Topic: artificial intelligence

CAREER: Numerically Literate AI via the Large Number Model and Foundational Data Curation Methods

Many modern Artificial Intelligence (AI) models can produce meaningful text, but they often fail on complex structural and numerical data involving different units, formulas, information describing the data (i.e., metadata), and hierarchies. These failures are especially concerning in areas such as medicine, finance, defense, and space, where even small quantitative mistakes can lead to misleading conclusions and significant negative consequences. This project aims to address this problem by developing a new AI model focused on accurate comprehension of complex numerical and structured data rather than natural language text. The project will help make scientific knowledge more transparent and accessible, while also supporting education through new teaching materials, student research opportunities, and outreach activities that engage learners in data reasoning. By improving the ability of AI to work correctly with complex numerical and structured data, the project advances the progress of science, supports health and welfare, and strengthens the nation’s capacity for trustworthy data-driven discovery and decision-making. The project develops the Large Number Model (LNM), a hybrid neural-symbolic model for reliable reasoning over numbers, units, formulas, and complex tabular data. The research includes three main activities: creating scalable methods to extract numerical and structured information from documents, designing model architectures that represent quantities and two-dimensional tabular structures more effectively than text-only systems, and incorporating symbolic validation to check algebraic, dimensional, and semantic consistency. The project will also develop methods for combining quantitative evidence across multiple sources and will evaluate the resulting system through controlled experiments, robustness tests, and benchmark datasets drawn from scientific and medical domains. The expected contribution is a new foundation for AI systems that are more accurate, interpretable, dependable, and compatible with the full data cycle when working with complex numerical and structured knowledge. This, in turn, is expected to maximize the utility of information resources. 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.

CAREER: OPENALIGN: Towards Open-World Preference Alignment for Large Language Models

As artificial intelligence (AI) systems are increasingly deployed in critical domains such as healthcare, scientific discovery, and autonomous decision-making, ensuring that foundation AI models such as large language models (LLMs) align with human values and preferences has become essential for their safe and beneficial deployment. However, most existing approaches rely on large amounts of high-quality labeled preference data and assume clean, stable, well-controlled environments. These assumptions are often violated in real-world scenarios, where data may be limited, noisy, or subject to change over time. This project addresses the fundamental problem of aligning LLMs with human preferences under such real or open-world settings. The outcomes are expected to improve the data efficiency and reliability of LLMs in high-impact applications, including medical diagnosis and molecular discovery, while also contributing to education and workforce development through the integration of research and training activities. The project focuses on three key aspects of aligning LLMs with human preferences in open-world settings. First, it develops novel data-efficient preference alignment algorithms that enable LLMs to maintain effective alignment in open-world environments with limited human-annotated preference data. Specifically, when LLMs encounter new tasks or domains, the proposed algorithms can strategically minimize reliance on extensive human or AI annotation while maximizing alignment performance across different low-data scenarios. Secondly, this project aims to enhance the reliability of LLM-based AI systems to maintain safe and robust alignment with human preferences when confronted with unreliable inputs. We will develop algorithmic solutions that can mitigate various forms of data-quality issues (e.g., distribution shifts, label noise, and human value shifts) while preserving alignment performance across different open-world environments. Lastly, this project will demonstrate successful deployment in high-stakes domains, including biochemistry and public health, and will reveal fundamental principles about domain-specific preference alignment while maintaining efficiency and reliability guarantees. This cross-interdisciplinary validation will open new research avenues for LLM alignment with human preferences in interdisciplinary research. Overall, the transition from closed-world to open-world preference alignment represents a fundamental paradigm shift that will provide critical insights about real-world deployment challenges, benefiting the entire AI community. 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.

CAREER: Frontiers of Knowledge in Foundation Models

Foundation models are large artificial intelligence (AI) systems trained on vast amounts of data to perform a wide range of tasks, including answering questions, generating content, and assisting decision-making. These models are increasingly used in areas that affect everyday life, such as healthcare, education, and environmental planning. However, despite their impressive capabilities, they often rely on patterns and correlations in data rather than true causal relationships. This limitation can lead to unreliable or misleading outputs, especially in high-stakes situations. For example, a model may incorrectly assume that one factor causes another simply because they frequently appear together in data. This project addresses this critical challenge by enabling foundation models to better understand and use causal knowledge, which describes how one factor directly influences another in the real world. By improving the ability of these models to reason about cause and effect, the project aims to make them more reliable, transparent, and aligned with human reasoning. The results will support safer and more effective use of AI in important societal domains, strengthen decision-making in complex environments, and contribute to education and workforce development by training students in emerging areas of trustworthy AI. This project develops a systematic framework for understanding, leveraging, editing, and applying causal knowledge in foundation models, organized into four complementary thrusts. The first thrust introduces methods to interpret causal relationships embedded within large-scale models by analyzing internal components that encode causal knowledge across language, vision, and multimodal systems. The second thrust designs approaches to incorporate external causal knowledge into model reasoning, improving performance in tasks such as question answering, causal reasoning, and video understanding. The third thrust establishes techniques for editing causal knowledge within models, enabling targeted updates to specific relationships while preserving overall model performance and consistency. The fourth thrust focuses on empirical evaluation and application of the proposed methods across diverse application domains, including healthcare, materials science, and environmental systems. The project integrates research with education through curriculum development, student mentoring, and outreach activities, and produces open-source tools and resources to support broader adoption. Together, these efforts advance the development of interpretable, controllable, and generalizable foundation models grounded in a causal perspective. 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.

CAREER: Analog Memory Leaks: Mitigating Remanence Side-Channels in Analog Compute-in-Memory Hardware

Next-generation artificial intelligence systems use analog arithmetic to vastly improve efficiency over traditional digital artificial intelligence systems. While the energy efficiency benefits of these analog systems are known, the security vulnerabilities of these new computing architectures are not. Digital systems are built on top of a wealth of defenses and research in cybersecurity, but they do not necessarily apply to analog systems. This project’s novelties are developing guidelines and mitigations for security vulnerabilities in analog computing systems. The project's broader significance and importance are increasing security by addressing vulnerabilities in future artificial intelligence systems, especially in analog computer security. The education and outreach plan addresses an existing and urgent chip design workforce shortage in the United States through increasing the number of digital design classrooms. This project studies next-generation artificial intelligence systems that use novel analog compute-in-memory circuits. Data storage, in normal operation, causes wearout in on-chip storage circuits that change fundamental device parameters like resistance and capacitance. The analog nature of these systems implies that wear, caused by storage (memory), is visible, and can be used to form a side-channel between a user and an adversary. This project will study wearout in a broad spectrum of memory technologies using physics-based simulators and off-the-shelf systems. Ultimately, this project will develop mitigations for current and future artificial intelligence systems. 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.

CAREER: Structure-Aware Learning from Weak Supervision for Knowledge Acquisition

Knowledge acquisition—the ability of artificial intelligence (AI) systems to extract actionable insights from vast amounts of unstructured text—is critical for advancements in healthcare, education, and scientific discovery. While Large Language Models (LLMs) have shown impressive capabilities, their reliability depends heavily on massive, perfectly curated datasets, which are expensive and often unavailable in specialized domains. This CAREER project addresses this bottleneck by developing a new paradigm called “structure-aware weak supervision.” Instead of relying on perfect human annotations, the project enables AI systems to learn autonomously from incomplete, noisy, and ambiguous data by discovering and utilizing underlying semantic structures, such as concept hierarchies and retrieval pathways. By reducing the dependency on expensive labeled data, this research democratizes the development of highly accurate, domain-specific AI tools for resource-constrained environments, such as public health agencies and community organizations. The project also integrates these research outcomes into new undergraduate and graduate curricula, open-source educational toolkits, and targeted K-12 outreach programs designed to broaden participation in computing and teach the next generation how to build reliable, human-centered AI systems. This project proposes a unified framework for learning under weak supervision by bridging unstructured language data with structured, interpretable knowledge representations. The research is organized into three synergistic thrusts. Thrust 1 tackles incomplete supervision by inducing latent ontologies from unlabeled corpora via a novel Spherical Hierarchical Expectation-Maximization (SHEM) algorithm, enabling scalable information extraction and classification without predefined schemas. Thrust 2 addresses noisy supervision by designing a Denoising Retrieval-Augmented Generation (DeRAG) framework. It integrates symbolic reasoning over the induced ontologies with Structure-Aware Contrastive Retrieval (SACRet) to actively filter distractors and reliably ground language model outputs. Thrust 3 tackles ambiguous supervision by modeling complex, multi-faceted human preferences. It introduces a Tree of Reward Models (TreeRM) and Hierarchical Dirichlet Thompson Sampling (HDTS) to capture both shared foundational values (e.g., safety, factuality) and personalized user preferences (e.g., tone), ensuring robust AI alignment. Together, these contributions advance the theoretical foundations and practical methodologies of knowledge-centric AI, creating systems that autonomously construct knowledge, dynamically adapt to supervision gaps, and reliably align with hierarchical human values. 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.

CAREER: A Safety-Aware Learning Framework for Identifying and Mitigating Risks in Human-LLM Interactions in Healthcare

Large language models are increasingly used in healthcare applications such as virtual assistants and decision support tools, offering new opportunities to improve access to care and patient outcomes. However, these systems can introduce new kinds of risks that arise not from the model alone, but from how people interact with it. For example, patients may rely too heavily on automated advice, receive responses shaped by harmful preconceptions or be unintentionally influenced toward unsafe decisions. These risks are especially concerning in sensitive settings such as mental health and addiction recovery, where errors can have serious consequences. This project addresses these challenges by developing new methods to make interactions between people and artificial intelligence systems safer and more trustworthy. The work aims to improve the reliability of healthcare technologies, support safer patient experiences, and contribute to the broader goal of responsible artificial intelligence. Educational activities include developing interdisciplinary coursework and engaging students from diverse backgrounds in research at the intersection of artificial intelligence and health. This project develops a unified, safety-aware learning framework for identifying and mitigating risks in human-large language model interactions in healthcare. The research investigates three integrated thrusts. First, it develops predictive models to detect fundamental and emerging interaction risks, such as overreliance, stereotyping, manipulation, and privacy violations, using supervised and contrastive learning techniques with interpretable outputs. Second, it introduces robust learning methods to mitigate these risks by incorporating user intent, clinical context, and interaction dynamics, including adversarial training and personalized reinforcement learning algorithms. Third, it designs an adaptive, closed-loop method that jointly optimizes risk identification and mitigation through self-supervised and continual learning, enabling generalization to evolving risks over time. The framework is evaluated using realistic digital simulation environments for addiction recovery and mental health support. The expected outcomes include new machine learning methodologies for AI safety, insights into safe deployment of AI in healthcare, and generalizable techniques for trustworthy human-AI interaction in high-stakes domains. 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.

CAREER: Harnessing the Power of Off-Dynamics Reinforcement Learning: Foundations, Algorithms, and Applications

Many important decisions in everyday life, such as choosing effective medical treatments, managing public health responses, or controlling autonomous systems, must be made step by step while learning from limited and imperfect information. Current artificial intelligence methods often require extensive trial-and-error interactions with the real world to learn effective strategies, which is impractical or unsafe in high-stakes settings where mistakes are costly or unethical. This project addresses this challenge by developing new approaches that allow intelligent systems to learn from simulated or indirect environments and reliably transfer that knowledge to real-world situations, even when conditions differ. By enabling safer and more data-efficient decision-making, the project has the potential to improve technologies in healthcare, robotics, and other critical domains, ultimately benefiting public health, economic productivity, and societal well-being. The project will also contribute to education and workforce development by training students at multiple levels, creating accessible learning materials, and conducting outreach activities to broaden participation in artificial intelligence. All resulting software, data resources, and educational materials will be made openly available to maximize their impact. This project develops a comprehensive theoretical and algorithmic framework for off-dynamics reinforcement learning, which studies how to train decision-making agents in a source domain, such as a simulator, and effectively deploy them in a target domain with different and potentially unknown transition dynamics. The research addresses fundamental challenges arising from distributional shifts between training and deployment environments. The work is organized into three main research activities: (1) developing distributionally robust learning methods that ensure reliable performance via minimax optimization over an uncertainty set of transitions when the target domain is unknown, supported by finite-sample performance guarantees; (2) designing algorithms that leverage partial access to target-domain data through data augmentation and cross-domain learning to improve transfer efficiency; and (3) establishing new frameworks for learning under limited interaction and high policy-switching costs, focusing on stability and efficiency in real-world deployment. The proposed methods will be analyzed theoretically to characterize statistical limits and performance guarantees and will be validated empirically on standard reinforcement learning benchmarks and real-world healthcare datasets. All developed algorithms will be released as open-source implementations to support reproducibility and broad adoption. 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.

