AI Institute: AI Research Institute for Fundamental Interactions

人工智能研究所：人工智能基础交互研究所

• 批准号: 2019786
• 负责人: Jesse Thaler
• 金额: $ 2000万
• 依托单位: Massachusetts Institute of Technology
• 依托单位国家: 美国
• 项目类别: Cooperative Agreement
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-11-01 至 2026-04-30
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2019786&HistoricalAwards=false
• 关键词: AI; Institute; Research; Fundamental; Interactions

The Institute for Artificial Intelligence and Fundamental Interactions (IAIFI) will enable physics discoveries and advance foundational artificial intelligence (AI) through the development of novel AI approaches that incorporate first principles from fundamental physics. AI is transforming many aspects of society, including the ways that scientists are pursuing groundbreaking discoveries. For many years, physicists have been at the forefront of applying AI methods to investigate fundamental questions about the Universe. As an example, AI played a key role in the discovery and study of the Higgs boson, the last missing ingredient in the Standard Model of particle physics. Further progress will require a revolutionary leap in AI, as both the complexity of physics problems and the size of physics datasets continue to grow. The goal of the IAIFI is to develop and deploy the next generation of AI technologies, based on the transformative idea that artificial intelligence can directly incorporate physics intelligence. IAIFI researchers will use these new AI technologies to tackle some of the most challenging problems in physics, from precision calculations of the structure of matter, to gravitational wave detection of merging black holes, to the extraction of new physical laws from noisy data. IAIFI researchers will also transfer these technologies to the broader AI community, since trustworthy AI is as important for physics discovery as it is for other applications of AI in society. To cultivate human intelligence, the IAIFI will promote training, education, and outreach at the intersection of physics and AI. In this way, the IAIFI will advance physics knowledge – from the smallest building blocks of nature to the largest structures in the Universe – and galvanize AI research innovation.The IAIFI will enable physics discoveries and advance foundational AI through the development of novel “Ab initio AI” approaches that incorporate first principles and best practices from fundamental physics. Ab initio AI will make intractable theoretical physics calculations feasible, predicting complex emergent phenomena that are computationally daunting to tackle even though the underlying physical laws are well understood. It will also transform many experimental physics applications, where ab initio principles will be used to design AI methods that are more easily verifiable using well-understood calibration data samples, leading to better quantification of uncertainties. While each physics use-inspired goal will present its own issues, the IAIFI’s focus will be on finding shared solutions, since these problems involve similar prior knowledge, are based on the same underlying ab initio principles, and face common experimental and theoretical challenges. The same challenges that arise in the development and deployment of AI methods across a broad spectrum of frontier physics research, including verification and interpretability of AI solutions, are also faced in other AI application domains. Therefore, by developing ab initio AI, the IAIFI will accelerate the pace of physics discovery, extend the frontiers of AI research, and develop new pathways for broad adoption. The IAIFI will make Cambridge and the surrounding Boston area a nexus point for collaborative efforts aimed at advancing both physics and AI and at connecting to industry and science partnerships. Part of the IAIFI mission will be to disseminate knowledge about (and enthusiasm for) physics, AI, and their intersection through various workforce development, digital learning, outreach, broadening participation, and knowledge transfer programs.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

