Kaitao Song, Yifei Shen, Dongsheng Li. CircuitNet: A Generic Neural Network to Realize Universal Circuit Motif Modeling. ICML 2023.
https://proceedings.mlr.press/v202/wang23k/wang23k.pdf (opens in new tab)
[4] Ke Yi, Yansen Wang, Kan Ren, Dongsheng Li. Learning Topology-Agnostic EEG Representations with Geometry-Aware Modeling. NeurIPS 2023.
https://openreview.net/attachment?id=hiOUySN0ub&name=pdf (opens in new tab)
[5] Wei-Bang Jiang, Yansen Wang, Bao-Liang Lu, Dongsheng Li. NeuroLM: A Universal Multi-task Foundation Model for Bridging the Gap between Language and EEG Signals, 2024. https://arxiv.org/abs/2409.00101 (opens in new tab)
[6] Xuanhao Liu, Yan-Kai Liu, Yansen Wang, Kan Ren, Hanwen Shi, Zilong Wang, Dongsheng Li, Bao-liang Lu, Wei-Long Zheng. EEG2Video: Towards Decoding Dynamic Visual Perception from EEG Signals. NeurIPS 2024. https://openreview.net/pdf?id=RfsfRn9OFd (opens in new tab)
[7] Dongqi Han, Kenji Doya, Dongsheng Li, Jun Tani. Synergizing habits and goals with variational Bayes. Nature Communications, volume 15, Article number: 4461 (2024). https://www.nature.com/articles/s41467-024-48577-7 (opens in new tab)
[8] Changze Lv, Yansen Wang, Dongqi Han, Xiaoqing Zheng, Xuanjing Huang, Dongsheng Li. Efficient and Effective Time-Series Forecasting with Spiking Neural Networks. ICML 2024. https://arxiv.org/abs/2402.01533 (opens in new tab)
[9] Changze Lv, Dongqi Han, Yansen Wang, Xiaoqing Zheng, Xuanjing Huang, Dongsheng Li. Advancing Spiking Neural Networks for Sequential Modeling with Central Pattern Generators. NeurIPS 2024. (Spotlight) https://arxiv.org/abs/2405.14362 (opens in new tab)
[10] PhysioPro Project: https://github.com/microsoft/PhysioPro (opens in new tab)
Theme Team:
Dongsheng Li, Principal Research Manager, Microsoft Research Asia
Yansen Wang, Senior Researcher, Microsoft Research Asia
Dongqi Han, Senior Researcher, Microsoft Research Asia
If you have any questions, please email Ms. Yanxuan Wu, program manager of the Microsoft Research Asia StarTrack Scholars Program, at v-yanxuanwu@microsoft.com
The post Towards a Synergy between AI and the Brain: Microsoft Research Asia StarTrack Scholars 2025 Explores the Frontiers of AI and Brain Science with You appeared first on Microsoft Research.