CAREER: Calibrating Human Trust in Artificial Intelligence through Real-Time Behavioral and Physiological Feedback in Healthcare Decision Making

Artificial intelligence is increasingly used to support decision making in healthcare, especially in time-sensitive settings such as emergency care and intensive care units. However, people do not always rely on these systems appropriately. Some users may place too much trust in incorrect recommendations, while others may ignore useful guidance. These mismatches can affect decision quality and patient safety. This project studies how people interact with artificial intelligence in such settings and explores ways to support more appropriate use. By improving how clinicians interpret and respond to artificial intelligence, the work aims to support safer and more reliable decision making. The project also contributes to education by engaging students in simulation-based learning and providing training opportunities in human-centered artificial intelligence. This project develops a framework to study how trust in artificial intelligence changes over time during decision making. The research combines behavioral data with physiological signals, including eye movements and brain activity, to better understand user responses. First, a mathematical model is developed to represent trust as a changing internal state influenced by task conditions and system performance. Second, machine learning methods are used to estimate this state in real time using data collected from clinicians interacting with simulated clinical scenarios. Third, the project explores interface strategies that provide targeted feedback to help users better align their decisions with the reliability of the system. These approaches are evaluated through controlled simulation studies with clinician participants. The project will generate data, models, and open resources to support future research on human interaction with artificial intelligence. 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.

CAREER: Designing Domain Knowledge-Guided Learning Architectures towards Wireless Network Security: Enabling Effective and Efficient Attacks and Countermeasures

Modern wireless networks have become increasingly complex and densely populated and can create a massive volume of operation data. As a result, extensive efforts from both academia and industry have focused on leveraging artificial intelligence (AI) in wireless network security related tasks such as (i) adversarial inference, (ii) adversarial generation, and (iii) data transformation. This project will focus on exploring new directions for incorporating additional domain knowledge to improve the efficiency of machine learning architecture design. The project's novelties are (i) investigating the wireless-domain knowledge used in mobile network design across different protocol layers and classifying this knowledge based on how it can be deterministically incorporated into learning model design; and (ii) designing specialized learning architectures that translate this domain knowledge into AI-friendly representations to improve both learning efficiency and performance. The project's broader significance and importance are advancing the state of the art in wireless network security, enhancing undergraduate student training opportunities, openly disseminating training materials, and carrying out outreach activities. This project targets three major categories of learning models: (i) typical centralized learning models, (ii) decentralized learning models, and (iii) large language models, and explores the incorporation of wireless-domain knowledge into each class to address different security tasks. Specifically, the project outlines three research thrusts based on both system design and practical evaluations: (i) creating a new training-efficient, wireless-specific learning model as a surrogate model for resource allocation attacks; (ii) developing a new large language model-powered multi-agent framework to enable attacks in cooperative spectrum sensing; (iii) designing a novel client-to-server parameter sharing strategy in federated learning to defend against membership inference attacks. This project will also perform comprehensive evaluations based on real-world wireless experiments to validate and improve the proposed designs. 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.

Center: National Synthesis Center for Organismal Resilience to Environmental Change (NSCORE)

The National Synthesis Center for Organismal Resilience (NSCORE) will inspire, unite, and train organismal biologists and computer scientists in a collaborative effort to leverage vast amounts of real-world data utilizing advanced Artificial Intelligence (AI) techniques to address the growing challenges in understanding and predicting organismal resilience. By integrating across diverse disciplines, NSCORE will transform how scientists study and predict organismal responses to change and make recommendations to improve resilience. Ultimately, its innovative programs and collaborations will empower a new generation of scientists to address one of the most critical and growing challenges of our time. Central to NSCORE’s mission is the education and training of the next generation of computational organismal biologists. To achieve this, NSCORE will implement a hybrid interdisciplinary education program for students and educators, from K-12 to university. At the core of this vertically integrated program will be the NSCORE Nexus, an online infrastructure for broad reach, that will be complemented by in-person events for deeper engagement. Initiatives will include AI-focused Bootcamps for biologists, Virtual K-12 Teacher Institutes, and Train-the-Trainer Programs for university instructors. This manifold menu of programs will emphasize opportunities for everyone, everywhere, reaching underserved urban and rural schools, minority-serving institutions, and underdeveloped regions globally. NSCORE will also engage the public through exhibits, open talks, and online resources, thereby broadening awareness and participation. Organisms respond to complex and dynamic external conditions through a series of interrelated behavioral, physiological, morphological, and molecular mechanisms. Advances in technology and increasingly deeper connections to multiple biological disciplines, including molecular, neuro-, and developmental biology, are revolutionizing the study of organisms in their natural environments. With access to vast and complex biological datasets—from genomes to behaviors—and detailed environmental data from satellite imagery and global weather networks, we are poised to revolutionize our understanding and prediction of organismal resilience. By merging organismal biology with computer science into the field of computational organismal biology, NSCORE will help advance our ability to explain and predict organismal resiliency to change by developing the tools and expertise needed to synthesize these varied datasets and bridge multiple scales using cutting-edge AI techniques. As a global hub for computational organismal biology, NSCORE will facilitate interdisciplinary synthesis through Working Groups and Catalysis Conferences that bring together experts and students. Nucleated at Columbia University, NSCORE will collaborate with academic and research institutions across the greater New York City area, North America, Europe, Asia, and beyond to foster an integrated community of researchers dedicated to achieving its goals. 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.

CAREER: Enabling Single-Step Additive Manufacturing of Ceramics via Laser-Triggered Flash Sintering and Scientific Artificial Intelligence-based Multiscale Modeling

Ceramics possess exceptional resistance to heat, wear, radiation and corrosion, yet their widespread adoption is limited because direct manufacturing complex ceramic parts requires extremely high temperatures and often leads to cracking, defects, and long processing times. This Faculty Early Career Development Program (CAREER) award supports research in additive manufacturing (AM) of high-performance ceramics to enable faster, more reliable production of components used in aerospace, nuclear energy, electronics, and biomedical systems. Current AM methods either rely on multi-step processes that are slow and prone to distortion or single-step methods that generate severe cracking and poor material quality. Research enabled by this award seeks to overcome these limitations by developing a new AM approach that enables rapid, defect-resistant fabrication of complex components. By advancing reliable manufacturing of high-performance ceramics, the award is expected to accelerate innovations in energy efficiency, advanced transportation, and resilient infrastructure, strengthening U.S. technological leadership, economic competitiveness, and national security.    This CAREER award aims to establish the scientific foundation for a transformative single-step ceramic AM process based on laser-triggered flash sintering (LTFS). A central challenge is the lack of fundamental understanding of how coupled laser heating and electric-field stimulation initiate flash sintering, govern densification kinetics, and influence microstructure evolution, defect formation, and process reliability. To address this gap, research is planned to develop an integrated experimental, computational, and data-driven framework. Specifically, the research tasks include (1) design and construct an LTFS-enabled AM testbed with in-situ monitoring for real-time process characterization; (2) investigate flash-sintering initiation, stability, and microstructure evolution through coordinated experiments and multiphysics microscale modeling; (3) establish a multiscale electro-thermal-mechanical modeling framework to quantify how manufacturing parameters influence densification, shrinkage, and resulting material properties; (4) develop a Scientific Artificial Intelligence (Sci-AI) framework that integrates in-situ data with physics-based models to capture process stochasticity, improve predictive accuracy, and enable intelligent process control; and (5) demonstrate manufacturing capability through fabrication and evaluation of complex, high-performance ceramic components. The outcomes are expected to establish quantitative process–structure–property relationships for LTFS-based ceramic manufacturing, enabling defect-controlled fabrication of advanced ceramics and advancing smart, data-driven manufacturing of high-performance materials. 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.

CAREER: Mechanism-Driven Machine Learning for Water and Wastewater Treatment

Clean and reliable water treatment is critical to protect public health. However, most treatment decisions still rely on simple models that do not reflect real‑world conditions. Artificial intelligence (AI) has the potential to help engineers make more informed decisions. This CAREER project will develop AI models that help engineers understand how and why treatment processes work at full-scale. The project will reduce risk, improve reliability, and expand access to advanced tools for communities with limited technical resources. The outcomes of this research will be shared with water utilities and used across many treatment systems. The project will also address a national need for a workforce that can use AI responsibly by integrating data science into environmental engineering education. This project will support safer water systems, prepare future engineers, and show how AI can be used as a tool for scientific discovery. This CAREER project will develop an application‑driven AI framework for modeling engineered environmental systems, using water and wastewater disinfection as a representative, high‑risk unit process. The research will integrate multi‑facility operational and water quality data with hybrid modeling approaches that combine physics‑based process models and machine learning (ML). These approaches will include mechanistic ordinary differential equation models coupled with ML components, physics‑informed neural networks, and embedded neural differential equation formulations that constrain learning using known physical, chemical, and biological relationships. Model development and evaluation will explicitly address challenges common to environmental datasets, including data sparsity, autocorrelation, measurement uncertainty, and site‑specific variability, through time‑aware validation, uncertainty quantification, and risk‑based performance metrics. Mechanistic insights inferred from the models will be tested using a pilot‑scale disinfection system to distinguish true process behavior from artifacts introduced by data collection or modeling practices. The project will also develop protocols for model reuse and adaptation using transfer learning and privacy‑preserving federated learning, enabling models trained on multi‑facility data to be applied in data‑limited systems without sharing raw data. Together, these methods will advance the scientific use of AI in environmental engineering by enabling mechanistically grounded discovery, improving generalizability across real systems, and establishing a foundation for trustworthy, reusable models for infrastructure applications. 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.

CAREER: Understanding nutrient dynamics in American rivers through remote sensing and artificial intelligence

This award supports the study of nutrient runoff and its effects on American river systems. Nitrogen and phosphorus are known to cause harmful algal blooms and other ecological impacts in rivers. However, knowledge about the causes of variability in nutrient loads across river networks remains limited. Through the application of artificial intelligence to data from satellite remote sensing, this project will generate a detailed map of nutrients in rivers across the United States over time. This map will then be analyzed to understand human and natural factors affecting nutrient variability. This research integrates with education for high school, undergraduate, and graduate students. Project outcomes will support water management, ecosystem protection, and public health. This project will pursue three objectives. (1) A novel modeling framework that integrates remote sensing and deep learning will be developed. This framework will be used to estimate daily, reach-level total phosphorus and total nitrogen concentration in American rivers over the past five decades. (2) Major drivers and controlling mechanisms for nutrient variability across space and time will be identified. (3) Relationships between riverine nutrients and harmful algal blooms across various settings will be quantified. Outcomes of these analyses will reveal spatial patterns of nutrient sensitivity and eutrophication risk. 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.