人工智能和基本相互作用研究所（IAIFI）将通过开发新的人工智能方法，结合基础物理学的第一原理，实现物理学发现和推进基础人工智能（AI）。人工智能正在改变社会的许多方面，包括科学家追求突破性发现的方式。 多年来，物理学家一直站在应用人工智能方法研究宇宙基本问题的最前沿。 例如，人工智能在希格斯玻色子的发现和研究中发挥了关键作用，希格斯玻色子是粒子物理学标准模型中最后一个缺失的成分。 进一步的进展将需要人工智能的革命性飞跃，因为物理问题的复杂性和物理数据集的规模都在不断增长。 IAIFI的目标是开发和部署下一代人工智能技术，基于人工智能可以直接融入物理智能的变革性想法。 IAIFI的研究人员将利用这些新的人工智能技术来解决物理学中一些最具挑战性的问题，从物质结构的精确计算，到黑洞合并的引力波探测，再到从嘈杂的数据中提取新的物理定律。 IAIFI的研究人员还将把这些技术转移到更广泛的人工智能社区，因为值得信赖的人工智能对于物理发现和人工智能在社会中的其他应用同样重要。 为了培养人类智能，IAIFI将促进物理学和人工智能交叉领域的培训、教育和推广。 通过这种方式，IAIFI将推进物理学知识--从自然界最小的构建块到宇宙中最大的结构--并激发人工智能研究创新。IAIFI将通过开发新颖的“从头算人工智能”方法，结合基础物理学的第一原理和最佳实践，实现物理学发现并推进基础人工智能。 从头算人工智能将使棘手的理论物理计算变得可行，预测复杂的新兴现象，这些现象在计算上令人生畏，即使基本的物理定律已经很好地理解。 它还将改变许多实验物理应用，其中从头算原理将用于设计AI方法，这些方法更容易使用易于理解的校准数据样本进行验证，从而更好地量化不确定性。 虽然每一个物理用途启发的目标将提出自己的问题，IAIFI的重点将是寻找共同的解决方案，因为这些问题涉及类似的先验知识，基于相同的基本从头算原理，并面临共同的实验和理论挑战。在广泛的前沿物理研究中开发和部署人工智能方法所面临的挑战，包括人工智能解决方案的验证和可解释性，也在其他人工智能应用领域面临。因此，通过开发从头开始的人工智能，IAIFI将加快物理发现的步伐，扩展人工智能研究的前沿，并为广泛采用开发新的途径。 IAIFI将使剑桥和周围的波士顿地区成为合作努力的连接点，旨在推进物理学和人工智能，并与工业和科学合作伙伴关系建立联系。 IAIFI的部分使命是通过各种劳动力发展、数字学习、推广、扩大参与和知识转移计划，传播有关物理学、人工智能及其交叉点的知识（和热情）。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

Going beyond the galaxy power spectrum: An analysis of BOSS data with wavelet scattering transforms

• DOI: 10.1103/physrevd.106.103509
• 发表时间: 2022-04
• 期刊: Physical Review D
• 影响因子: 5
• 作者: Georgios Valogiannis;C. Dvorkin

Single electrons on solid neon as a solid-state qubit platform

固体氖上的单电子作为固态量子位平台

• DOI: 10.1038/s41586-022-04539-x
• 发表时间: 2022
• 期刊: Nature
• 影响因子: 64.8
• 作者: Zhou, Xianjing;Koolstra, Gerwin;Zhang, Xufeng;Yang, Ge;Han, Xu;Dizdar, Brennan;Li, Xinhao;Divan, Ralu;Guo, Wei;Murch, Kater W.
• 通讯作者: Murch, Kater W.

Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields

• DOI: 10.1109/cvpr52688.2022.00541
• 发表时间: 2021-12
• 期刊: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
• 作者: Dor Verbin;Peter Hedman;B. Mildenhall;Todd E. Zickler;J. Barron;Pratul P. Srinivasan

Topogivity: A Machine-Learned Chemical Rule for Discovering Topological Materials

• DOI: 10.1021/acs.nanolett.2c03307
• 发表时间: 2022-02
• 期刊: Nano letters
• 影响因子: 10.8
• 作者: Andrew Ma;Yang Zhang;Thomas Christensen;H. Po;Li Jing;L. Fu;M. Soljavci'c

Flow-based sampling for fermionic lattice field theories

• DOI: 10.1103/physrevd.104.114507
• 发表时间: 2021-06
• 期刊: ArXiv
• 作者: M. S. Albergo;G. Kanwar;S. Racanière;Danilo Jimenez Rezende;Julian M. Urban;D. Boyda;Kyle Cranmer;D. Hackett;P. Shanahan