CAREER: Towards Semantic-Centric Wireless Foundations for Swarm AI

Artificial intelligence (AI) is rapidly expanding from centralized computing infrastructures into the physical world, where networks of distributed devices must sense, reason, and act together in dynamic and uncertain environments. This transformation gives rise to swarm AI, an emerging form of intelligent infrastructure that supports applications such as disaster response, environmental monitoring, precision agriculture, and autonomous mobility. Unlike traditional systems that rely on stable wired connections, swarm intelligence operates over wireless links that are intermittent, noisy, and resource constrained. Communication, therefore, becomes a central bottleneck that limits reliability, efficiency, and coordination. This project establishes a new wireless foundation for swarm AI by prioritizing the meaning and task relevance of transmitted information rather than raw bit accuracy alone. By strengthening how distributed agents share mission-critical information under challenging wireless conditions, the research enhances the resilience, scalability, and interoperability of next-generation intelligent systems. The project integrates research and education through curriculum development in communication-aware AI, hands-on mentoring of undergraduate and graduate students, outreach to K-12 learners, and open dissemination of research outcomes. These activities broaden participation in advanced wireless and intelligent systems research and contribute to workforce development in emerging communication and intelligent system technologies. The project addresses a fundamental gap between AI systems that assume ideal connectivity and wireless communication protocols that optimize bit-level fidelity without accounting for task intent. The scientific problem is how to design wireless architectures that are aware of semantic content, resilient to time-varying channel impairments, and adaptive to heterogeneous device capabilities in swarm settings. The research establishes a semantic-centric communication framework organized into three integrated thrusts: (i) robust semantic transceiver principles that identify and protect task-relevant information against dynamic wireless distortion, ensuring reliable semantic delivery under feature-dependent channel impairments; (ii) swarm-aware radio orchestration strategies that align spectrum allocation and scheduling with collective task objectives through utility-driven coordination; and (iii) heterogeneity-aware collaborative reasoning architectures that enable progressive semantic compression and resource-adaptive inference across devices with diverse sensing, computing, and communication constraints. The research combines theoretical analysis, algorithm design, and experimental validation to advance communication-aware intelligence across layers of the wireless stack. Together, these advances provide a principled foundation for building resilient, scalable, and interoperable swarm AI infrastructures. 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.

CAREER: Inferring Specifications for AI by Modeling Humans

Artificial intelligence (AI) systems increasingly influence both high stakes and everyday decisions across many sectors of the economy. These systems, however, are not developed in isolation. Instead, they depend on people to provide instructions that describe what the system should do and what outcomes it should avoid. These instructions can take many forms. They may be written explicitly by domain experts or learned from data such as human preferences over outcomes. However, providing clear and reliable instructions for intelligent systems is difficult even for relatively narrow applications. Instructions can be too rigid, too vague, or simply incorrect, and any of these problems can cause systems to behave in unintended ways. These failures occur because instructions are created by people, and human reasoning is shaped by limited information, context, and common cognitive mistakes. As AI becomes more widespread, improving how systems interpret human intent will be essential for safety and reliability. This project addresses that challenge by studying how people communicate goals to machines and by designing AI systems that can interpret imperfect instructions by reasoning about the intent behind them. The expected outcomes include safer decision-making technologies and new tools that help organizations deploy AI more effectively. This project develops computational foundations for learning AI specifications from imperfect human input. The research integrates reinforcement learning, Bayesian inference, and computational cognitive modeling with empirical studies of human decision making to better characterize how people communicate goals and where specification errors arise. The work is organized around three research thrusts. The first thrust, Modeling and Inferring AI Specifications, develops probabilistic models of human reasoning that capture systematic specification errors and uses these models to enable AI systems to infer more accurate goals from flawed instructions. The second thrust, Richer Inputs and Representations, expands how AI systems learn from people by incorporating different forms of input such as preferences, demonstrations, explanations, gestures, and structured debate. New algorithms and elicitation interfaces will integrate these signals and resolve inconsistencies across modalities. The third thrust, Personalization and Governance, develops methods for learning multiple reward models that reflect differences in human preferences, enabling scalable personalization and avoiding one-size-fits-all objectives. In parallel, the project will develop educational programs that prepare students to design and govern AI systems. These activities include revising an undergraduate AI course to emphasize human decision-making in the design of AI systems, creating a graduate course on AI policy and governance, and expanding the AI Policy Summer School to help build a national workforce that is fluent in both AI technology and public policy. 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.

Postdoctoral Fellowship: MSPRF: Traveling Water Waves and Stellar Collapse: Nonlinear Analysis of Free Boundary Problems in Fluid Mechanics

This award is made as part of the FY 2026 Mathematical Sciences Postdoctoral Research Fellowships Program. Each of the fellowships supports a research and training project at a host institution in the mathematical sciences, including applications to other disciplines such as Artificial Intelligence and Quantum Information Science, under the mentorship of a sponsoring scientist. The title of the project for this fellowship to Noah Stevenson is “Traveling Water Waves and Stellar Collapse: Nonlinear Analysis of Free Boundary Problems in Fluid Mechanics”. The host institution for the fellowship is ETH Zurich and the sponsoring scientist is Mikaela Iacobelli. 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.

Postdoctoral Fellowship: MSPRF: Homotopy Invariants of Orbifolds

This award is made as part of the FY 2026 Mathematical Sciences Postdoctoral Research Fellowships Program. Each of the fellowships supports a research and training project at a host institution in the mathematical sciences, including applications to other disciplines such as Artificial Intelligence and Quantum Information Science, under the mentorship of a sponsoring scientist. The title of the project for this fellowship to Maxine Calle is “Homotopy Invariants of Orbifolds”. The host institution for the fellowship is Brown University and the sponsoring scientist is Thomas G. Goodwillie. 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.

Postdoctoral Fellowship: MSPRF: Coupled Multispecies Systems

This award is made as part of the FY 2026 Mathematical Sciences Postdoctoral Research Fellowships Program. Each of the fellowships supports a research and training project at a host institution in the mathematical sciences, including applications to other disciplines such as Artificial Intelligence and Quantum Information Science, under the mentorship of a sponsoring scientist. The title of the project for this fellowship to Lauren Conger is “Coupled Multispecies Systems”. The host institution for the fellowship is Stanford University and the sponsoring scientist is Lenya Ryzhik. 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.

Postdoctoral Fellowship: MSPRF: Combinatorial Models and Scaling Limits in Liouville Quantum Gravity and KPZ

This award is made as part of the FY 2026 Mathematical Sciences Postdoctoral Research Fellowships Program. Each of the fellowships supports a research and training project at a host institution in the mathematical sciences, including applications to other disciplines such as Artificial Intelligence and Quantum Information Science, under the mentorship of a sponsoring scientist. The title of the project for this fellowship to Andres Contreras Hip is “Combinatorial Models and Scaling Limits in Liouville Quantum Gravity and KPZ”. The host institution for the fellowship is Columbia University and the sponsoring scientist is Ivan Corwin. 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.

Postdoctoral Fellowship: MSPRF: Foundations of Positive Geometry

This award is made as part of the FY 2026 Mathematical Sciences Postdoctoral Research Fellowships Program. Each of the fellowships supports a research and training project at a host institution in the mathematical sciences, including applications to other disciplines such as Artificial Intelligence and Quantum Information Science, under the mentorship of a sponsoring scientist. The title of the project for this fellowship to Elizabeth Pratt is “Foundations of Positive Geometry”. The host institution for the fellowship is Princeton University and the sponsoring scientist is June Huh. 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.

Postdoctoral Fellowship: MSPRF: Newton Polygons and Higher Quasi-F-Injective Singularities

This award is made as part of the FY 2026 Mathematical Sciences Postdoctoral Research Fellowships Program. Each of the fellowships supports a research and training project at a host institution in the mathematical sciences, including applications to other disciplines such as Artificial Intelligence and Quantum Information Science, under the mentorship of a sponsoring scientist. The title of the project for this fellowship to Jack Garzela is “Newton Polygons and Higher Quasi-F-Injective Singularities”. The host institution for the fellowship is the University of Illinois at Chicago and the sponsoring scientist is Kevin Tucker. 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.

Postdoctoral Fellowship: MSPRF: Theory and Algorithms for Scalable Decision-Making under Uncertainty

This award is made as part of the FY 2026 Mathematical Sciences Postdoctoral Research Fellowships Program. Each of the fellowships supports a research and training project at a host institution in the mathematical sciences, including applications to other disciplines such as Artificial Intelligence and Quantum Information Science, under the mentorship of a sponsoring scientist. The title of the project for this fellowship to Graham Pash is “Theory and Algorithms for Scalable Decision-Making under Uncertainty”. The host institution for the fellowship is the Massachusetts Institute of Technology and the sponsoring scientist is Youssef Marzouk. 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.

REU Site: Microbiology at the host-pathogen interface

This REU Site award to The University of Iowa, located in Iowa City, IA, will support the training of 10 students for 10 weeks during the summers of 2027-2029. The goals of the program are to generate high-impact discoveries about the fundamental biology of host-microbe interactions and to train the next generation of microbial scientists. Achieving these goals is important for strengthening our bioeconomy, enhancing agriculture and mitigating the threats posed by pandemics and antibiotic resistance. The students will learn to design, conduct and interpret microbiology experiments; many will have the opportunity to present their findings at scientific conferences. Assessment of the program will use a version of the Undergraduate Research Student Self-assessment, a validated tool for measuring student learning gains. In addition, students will be tracked after the program to determine their career paths. Students will apply to the REU site using NSF ETAP (Education and Training Application: https://etap.nsf.gov). The training students will receive is aligned with NSF priorities in Artificial Intelligence and Biotechnology. The focus of the program is host interactions with bacteria, viruses, and parasites. Each student will conduct an independent laboratory research project under the joint guidance of a faculty and a graduate student or postdoctoral co-mentor from the Department of Microbiology and Immunology. Participants will be instructed in communicating their research in short talks, a written report, and a campus-wide poster session. Participants will attend workshops and seminars to broaden their understanding of microbiology and learn how to use AI tools to accelerate discovery. Additional professional development activities will cover graduate school, career options, and responsible conduct of research. Applications will include a form, a personal statement, information on career goals, research interests, two letters of recommendation, and college transcripts. Prior research experience is not required. Students will be selected by the program directors based on their fit for the program’s objectives and potential for outstanding careers involving microbiology research. More information about the program is available by visiting https://microbiology.medicine.uiowa.edu/undergraduate-education/research-opportunities/summer-undergraduate-research, or by contacting the PI (Dr. David Weiss at david-weiss@uiowa.edu) or the co-PI (Dr. Gina McGrane at regina-mcgrane@uiowa.edu). 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.

Collaborative Research: ARTS: Training A New Generation of Systematists and Completing the Monograph of Protium (Burseraceae) a Hyperdiverse Tropical Tree Clade

Newly developed and established methods will be integrated to complete a comprehensive study of the genus Protium (in the plant family Burseraceae, which includes frankincense and myrrh), revising the taxonomy of all 183 currently published species and describing all 52 remaining unpublished species. Protium represents one of the most important tree genera in the Americas. In Amazonia, Protium includes more than 125 species, and it serves as an excellent model system to understand the processes underlying the origins and maintenance of tropical tree diversity. New field collections in six countries will complement existing informative specimens of Protium for morphological and molecular studies that will identify and characterize all the species. Cutting-edge genomic approaches will be used to understand how each species is related to each other and the evolutionary history of the diversification in the group. New taxonomic tools such as leaf architecture (to find fingerprint-like leaf vein patterns) and near-infrared (NIR) spectral signatures, which quantifies light reflectance of dried leaves, will be integrated with Artificial Intelligence (AI) tools to develop an interactive, image-driven, multi-access electronic key to all species that will be available online. The wealth of knowledge about species and traits that will be integrated by this project will help make tropical forests more understood and better protected. Moreover, this work will be conducted as part of a training program for tropical plant systematics. Training modern taxonomists is one of the most urgent priorities for tropical biology and conservation. New systematists will be trained who can build on foundations of fieldwork and a solid background of traditional plant anatomy and morphology, but also become experts in genomics, bioinformatics, web-based interactive keys, and NIR spectroscopy, so that they can become leaders in tropical botany for the rest of this century. This project will train two PhD students and one postdoctoral scholar and give several undergraduate interns the foundation to enter the field– a small but mighty investment in the future of tropical botany. This project will contribute both a taxonomic monograph and a complete phylogeny of Protium (Burseraceae), one of the largest and most important clades of tropical trees. New field work in poorly documented regions of Panama, Ecuador, Colombia, Peru, Brazil, and Guyana will be conducted to collect silica-dried leaf material and highly informative herbarium specimens and integrated with previously collected material. The project will use hybrid enrichment sequencing (Hyb-Seq), an approach using targeted sequence capture strategies with probes designed to capture low copy nuclear loci. Use of Hyb-Seq will produce a fossil-calibrated comprehensive phylogeny of the genus Protium. Leveraging this information with data on relative abundance and functional traits of Protium will lead to an increased understanding of the factors that influence commonness and rarity in tropical forests, a critical focus of conservation strategies. A growing track record of success with Protium as a model organism in developing and applying leaf architecture, NIR spectrometry, interactive keys, GIS mapping, and other resources for effective identification and characterization of trees should have profound implications for sustainable forest management and conservation. 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.

REU Site: Engaging Undergraduates in Interdisciplinary Evolutionary Science

This REU Site award to the University of Iowa, located in Iowa City, IA, will support the training of 10 students for 10 weeks during the summers of 2027-2029. It is anticipated that a total of 30 students, primarily from schools with limited research opportunities, will be trained in the program. Students will be trained by faculty mentors how research is conducted in evolutionary science, will participate directly in that research, and will learn how to communicate science to public audiences. Potential career paths that exist for evolutionary scientists and for the application of evolutionary science will be discussed and explored. Many participants will present the results of their work at scientific conferences. Required formal mentor training of faculty mentors will have a lasting effect on their future mentoring efforts. Assessment of the program will be done through online surveys. Students will be tracked after the program to determine their career paths. Students should apply to the REU site using NSF ETAP (Education and Training Application: https://etap.nsf.gov). The training students will receive is aligned with NSF priorities in Artificial Intelligence and Biotechnology. The focus of this REU is evolutionary science, with students conducting research projects across several disciplines. Scientist-mentors in seven academic departments will offer research projects that span a wide range of topics, including evolutionary ecology, behavior, paleontology, genomics, bioinformatics, evolution of infectious disease, and developmental biology. Students will work on evolution-themed projects in one of these specific areas and will also work as a cohort to make broad connections among disciplines. As part of the program, students will receive training in ethical and responsible conduct in research, participate in career workshops, make formal research presentations based on their work, and create an interactive digital research poster. All students will be encouraged to participate in a series of three optional short courses in computational methods and phylogenetics. Students will be selected by program directors based on previous academic performance, enthusiasm for conducting research, interest in specific faculty research projects, and potential for future success in a research-related career. Students who have limited research opportunities at their home institution will be especially encouraged to apply. More information about the program is available by visiting https://biology.uiowa.edu/reu, or by contacting the PI (Dr. John Logsdon at john-logsdon@uiowa.edu) or the co-PI (Dr. Andrew Kitchen at andrew-kitchen@uiowa.edu). 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.

Travel: NSF Student Travel Grant for the Programming Languages Mentoring Workshop at ACM SIGPLAN Symposium on Principles of Programming Languages, 2027-2029

A Programming Languages Mentoring Workshop (PLMW) is organized as part of the ACM SIGPLAN Symposium on Principles of Programming Languages (POPL), the flagship conference in the field of programming language theory, and one of the premier conferences in all of computer science. The 2027 conference will be held in Mexico City, Mexico. Many POPL papers directly address administration priorities in artificial intelligence (AI) and Quantum computing. The impact of the award relates to providing opportunities for students to receive mentoring from leading researchers, and building the next generation of researchers and knowledgeable practitioners in programming languages. The award's broader significance and importance include building international community, lasting professional connections to design novel programming languages and implement tools, and enhancing education of US students. The workshop also provides students exposure to and multiple opportunities to interact with leading-edge research and researchers. By supporting students, the workshop thus imparts training to the next generation of researchers in programming languages and systems and contributes to building a national workforce in these topics. 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.

CICI: IPAAI: XAI-Driven Data Forensics for Nuclear Edge LLMs: Provenance, Detection, and Recovery

Nuclear reactors supply a large share of reliable electricity, and reactor operators increasingly turn to large language models to help interpret procedures, sensor logs, and alarm messages in real time. Running these models on-site, close to the reactor, keeps sensitive data in-house and speeds decision-making, but it also opens new ways for these systems to fail or be attacked. Poisoned or low-quality training data, foundation models that carry hidden backdoors, and manipulated prompts can each push a model toward confident but incorrect recommendations, a serious hazard that could threaten safety or force a costly shutdown. This project builds an explainable artificial intelligence framework that secures the entire workflow of nuclear-domain large language models. The team develops interpretability methods that trace a model's outputs back to specific inputs and internal representations, then uses those signals to detect and remove poisoned or low-quality training data through machine unlearning. To confirm a model's origin, hardware-based fingerprinting techniques are employed. At inference time, layered defenses screen inputs, flag risky prompts, and validate outputs before any recommendation reaches an operator. The methods are validated with real operational data from a research nuclear reactor. By establishing a unified approach to data integrity, model provenance, and output reliability, the project advances the safe deployment of large language models in critical infrastructure, protecting public welfare and national security. Openly shared tools, datasets, and a fingerprint library, along with new university courses and K–12 outreach, will help build a security-aware technical workforce. 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.

REU Site: Cybersecurity Research of Unmanned Aerial Vehicles

This award supports a Research Experience for Undergraduates (REU) Site at Embry-Riddle Aeronautical University (ERAU), Daytona Beach, Florida. Each summer, ten motivated undergraduates from universities across the United States participate in a rigorous 10-week Unmanned Aerial Vehicles (UAV) cybersecurity research program. The program combines faculty-mentored research with professional development activities designed to prepare students for careers in cybersecurity and/or graduate school. The project's novelties include vulnerability assessment and the development and testing of new cyber-resilient algorithms to safeguard UAVs. More broadly, the Site advances UAV security research while equipping students with essential research skills through structured faculty mentorships. The training is designed to prepare students with skills and tools to thrive in graduate school and future careers in the cybersecurity field. The REU Site focuses on enhancing UAV security through faculty-guided research projects. Participants investigate cyber-resilient UAV operation and simulations; secure navigation including Global Positioning System (GPS) spoofing detection and mitigation; and AI-based anomaly and intrusion detection. The research process includes systematic literature reviews, hypothesis development, testbed creation for data collection, data processing, technical seminars, and workshops in Artificial Intelligence (AI) and cybersecurity. Throughout the program, students gain valuable experience presenting their findings, with mid-term results showcased at the Daytona Museum of Arts and Sciences (MOAS) and final outcomes at the ERAU Summer Symposium. The key objectives of this Site are: (1) Increase the number of high-quality cybersecurity professionals, (2) Broaden STEM participation by enhancing recruitment efforts for veterans, individuals with diverse socioeconomical backgrounds and under-resources institutions, (3) Inspire and empower undergraduates to confidently pursue graduate degrees, and (4) Provide undergraduates with professional skills for their future careers. By leveraging Embry-Riddle’s state-of-the-art facilities, research labs, and faculty expertise, the program cultivates interest in cybersecurity and develops the research skills of undergraduate students, contributing to cybersecurity education, training, and workforce development. 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.

Integrating Data Science and AI-Aware Learning into Introductory Chemistry and Biology Pathways at a Two-Year College

This project aims to serve the national interest by providing two-year college students with early exposure to data, computing, and AI-related skills. Many STEM fields depend on data and artificial intelligence (AI), yet many students may not get early chances to build these skills. This Track 1 ITYC project will design materials for chemistry and biology classrooms that help STEM and non-STEM students engage with analysis and interpretation of real data that connects to everyday life. The project will also document the impact of these experiences on students’ interest in STEM, their understanding of the role of data and AI in scientific work, and their perspectives on future academic and workforce pathways. Partnerships with regional research institutions will provide authentic data sets and provide input on current workforce applications. The project will design, test and study a set of interdisciplinary learning modules for introductory chemistry and biology courses at a two-year college. These modules will be guided by the Interdisciplinary Data Investigation framework and immerse students in data analysis, visualization and interpretation using real scientific and societal contexts. The project has two main goals: 1) To develop a strategic and sustainable approach for embedding data- and AI-aware learning across early STEM pathways, creating a model that strengthens faculty practice, supports institutional capacity building, and advances data-rich undergraduate STEM education, and 2) To cultivate students' understanding and curiosity about the data-driven and AI enabled nature of contemporary science, fostering awareness of how discovery, innovation, and interdisciplinary collaboration define today’s STEM enterprise. This research will use a mixed-methods approach, utilizing surveys, assessments, interviews and classroom observations to study student learning, engagement and course outcomes. Partnerships with research and health organizations will provide real data sets and relevant applications. The projects will produce practical strategies and a scalable model that other institutions can use to bring data-driven learning into STEM education. The NSF IUSE: Innovation in Two-Year College STEM Education (ITYC) Program seeks to accelerate and advance knowledge about the impact of emerging and evidence-based practices in undergraduate STEM education at two-year colleges. 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.

Collaborative Research: Elements: Analysis Collaborative Ecosystem for High Energy Physics (ACEHEP)

The Large Hadron Collider (LHC) is the largest international scientific project partially supported by the United States. The goal of scientists at the LHC is to understand how matter behaves under the extreme conditions thought to have existed in the very early universe. This understanding informs the current scientific understanding of the evolution of the universe from the distant past to the present day. The behavior of matter under such conditions is investigated by analyzing the data from billions of high-energy particle collisions and comparing the results of these analyses to predictions made by theoretical physicists. LHC data are analyzed using many different software systems, called analysis frameworks, written by different research groups from around the world. While these frameworks are individually powerful tools, it is difficult for a scientist to form a global understanding of all the data from the LHC. This project will create a unifying software system, based upon a new programming language that allows scientists to describe their analysis ideas in a natural way. The system will advance science by making it considerably easier to run multiple analyses of data in an organized fashion, thereby yielding results that are impossible to obtain otherwise. The new language, called Analysis Description Language (ADL), reflects the way scientists at the LHC think about the analysis of data. Crucially, ADL does so in a way that is independent of the analysis frameworks in which the analyses are ultimately run. ADL also serves as a reliable intermediary between high energy physics analyses and natural language interfaces through Large Language Models and artificial intelligence-based assistants. For this reason, the proposed software system opens the way for students and science teachers to participate in the exciting research at the LHC by lowering the barrier to exploring the scientific principles and reasoning that underpin the research without requiring training in complex analyses frameworks. Many data analyses at the Large Hadron Collider (LHC) at the European Organization for Nuclear Research (CERN) have yielded small deviations from the predictions of the Standard Model, the best current theory of particle physics. To further investigate these deviations, a potential discovery at the LHC requires global assessments of the results of multiple analyses. To date, no cyberinfrastructure exists to make such a task routine. The project addresses this need through cyberinfrastructure called the Analysis Collaborative Ecosystem for High-Energy Physics (ACEHEP). This cyberinfrastructure builds on the Analysis Description Language (ADL), a declarative, domain-specific language that expresses the physics algorithms of analyses in a transparent, structured way, decoupled from the analysis (software) frameworks in which analyses are executed. The cyberinfrastructure will enable the meta-analysis of large collections of LHC analyses, enable analysis queries, and facilitate the coordinated execution of multiple analyses. The latter is possible because the analysis descriptions are executable. This project will complete interfaces between ADL and three modern high-energy physics (HEP) analysis frameworks (TIMBER, RDataFrame, and Coffea); complete an interpreter for the language; build tools to query databases of analyses written in ADL; and build a fine-tuned large-language model that can translate between ADL and English, assisting in analysis queries. A key component of this project is to bring to production level the compiler infrastructure required to build these interfaces and the interpreter. This award by the Office of Advanced Cyberinfrastructure is jointly supported by the Physics at the Information Frontier Program in the Physics Section of the Directorate for Mathematical and Physical Sciences. 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.

Collaborative Research: Elements: Analysis Collaborative Ecosystem for High Energy Physics (ACEHEP)

The Large Hadron Collider (LHC) is the largest international scientific project partially supported by the United States. The goal of scientists at the LHC is to understand how matter behaves under the extreme conditions thought to have existed in the very early universe. This understanding informs the current scientific understanding of the evolution of the universe from the distant past to the present day. The behavior of matter under such conditions is investigated by analyzing the data from billions of high-energy particle collisions and comparing the results of these analyses to predictions made by theoretical physicists. LHC data are analyzed using many different software systems, called analysis frameworks, written by different research groups from around the world. While these frameworks are individually powerful tools, it is difficult for a scientist to form a global understanding of all the data from the LHC. This project will create a unifying software system, based upon a new programming language that allows scientists to describe their analysis ideas in a natural way. The system will advance science by making it considerably easier to run multiple analyses of data in an organized fashion, thereby yielding results that are impossible to obtain otherwise. The new language, called Analysis Description Language (ADL), reflects the way scientists at the LHC think about the analysis of data. Crucially, ADL does so in a way that is independent of the analysis frameworks in which the analyses are ultimately run. ADL also serves as a reliable intermediary between high energy physics analyses and natural language interfaces through Large Language Models and artificial intelligence-based assistants. For this reason, the proposed software system opens the way for students and science teachers to participate in the exciting research at the LHC by lowering the barrier to exploring the scientific principles and reasoning that underpin the research without requiring training in complex analyses frameworks. Many data analyses at the Large Hadron Collider (LHC) at the European Organization for Nuclear Research (CERN) have yielded small deviations from the predictions of the Standard Model, the best current theory of particle physics. To further investigate these deviations, a potential discovery at the LHC requires global assessments of the results of multiple analyses. To date, no cyberinfrastructure exists to make such a task routine. The project addresses this need through cyberinfrastructure called the Analysis Collaborative Ecosystem for High-Energy Physics (ACEHEP). This cyberinfrastructure builds on the Analysis Description Language (ADL), a declarative, domain-specific language that expresses the physics algorithms of analyses in a transparent, structured way, decoupled from the analysis (software) frameworks in which analyses are executed. The cyberinfrastructure will enable the meta-analysis of large collections of LHC analyses, enable analysis queries, and facilitate the coordinated execution of multiple analyses. The latter is possible because the analysis descriptions are executable. This project will complete interfaces between ADL and three modern high-energy physics (HEP) analysis frameworks (TIMBER, RDataFrame, and Coffea); complete an interpreter for the language; build tools to query databases of analyses written in ADL; and build a fine-tuned large-language model that can translate between ADL and English, assisting in analysis queries. A key component of this project is to bring to production level the compiler infrastructure required to build these interfaces and the interpreter. This award by the Office of Advanced Cyberinfrastructure is jointly supported by the Physics at the Information Frontier Program in the Physics Section of the Directorate for Mathematical and Physical Sciences. 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.

RET Site: Research Experiences for Teachers in Cybersecurity and Artificial Intelligence

Computing and cybersecurity skills are increasingly essential across nearly every sector of the modern economy, yet many students in Mississippi and similar rural states have limited exposure to these fields, particularly in schools with fewer resources. Teachers play a pivotal role in shaping student interest in science and technology careers, but many lack the research experience needed to bring authentic computing content into their classrooms. This project addresses that gap by immersing high school and community college educators in hands-on cybersecurity research at Mississippi State University, enabling them to return to their classrooms with deeper knowledge, stronger confidence, and ready-to-use instructional materials. By strengthening teacher expertise in computing and cybersecurity, the project generates a multiplier effect that reaches hundreds of students annually, expanding access to computing pathways in communities. The project also supports workforce development by preparing a pipeline of students who are better equipped for careers in computing, national security, and related fields that are critical to the nation's economic strength and security. This award supports a Research Experiences for Teachers (RET) site in which thirty educators participate over three years in a six-week summer research program at Mississippi State University. Participants are embedded in faculty-led research groups focused on areas such as machine learning for radio frequency interference detection, algorithmic resilience in wireless networks, software reverse engineering, adversarial attacks on robot vision systems, and cybersecurity of electric vehicle charging infrastructure. The first week provides foundational training in computing concepts, research ethics, and laboratory practices, after which participants engage directly in ongoing research projects under faculty mentorship. Beginning in the fourth week, teachers develop standards-aligned curricular modules that connect their research experiences to mathematics and computing courses at the secondary and community college level. Academic-year follow-up activities include faculty classroom visits, virtual mentoring, peer collaboration through a shared digital platform, and presentations at state and regional teacher conferences. Program effectiveness is assessed through a mixed-methods evaluation framework examining teacher self-efficacy, curricular integration, and long-term changes in classroom practice. The project is led by faculty in the Department of Computer Science and Engineering at Mississippi State University, in partnership with the Starkville Oktibbeha Consolidated School District, Hinds Community College, and others. 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.

REU Site: Undergraduate Research Experience in Edge Intelligence

This project establishes a Research Experiences for Undergraduates (REU) site at the University of Nevada, Reno, to provide undergraduate students nationwide with immersive research experiences in the field of edge intelligence. Edge intelligence, where data collection, processing, and analysis are performed close to the data source rather than on centralized cloud-based systems, is a rapidly advancing paradigm critical for applications such as autonomous vehicles, smart cities, industrial automation, and healthcare systems. However, deploying artificial intelligence (AI) on resource-limited edge devices presents substantial challenges in accuracy, efficiency, and robustness, with significant implications for public health, national security, and economic competitiveness. By engaging students in hands-on research and equipping them with the skills to address these challenges, this project aims to develop a highly skilled STEM workforce capable of advancing practical AI technologies. The program will foster sustained interest in research and encourage participants to pursue graduate studies and careers in STEM. Through the development of accurate, efficient, and trustworthy AI systems, the project will contribute to technological innovation, strengthen public trust in AI, and support societal well-being and national security. This REU site develops undergraduate research capacity in edge intelligence by leveraging the interdisciplinary expertise of University of Nevada, Reno faculty in AI, robotics, and cyber-physical systems. The research is structured around three core thrusts in edge-based machine learning (ML) systems: (1) algorithms and architectures for resource-constrained devices, (2) scalable and robust learning in heterogeneous edge environments, and (3) resource management for energy-efficient and sustainable operation. Representative projects include accelerating federated learning in heterogeneous wireless networks, developing memory-efficient training of large language models for edge devices, designing decentralized federated learning with workload balancing, federated fine-tuning of vision models for wildfire detection, and collaborative multi-robot deep reinforcement learning. Guided by experienced faculty and graduate student mentors, students will engage in cutting-edge research and contribute to the development of ML systems for real-world applications such as smart Internet-of-Things devices, autonomous vehicles, and robotics. The program is complemented by training in research methodology, scientific communication, and career development, preparing participants for graduate study and careers in STEM fields. 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.

Collaborative Research: Frameworks: ML4GW, a machine learning ecosystem for gravitational wave data analysis

Since the first direct observation of gravitational waves in 2015, a new field of astronomy has fundamentally changed how the universe can be explored. A gravitational wave is a ripple in spacetime produced when two extremely dense objects, such as black holes or neutron stars, collide. The detectors that observe these signals now record hundreds of such events per year, and the rate is expected to grow to several events per day within the next few years. Sifting through this flood of data to find the rare astrophysical signals buried in detector noise, and then alerting telescopes around the world quickly enough to capture the light produced by a collision, is one of the most demanding computational problems in modern science. This project develops an open, community machine learning framework that makes gravitational-wave discovery faster, more accurate, and more accessible. The framework will reduce the time between the crossing of a gravitational wave through the detectors and a public astronomical alert to less than a second, make it possible to detect events that traditional analysis methods miss, and lower the computing cost of these analyses by orders of magnitude. The project trains the next generation of scientists, including high-school students, undergraduates, graduate students, and researchers at smaller institutions, by sharing open code, open data, open trained models, and open lessons. It also strengthens shared national computing infrastructure that benefits not just gravitational-wave science but also particle physics, neutrino astronomy, and time-domain astronomy more broadly, advancing the national interest by accelerating discovery and broadening participation in science. The project develops ml4gw, an open-source PyTorch-based machine learning (ML) framework for gravitational wave (GW) data analysis. ml4gw provides Graphics Processing Unit (GPU)-accelerated implementations of the data ingestion, signal-processing, waveform-generation, and inference operations that historically ran on central-processing-unit clusters of the Laser Interferometer Gravitational-wave Observatory (LIGO) Data Grid, and integrates the resulting models into the international low-latency alert pipeline. The award covers three coordinated work packages. The first extends model coverage to long-duration binary-neutron-star and neutron-star-black-hole signals using multi-rate and multi-band processing, integrates auxiliary detector channels through multimodal architectures, and develops state-space and ensemble models together with a shared foundation backbone from which task-specific models can be fine-tuned. The second work package builds production-grade cyberinfrastructure that targets the National Artificial Intelligence Research Resource, the National Research Platform, and the Open Science Data Federation, including an Inference-as-a-Service deployment built on the NVIDIA Triton inference server and on the SuperSONIC service that already supports particle and neutrino physics experiments. The third work package delivers community resources: standardized benchmark datasets with persistent digital object identifiers on Zenodo, versioned reference models on Hugging Face, comprehensive documentation and tutorials hosted on Read the Docs, containerized release artifacts, an upgraded continuous integration system that reduces test runtime by an order of magnitude, and an agent-driven development scaffold for community-led code contribution. Training and outreach activities include hands-on tutorials at international collaboration meetings, an annual hands-on lesson at the University of Minnesota Time-Domain Astrophysics Summer School, and a public machine learning challenge focused on binary-neutron-star detection that builds on a prior challenge that engaged hundreds of teams and roughly a thousand individual participants. This award by the Office of Advanced Cyberinfrastructure is jointly supported by the Physics at the Information Frontier Program within the Physics Section of the Directorate for Mathematical and Physical Sciences. 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.

RET Site: Project-Based Learning for Rural Alabama STEM Middle School Teachers in Artificial Intelligence, Machine Learning, and Robotics

This award establishes a renewed Research Experience for Teachers (RET) Site at Auburn University. The site will provide unique and holistic research experiences for 24 middle school math and science teachers in the 7th-8th grades from rural areas of Alabama. The research focus is on smart humanoid and mobile robots enabled by cutting-edge technologies of artificial intelligence (AI) and machine learning (ML). The goals of the site are to equip teachers with knowledge and skills in AI and ML and robotics and promote their interests in these areas and facilitate teachers’ development and implementation of engaging project-based curricular modules for their classrooms. The site will provide research experiences to eight (8) middle school math and science teachers in the 7th-8th grades each year via a six-week summer program and nine-month academic year follow-up, with the research focused on smart mobile robots based on AI and ML. The site has five primary objectives to reach its goals of providing a rigorous and engaging RET experience: 1) provide education and training activities on the fundamentals of AI/ML and robotics, and novel platforms of ML-based smart humanoid and mobile robots for research and education; 2) engage teachers in hands-on research projects on ML-based smart robots that match well with faculty mentors’ research projects; 3) allow teachers to collaborate with engineering and STEM education faculty to develop the project-based curricular modules; 4) foster teachers’ leadership and pedagogical skills via teacher leader mentoring and practice of teaching the RET curricular modules; 5) assist teachers to implement the RET curricular modules via academic year follow-up. 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.

REU Site: Artificial Intelligence in Multidisciplinary Engineering Research for Guided Exploration (AI-MERGE)

The REU Site Artificial Intelligence in Multidisciplinary Engineering Research for Guided Exploration (AI-MERGE) at the University of Memphis is a 10-week summer program that provides undergraduate students with hands-on research experiences applying artificial intelligence (AI) to engineering problems. Advances in AI are transforming the design and operation of infrastructure, biomedical technologies, and sensing systems, creating a need for engineers with expertise in data-driven methods and computational tools. The program recruits undergraduate students, especially those from institutions with limited research opportunities. This REU program is open and available to all American undergraduate students. Participants are engaged in interdisciplinary, faculty-led teams addressing engineering challenges in biomedical systems, imaging, smart infrastructure, and resilient systems. The program integrates research with structured mentoring, training in AI tools, and professional development, and includes opportunities for students to present their work. By expanding access to research and strengthening pathways to graduate education, the project contributes to the development of a strong engineering workforce and supports the nation’s technological competitiveness. The AI-MERGE REU Site provides undergraduate research training in artificial intelligence methods applied to multidisciplinary engineering problems. Research is organized around applications in imaging and biomedical systems, smart infrastructure, and security and resilience, where AI techniques are used for data analysis, modeling, and system optimization. Students participate in faculty-led projects involving machine learning and computational methods, including problem formulation, algorithm development, and performance evaluation using real or simulated data. The program leverages the Vertically Integrated Projects (VIP) framework to support team-based research with tiered mentoring from faculty, graduate students, and experienced undergraduates. Students are embedded in ongoing research efforts and contribute to project outcomes, producing presentations and written reports. The REU Site supports annual cohorts and provides training through research activities, seminars, and professional development. Expected outcomes include student-authored presentations and contributions to engineering research. The project advances AI-enabled approaches for engineering systems while preparing undergraduate participants for graduate study and research careers. 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.

REU Site: Engineers for Exploration

Engineers for Exploration is an experiential program that enables undergraduates to conduct computer systems research that addresses technology gaps in scientific exploration. These undergraduates are provided with an impactful research project, supported by a multi-tiered mentor network and structured collaboration with researchers, scientists, and explorers. Research problems they get to work on are framed by real-world, high-impact applications, providing participants with an inspiring, multidisciplinary research experience. The Engineers for Exploration Research Experience for Undergraduates Site offers research projects rooted in real-world applications and guided by mentorship from scientists and explorers. These use-inspired artificial intelligence research projects include the creation of machine learning and digital signal processing algorithms to automatically classify bird calls from acoustic wildlife sensors, the building of novel sensing strategies on the Smartfin embedded computing system to allow surfers to collect coastal oceanographic data, and the development of FishSense underwater 3D camera systems that automate the collection of fish measurements to turn recreational divers into citizen scientists aiding in fisheries management. Engineers for Exploration involves participants in multidisciplinary research projects spanning embedded systems, remote sensing, digital signal processing, computer vision, machine learning, robotics, and data analytics. The projects develop innovative computer systems that support exploration, and are deployed worldwide by scientists and explorers. For participants, the site fosters mentorship, leadership, and teamwork skills, providing a positive, memorable experience that encourages them to pursue further opportunities, such as research training or graduate education. 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.

REU Site: Research on Prescriptive Analytics for AI-enabled Operations Engineering

This Research Experiences for Undergraduates Site renewal will engage 10 undergraduate students each year in a 9-week summer research program on artificial intelligence (AI)-enabled operations engineering. Many important systems in transportation, healthcare, manufacturing, and services operate under uncertainty, limited resources, and changing conditions. These systems increasingly depend on analytical and computational methods that combine operations research and artificial intelligence to improve planning, coordination, and real-time decision-making. The project will provide students with mentored research, technical training, and professional development that strengthen preparation for graduate study and technical careers in computing, operations research, analytics, and intelligent systems. The research activities will focus on AI-enabled operations engineering, including prescriptive analytics methods that combine optimization, simulation, machine learning, and related computational tools to support complex operational decisions. Students will engage with decision problems in healthcare, next-generation transportation systems, advanced manufacturing, and contested logistics. They will receive training in problem formulation, data analysis, optimization, simulation, algorithm development, machine learning, computational experimentation, and solution evaluation. Sample projects that are computationally challenging include collaborative blood inventory planning, network-aware drone logistics, urban air mobility coordination, additive manufacturing scheduling, and decision-making for logistics operations under disrupted or contested conditions. The program will strengthen student preparation in both computing and operations research methods while helping build a workforce prepared to address complex real-world challenges in critical operational domains. 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.

Collaborative Research: Elements: Physics-Informed Digitized Cyberinfrustructure Towards Next-G Underwater Networks

Underwater wireless communication and networking are critical for monitoring aquatic environments, improving maritime safety, supporting offshore exploration, and enhancing national security. However, research progress in this field has been slow compared with terrestrial wireless systems because conducting underwater wireless experiments is challenging. Natural underwater environments are uncontrollable, and indoor pools and tanks are static and small, which significantly limits research reproducibility, innovation, and public accessibility. To bridge this gap, this project implements a remotely accessible underwater communication and networking platform in a water tunnel that enables experimentation, dataset collection, and artificial intelligence model building under a range of controlled, reconfigurable, reproducible conditions. The testbed, datasets, and developed software enable new wireless communication technology development without the high cost and complications of natural underwater deployments. By sharing advanced experimental tools, datasets, and software with the research community, this project advances scientific discovery and strengthens national leadership in next-generation underwater communication and networking systems. In addition, this project integrates research with education and actively trains students in communication, networking, sensing, and artificial intelligence to support workforce development and address critical national needs. This project designs and deploys a hybrid underwater acoustic, magnetic, and visible light networking system that integrates a reconfigurable water-tunnel testbed, physics-informed multi-modal deep generative channel models, and a scalable digital twin for dynamic underwater networking simulation and optimal control. First, the remotely accessible, reconfigurable testbed instrument enables the collection of acoustic, magnetic, and visible light communication channel data under dynamic water flow and blockage conditions. Second, the collected datasets are used to train physics-informed deep generative channel models that extend beyond the physical testbed to enable large-scale, measurement-driven simulations. Last, the physical testbed and channel models are integrated to develop a networking digital twin, which allows researchers to evaluate multi-modal scheduling strategies, resource allocation schemes, and networking protocols under realistic dynamic underwater conditions. All software, datasets, models, and documentation will be publicly released through open repositories and public websites. By linking physical experimentation with scalable digital simulation, this project will provide sustainable cyberinfrastructure that accelerates data-driven and artificial intelligence-enabled innovation in underwater wireless communication and networking. 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.

REU Site: AI in Sensing, Robotics, and Healthcare (AI-CARE)

During a ten-week summer session, undergraduate participants from institutions nationwide will engage in exciting and challenging projects in multiple engineering disciplines. Each participant will be matched with a research project in one of the many laboratories at Johns Hopkins University (JHU) working at the intersection of artificial intelligence (AI) with robotics, sensing, and/or healthcare. These students will work closely with teams composed of a professor, postdoctoral fellow and/or graduate student, other undergraduate students and, possibly, high school students and/or K-12 educators. These research projects aim to help healthcare providers improve patient care, including projects that help surgeons make operations more effective and less error-prone or help individuals with disabilities regain previously lost functions after limb amputation or stroke. Undergraduate participants will also receive training in technical writing, oral presentations, and research ethics, which will include AI ethics. They will tour laboratories at JHU and the Johns Hopkins Hospital (JHH), the Applied Physics Laboratory, and local robotics and biotech companies. They will have opportunities to perform laparoscopic procedures at the JHH Minimally Invasive Surgical Training and Innovation Center. As a result of these activities, this grant will help prepare a workforce for an important sector of the US economy at the interface of healthcare delivery and engineering. This multidisciplinary AI-CARE REU enables participants to conduct research in multiple engineering fields, while developing strong teamwork and collaboration skills. Faculty mentors offer projects either created specifically for the program or designed to fulfill a facet of ongoing work. Because the Laboratory for Computational Sensing and Robotics (LCSR) has close ties with JHH, the Malone Center for Engineering in Healthcare, and the new Data Science and AI Institute, participants will experience cutting-edge research that is designed to aid medical diagnoses and interventions, while contributing to “the future of medicine.” With 25 years of success managing Supplemental, LSAMP, and Site REU programs within LCSR and its parent entity (i.e., the NSF-funded ERC on Computer-Integrated Surgical Systems and Technology), the AI-CARE REU will modernize previous approaches by offering a new focus on AI, which now intersects multiple facets of traditional computational sensing and medical robotics training. Participants will be equipped to make ethical judgment calls about the responsible use of AI when engaging in research activities related or unrelated to AI (e.g., when performing literature reviews, preparing presentations and papers, and solving research problems). This training will be tightly coupled with more traditional topics, including computational sensing, medical imaging, medical robotics, prosthetics, computer-integrated surgery, and biologically inspired robotics research. The quality of this programming is assured through mentor training, formative and summative assessments, and longitudinal tracking of students. The broader impact of the AI-CARE REU focuses on addressing a vital national need to improve the delivery of healthcare by developing new sensing, imaging, and robotic systems, as well as new techniques designed to introduce, plan, and execute medical procedures. The Site will help develop a pipeline of qualified individuals who will contribute to the STEM workforce, particularly in the multi- and interdisciplinary subjects encountered in technology-enhanced biomedical research, clinical interventions, and basic biological and life sciences. Participants will be well-trained in research communications and ethics (including AI ethics), which are essential skills for success in today’s biotechnology, biomedical science, and healthcare workforce. 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.

Moritz Grosse-Wentrup

We have an open position for a postdoctoral researcher with experience in brain-computer interfacing and artificial intelligence to further advance our new class of Brain-Artificial Intelligence (BAI) interfaces. A central part of your research would be to further develop our BAI for single-unit data recorded in language areas of a post-stroke aphasia patient, a project we carry out in close collaboration with the Translational NeuroTechnology Lab at TUM, headed by Simon Jacob.

Mitra Baratchi

We are looking for an excellent candidate with a master’s degree in MSc in Artificial Intelligence, Computer Science, Mathematics, Statistics, or a closely related field to join a project focused on developing an advanced transparent machine learning framework with application on movement behavioural analysis. Smartwatches and other wearable technologies allow us to continuously collect data on our daily movement behaviour patterns. We would like to understand how machine learning techniques can be used to learn causal effects from time-series data to identify and recommend effective changes in daily activities (i.e., possible behavioural interventions) that are expected to result in concrete health improvements (e.g., improving cardiorespiratory fitness). This research, at the intersection of machine learning and causality, aims to develop algorithms for finding causal relations between behavioural indicators learned from the time series data and associated health-outcomes.

A personal journey on understanding intelligence

The focus of this talk is not about my research in AI or Robotics but my own journey on trying to do research and understand intelligence in a rapidly evolving research landscape. I will trace my path from conducting early-stage research during graduate school, to working on practical solutions within a startup environment, and finally to my current role where I participate in more structured research at a major tech company. Through these varied experiences, I will provide different perspectives on research and talk about how my core beliefs on intelligence have changed and sometimes even been compromised. There are no lessons to be learned from my stories, but hopefully they will be entertaining.

Short and Synthetically Distort: Investor Reactions to Deepfake Financial News

Recent advances in artificial intelligence have led to new forms of misinformation, including highly realistic “deepfake” synthetic media. We conduct three experiments to investigate how and why retail investors react to deepfake financial news. Results from the first two experiments provide evidence that investors use a “realism heuristic,” responding more intensely to audio and video deepfakes as their perceptual realism increases. In the third experiment, we introduce an intervention to prompt analytical thinking, varying whether participants make analytical judgments about credibility or intuitive investment judgments. When making intuitive investment judgments, investors are strongly influenced by both more and less realistic deepfakes. When making analytical credibility judgments, investors are able to discern the non-credibility of less realistic deepfakes but struggle with more realistic deepfakes. Thus, while analytical thinking can reduce the impact of less realistic deepfakes, highly realistic deepfakes are able to overcome this analytical scrutiny. Our results suggest that deepfake financial news poses novel threats to investors.

Memory Decoding Journal Club: Reconstructing a new hippocampal engram for systems reconsolidation and remote memory updating

Join us for the Memory Decoding Journal Club, a collaboration between the Carboncopies Foundation and BPF Aspirational Neuroscience. This month, we're diving into a groundbreaking paper: 'Reconstructing a new hippocampal engram for systems reconsolidation and remote memory updating' by Bo Lei, Bilin Kang, Yuejun Hao, Haoyu Yang, Zihan Zhong, Zihan Zhai, and Yi Zhong from Tsinghua University, Beijing Academy of Artificial Intelligence, IDG/McGovern Institute of Brain Research, and Peking Union Medical College. Dr. Randal Koene will guide us through an engaging discussion on these exciting findings and their implications for neuroscience and memory research.

Active Predictive Coding and the Primacy of Actions in Natural and Artificial Intelligence

Brain Emulation Challenge Workshop

Brain Emulation Challenge workshop will tackle cutting-edge topics such as ground-truthing for validation, leveraging artificial datasets generated from virtual brain tissue, and the transformative potential of virtual brain platforms, such as applied to the forthcoming Brain Emulation Challenge.

The Brain Prize winners' webinar

This webinar brings together three leaders in theoretical and computational neuroscience—Larry Abbott, Haim Sompolinsky, and Terry Sejnowski—to discuss how neural circuits generate fundamental aspects of the mind. Abbott illustrates mechanisms in electric fish that differentiate self-generated electric signals from external sensory cues, showing how predictive plasticity and two-stage signal cancellation mediate a sense of self. Sompolinsky explores attractor networks, revealing how discrete and continuous attractors can stabilize activity patterns, enable working memory, and incorporate chaotic dynamics underlying spontaneous behaviors. He further highlights the concept of object manifolds in high-level sensory representations and raises open questions on integrating connectomics with theoretical frameworks. Sejnowski bridges these motifs with modern artificial intelligence, demonstrating how large-scale neural networks capture language structures through distributed representations that parallel biological coding. Together, their presentations emphasize the synergy between empirical data, computational modeling, and connectomics in explaining the neural basis of cognition—offering insights into perception, memory, language, and the emergence of mind-like processes.

LLMs and Human Language Processing

This webinar convened researchers at the intersection of Artificial Intelligence and Neuroscience to investigate how large language models (LLMs) can serve as valuable “model organisms” for understanding human language processing. Presenters showcased evidence that brain recordings (fMRI, MEG, ECoG) acquired while participants read or listened to unconstrained speech can be predicted by representations extracted from state-of-the-art text- and speech-based LLMs. In particular, text-based LLMs tend to align better with higher-level language regions, capturing more semantic aspects, while speech-based LLMs excel at explaining early auditory cortical responses. However, purely low-level features can drive part of these alignments, complicating interpretations. New methods, including perturbation analyses, highlight which linguistic variables matter for each cortical area and time scale. Further, “brain tuning” of LLMs—fine-tuning on measured neural signals—can improve semantic representations and downstream language tasks. Despite open questions about interpretability and exact neural mechanisms, these results demonstrate that LLMs provide a promising framework for probing the computations underlying human language comprehension and production at multiple spatiotemporal scales.

Llama 3.1 Paper: The Llama Family of Models

Modern artificial intelligence (AI) systems are powered by foundation models. This paper presents a new set of foundation models, called Llama 3. It is a herd of language models that natively support multilinguality, coding, reasoning, and tool usage. Our largest model is a dense Transformer with 405B parameters and a context window of up to 128K tokens. This paper presents an extensive empirical evaluation of Llama 3. We find that Llama 3 delivers comparable quality to leading language models such as GPT-4 on a plethora of tasks. We publicly release Llama 3, including pre-trained and post-trained versions of the 405B parameter language model and our Llama Guard 3 model for input and output safety. The paper also presents the results of experiments in which we integrate image, video, and speech capabilities into Llama 3 via a compositional approach. We observe this approach performs competitively with the state-of-the-art on image, video, and speech recognition tasks. The resulting models are not yet being broadly released as they are still under development.

Trends in NeuroAI - Brain-like topography in transformers (Topoformer)

Dr. Nicholas Blauch will present on his work "Topoformer: Brain-like topographic organization in transformer language models through spatial querying and reweighting". Dr. Blauch is a postdoctoral fellow in the Harvard Vision Lab advised by Talia Konkle and George Alvarez. Paper link: https://openreview.net/pdf?id=3pLMzgoZSA Trends in NeuroAI is a reading group hosted by the MedARC Neuroimaging & AI lab (https://medarc.ai/fmri | https://groups.google.com/g/medarc-fmri).

Generative models for video games (rescheduled)

Developing agents capable of modeling complex environments and human behaviors within them is a key goal of artificial intelligence research. Progress towards this goal has exciting potential for applications in video games, from new tools that empower game developers to realize new creative visions, to enabling new kinds of immersive player experiences. This talk focuses on recent advances of my team at Microsoft Research towards scalable machine learning architectures that effectively capture human gameplay data. In the first part of my talk, I will focus on diffusion models as generative models of human behavior. Previously shown to have impressive image generation capabilities, I present insights that unlock applications to imitation learning for sequential decision making. In the second part of my talk, I discuss a recent project taking ideas from language modeling to build a generative sequence model of an Xbox game.

Generative models for video games

Developing agents capable of modeling complex environments and human behaviors within them is a key goal of artificial intelligence research. Progress towards this goal has exciting potential for applications in video games, from new tools that empower game developers to realize new creative visions, to enabling new kinds of immersive player experiences. This talk focuses on recent advances of my team at Microsoft Research towards scalable machine learning architectures that effectively capture human gameplay data. In the first part of my talk, I will focus on diffusion models as generative models of human behavior. Previously shown to have impressive image generation capabilities, I present insights that unlock applications to imitation learning for sequential decision making. In the second part of my talk, I discuss a recent project taking ideas from language modeling to build a generative sequence model of an Xbox game.

Learning produces a hippocampal cognitive map in the form of an orthogonalized state machine

Cognitive maps confer animals with flexible intelligence by representing spatial, temporal, and abstract relationships that can be used to shape thought, planning, and behavior. Cognitive maps have been observed in the hippocampus, but their algorithmic form and the processes by which they are learned remain obscure. Here, we employed large-scale, longitudinal two-photon calcium imaging to record activity from thousands of neurons in the CA1 region of the hippocampus while mice learned to efficiently collect rewards from two subtly different versions of linear tracks in virtual reality. The results provide a detailed view of the formation of a cognitive map in the hippocampus. Throughout learning, both the animal behavior and hippocampal neural activity progressed through multiple intermediate stages, gradually revealing improved task representation that mirrored improved behavioral efficiency. The learning process led to progressive decorrelations in initially similar hippocampal neural activity within and across tracks, ultimately resulting in orthogonalized representations resembling a state machine capturing the inherent struture of the task. We show that a Hidden Markov Model (HMM) and a biologically plausible recurrent neural network trained using Hebbian learning can both capture core aspects of the learning dynamics and the orthogonalized representational structure in neural activity. In contrast, we show that gradient-based learning of sequence models such as Long Short-Term Memory networks (LSTMs) and Transformers do not naturally produce such orthogonalized representations. We further demonstrate that mice exhibited adaptive behavior in novel task settings, with neural activity reflecting flexible deployment of the state machine. These findings shed light on the mathematical form of cognitive maps, the learning rules that sculpt them, and the algorithms that promote adaptive behavior in animals. The work thus charts a course toward a deeper understanding of biological intelligence and offers insights toward developing more robust learning algorithms in artificial intelligence.

Trends in NeuroAI - Unified Scalable Neural Decoding (POYO)

Lead author Mehdi Azabou will present on his work "POYO-1: A Unified, Scalable Framework for Neural Population Decoding" (https://poyo-brain.github.io/). Mehdi is an ML PhD student at Georgia Tech advised by Dr. Eva Dyer. Paper link: https://arxiv.org/abs/2310.16046 Trends in NeuroAI is a reading group hosted by the MedARC Neuroimaging & AI lab (https://medarc.ai/fmri | https://groups.google.com/g/medarc-fmri).

Reimagining the neuron as a controller: A novel model for Neuroscience and AI

We build upon and expand the efficient coding and predictive information models of neurons, presenting a novel perspective that neurons not only predict but also actively influence their future inputs through their outputs. We introduce the concept of neurons as feedback controllers of their environments, a role traditionally considered computationally demanding, particularly when the dynamical system characterizing the environment is unknown. By harnessing a novel data-driven control framework, we illustrate the feasibility of biological neurons functioning as effective feedback controllers. This innovative approach enables us to coherently explain various experimental findings that previously seemed unrelated. Our research has profound implications, potentially revolutionizing the modeling of neuronal circuits and paving the way for the creation of alternative, biologically inspired artificial neural networks.

Trends in NeuroAI - Meta's MEG-to-image reconstruction

Trends in NeuroAI is a reading group hosted by the MedARC Neuroimaging & AI lab (https://medarc.ai/fmri). Title: Brain-optimized inference improves reconstructions of fMRI brain activity Abstract: The release of large datasets and developments in AI have led to dramatic improvements in decoding methods that reconstruct seen images from human brain activity. We evaluate the prospect of further improving recent decoding methods by optimizing for consistency between reconstructions and brain activity during inference. We sample seed reconstructions from a base decoding method, then iteratively refine these reconstructions using a brain-optimized encoding model that maps images to brain activity. At each iteration, we sample a small library of images from an image distribution (a diffusion model) conditioned on a seed reconstruction from the previous iteration. We select those that best approximate the measured brain activity when passed through our encoding model, and use these images for structural guidance during the generation of the small library in the next iteration. We reduce the stochasticity of the image distribution at each iteration, and stop when a criterion on the "width" of the image distribution is met. We show that when this process is applied to recent decoding methods, it outperforms the base decoding method as measured by human raters, a variety of image feature metrics, and alignment to brain activity. These results demonstrate that reconstruction quality can be significantly improved by explicitly aligning decoding distributions to brain activity distributions, even when the seed reconstruction is output from a state-of-the-art decoding algorithm. Interestingly, the rate of refinement varies systematically across visual cortex, with earlier visual areas generally converging more slowly and preferring narrower image distributions, relative to higher-level brain areas. Brain-optimized inference thus offers a succinct and novel method for improving reconstructions and exploring the diversity of representations across visual brain areas. Speaker: Reese Kneeland is a Ph.D. student at the University of Minnesota working in the Naselaris lab. Paper link: https://arxiv.org/abs/2312.07705

Trends in NeuroAI - Meta's MEG-to-image reconstruction

Trends in NeuroAI is a reading group hosted by the MedARC Neuroimaging & AI lab (https://medarc.ai/fmri). This will be an informal journal club presentation, we do not have an author of the paper joining us. Title: Brain decoding: toward real-time reconstruction of visual perception Abstract: In the past five years, the use of generative and foundational AI systems has greatly improved the decoding of brain activity. Visual perception, in particular, can now be decoded from functional Magnetic Resonance Imaging (fMRI) with remarkable fidelity. This neuroimaging technique, however, suffers from a limited temporal resolution (≈0.5 Hz) and thus fundamentally constrains its real-time usage. Here, we propose an alternative approach based on magnetoencephalography (MEG), a neuroimaging device capable of measuring brain activity with high temporal resolution (≈5,000 Hz). For this, we develop an MEG decoding model trained with both contrastive and regression objectives and consisting of three modules: i) pretrained embeddings obtained from the image, ii) an MEG module trained end-to-end and iii) a pretrained image generator. Our results are threefold: Firstly, our MEG decoder shows a 7X improvement of image-retrieval over classic linear decoders. Second, late brain responses to images are best decoded with DINOv2, a recent foundational image model. Third, image retrievals and generations both suggest that MEG signals primarily contain high-level visual features, whereas the same approach applied to 7T fMRI also recovers low-level features. Overall, these results provide an important step towards the decoding - in real time - of the visual processes continuously unfolding within the human brain. Speaker: Dr. Paul Scotti (Stability AI, MedARC) Paper link: https://arxiv.org/abs/2310.19812

Trends in NeuroAI - SwiFT: Swin 4D fMRI Transformer

Trends in NeuroAI is a reading group hosted by the MedARC Neuroimaging & AI lab (https://medarc.ai/fmri). Title: SwiFT: Swin 4D fMRI Transformer Abstract: Modeling spatiotemporal brain dynamics from high-dimensional data, such as functional Magnetic Resonance Imaging (fMRI), is a formidable task in neuroscience. Existing approaches for fMRI analysis utilize hand-crafted features, but the process of feature extraction risks losing essential information in fMRI scans. To address this challenge, we present SwiFT (Swin 4D fMRI Transformer), a Swin Transformer architecture that can learn brain dynamics directly from fMRI volumes in a memory and computation-efficient manner. SwiFT achieves this by implementing a 4D window multi-head self-attention mechanism and absolute positional embeddings. We evaluate SwiFT using multiple large-scale resting-state fMRI datasets, including the Human Connectome Project (HCP), Adolescent Brain Cognitive Development (ABCD), and UK Biobank (UKB) datasets, to predict sex, age, and cognitive intelligence. Our experimental outcomes reveal that SwiFT consistently outperforms recent state-of-the-art models. Furthermore, by leveraging its end-to-end learning capability, we show that contrastive loss-based self-supervised pre-training of SwiFT can enhance performance on downstream tasks. Additionally, we employ an explainable AI method to identify the brain regions associated with sex classification. To our knowledge, SwiFT is the first Swin Transformer architecture to process dimensional spatiotemporal brain functional data in an end-to-end fashion. Our work holds substantial potential in facilitating scalable learning of functional brain imaging in neuroscience research by reducing the hurdles associated with applying Transformer models to high-dimensional fMRI. Speaker: Junbeom Kwon is a research associate working in Prof. Jiook Cha’s lab at Seoul National University. Paper link: https://arxiv.org/abs/2307.05916

Use of Artificial Intelligence by Law Enforcement Authorities in the EU

Recently, artificial intelligence (AI) has become a global priority. Rapid and ongoing technological advancements in AI have prompted European legislative initiatives to regulate its use. In April 2021, the European Commission submitted a proposal for a Regulation that would harmonize artificial intelligence rules across the EU, including the law enforcement sector. Consequently, law enforcement officials await the outcome of the ongoing inter-institutional negotiations (trilogue) with great anticipation, as it will define how to capitalize on the opportunities presented by AI and how to prevent criminals from abusing this emergent technology.

BrainLM Journal Club

Connor Lane will lead a journal club on the recent BrainLM preprint, a foundation model for fMRI trained using self-supervised masked autoencoder training. Preprint: https://www.biorxiv.org/content/10.1101/2023.09.12.557460v1 Tweeprint: https://twitter.com/david_van_dijk/status/1702336882301112631?t=Q2-U92-BpJUBh9C35iUbUA&s=19

Foundation models in ophthalmology

Abstract to follow.

Cognitive Computational Neuroscience 2023

CCN is an annual conference that serves as a forum for cognitive science, neuroscience, and artificial intelligence researchers dedicated to understanding the computations that underlie complex behavior.

Algonauts 2023 winning paper journal club (fMRI encoding models)

Algonauts 2023 was a challenge to create the best model that predicts fMRI brain activity given a seen image. Huze team dominated the competition and released a preprint detailing their process. This journal club meeting will involve open discussion of the paper with Q/A with Huze. Paper: https://arxiv.org/pdf/2308.01175.pdf Related paper also from Huze that we can discuss: https://arxiv.org/pdf/2307.14021.pdf

1.8 billion regressions to predict fMRI (journal club)

Public journal club where this week Mihir will present on the 1.8 billion regressions paper (https://www.biorxiv.org/content/10.1101/2022.03.28.485868v2), where the authors use hundreds of pretrained model embeddings to best predict fMRI activity.

In search of the unknown: Artificial intelligence and foraging

Diverse applications of artificial intelligence and mathematical approaches in ophthalmology

Ophthalmology is ideally placed to benefit from recent advances in artificial intelligence. It is a highly image-based specialty and provides unique access to the microvascular circulation and the central nervous system. This talk will demonstrate diverse applications of machine learning and deep learning techniques in ophthalmology, including in age-related macular degeneration (AMD), the leading cause of blindness in industrialized countries, and cataract, the leading cause of blindness worldwide. This will include deep learning approaches to automated diagnosis, quantitative severity classification, and prognostic prediction of disease progression, both from images alone and accompanied by demographic and genetic information. The approaches discussed will include deep feature extraction, label transfer, and multi-modal, multi-task training. Cluster analysis, an unsupervised machine learning approach to data classification, will be demonstrated by its application to geographic atrophy in AMD, including exploration of genotype-phenotype relationships. Finally, mediation analysis will be discussed, with the aim of dissecting complex relationships between AMD disease features, genotype, and progression.

Consciousness in the age of mechanical minds

We are now clearly entering a new age in our relationship with machines. The power of AI natural language processors and image generators has rapidly exceeded the expectations of even those who developed them. Serious questions are now being asked about the extent to which machines could become — or perhaps already are — sentient or conscious. Do AI machines understand the instructions they are given and the answers they provide? In this talk I will consider the prospects for conscious machines, by which I mean machines that have feelings, know about their own existence, and about ours. I will suggest that the recent focus on information processing in models of consciousness, in which the brain is treated as a kind of digital computer, have mislead us about the nature of consciousness and how it is produced in biological systems. Treating the brain as an energy processing system is more likely to yield answers to these fundamental questions and help us understand how and when machines might become minds.

How AI is advancing Clinical Neuropsychology and Cognitive Neuroscience

This talk aims to highlight the immense potential of Artificial Intelligence (AI) in advancing the field of psychology and cognitive neuroscience. Through the integration of machine learning algorithms, big data analytics, and neuroimaging techniques, AI has the potential to revolutionize the way we study human cognition and brain characteristics. In this talk, I will highlight our latest scientific advancements in utilizing AI to gain deeper insights into variations in cognitive performance across the lifespan and along the continuum from healthy to pathological functioning. The presentation will showcase cutting-edge examples of AI-driven applications, such as deep learning for automated scoring of neuropsychological tests, natural language processing to characeterize semantic coherence of patients with psychosis, and other application to diagnose and treat psychiatric and neurological disorders. Furthermore, the talk will address the challenges and ethical considerations associated with using AI in psychological research, such as data privacy, bias, and interpretability. Finally, the talk will discuss future directions and opportunities for further advancements in this dynamic field.

Deep learning applications in ophthalmology

Deep learning techniques have revolutionized the field of image analysis and played a disruptive role in the ability to quickly and efficiently train image analysis models that perform as well as human beings. This talk will cover the beginnings of the application of deep learning in the field of ophthalmology and vision science, and cover a variety of applications of using deep learning as a method for scientific discovery and latent associations.

AI for Multi-centre Epilepsy Lesion Detection on MRI

Epilepsy surgery is a safe but underutilised treatment for drug-resistant focal epilepsy. One challenge in the presurgical evaluation of patients with drug-resistant epilepsy are patients considered “MRI negative”, i.e. where a structural brain abnormality has not been identified on MRI. A major pathology in “MRI negative” patients is focal cortical dysplasia (FCD), where lesions are often small or subtle and easily missed by visual inspection. In recent years, there has been an explosion in artificial intelligence (AI) research in the field of healthcare. Automated FCD detection is an area where the application of AI may translate into significant improvements in the presurgical evaluation of patients with focal epilepsy. I will provide an overview of our automated FCD detection work, the Multicentre Epilepsy Lesion Detection (MELD) project and how AI algorithms are beginning to be integrated into epilepsy presurgical planning at Great Ormond Street Hospital and elsewhere around the world. Finally, I will discuss the challenges and future work required to bring AI to the forefront of care for patients with epilepsy.

Does subjective time interact with the heart rate?

Decades of research have investigated the relationship between perception of time and heart rate with often mixed results. In search of such a relationship, I will present my far journey between two projects: from time perception in the realistic VR experience of crowded subway trips in the order of minutes (project 1); to the perceived duration of sub-second white noise tones (project 2). Heart rate had multiple concurrent relationships with subjective temporal distortions for the sub-second tones, while the effects were lacking or weak for the supra-minute subway trips. What does the heart have to do with sub-second time perception? We addressed this question with a cardiac drift-diffusion model, demonstrating the sensory accumulation of temporal evidence as a function of heart rate.

On the link between conscious function and general intelligence in humans and machines

In popular media, there is often a connection drawn between the advent of awareness in artificial agents and those same agents simultaneously achieving human or superhuman level intelligence. In this talk, I will examine the validity and potential application of this seemingly intuitive link between consciousness and intelligence. I will do so by examining the cognitive abilities associated with three contemporary theories of conscious function: Global Workspace Theory (GWT), Information Generation Theory (IGT), and Attention Schema Theory (AST), and demonstrating that all three theories specifically relate conscious function to some aspect of domain-general intelligence in humans. With this insight, we will turn to the field of Artificial Intelligence (AI) and find that, while still far from demonstrating general intelligence, many state-of-the-art deep learning methods have begun to incorporate key aspects of each of the three functional theories. Given this apparent trend, I will use the motivating example of mental time travel in humans to propose ways in which insights from each of the three theories may be combined into a unified model. I believe that doing so can enable the development of artificial agents which are not only more generally intelligent but are also consistent with multiple current theories of conscious function.

Do large language models solve verbal analogies like children do?

Analogical reasoning –learning about new things by relating it to previous knowledge– lies at the heart of human intelligence and creativity and forms the core of educational practice. Children start creating and using analogies early on, making incredible progress moving from associative processes to successful analogical reasoning. For example, if we ask a four-year-old “Horse belongs to stable like chicken belongs to …?” they may use association and reply “egg”, whereas older children will likely give the intended relational response “chicken coop” (or other term to refer to a chicken’s home). Interestingly, despite state-of-the-art AI-language models having superhuman encyclopedic knowledge and superior memory and computational power, our pilot studies show that these large language models often make mistakes providing associative rather than relational responses to verbal analogies. For example, when we asked four- to eight-year-olds to solve the analogy “body is to feet as tree is to …?” they responded “roots” without hesitation, but large language models tend to provide more associative responses such as “leaves”. In this study we examine the similarities and differences between children's and six large language models' (Dutch/multilingual models: RobBERT, BERT-je, M-BERT, GPT-2, M-GPT, Word2Vec and Fasttext) responses to verbal analogies extracted from an online adaptive learning environment, where >14,000 7-12 year-olds from the Netherlands solved 20 or more items from a database of 900 Dutch language verbal analogies.

Lifelong Learning AI via neuro inspired solutions

AI embedded in real systems, such as in satellites, robots and other autonomous devices, must make fast, safe decisions even when the environment changes, or under limitations on the available power; to do so, such systems must be adaptive in real time. To date, edge computing has no real adaptivity – rather the AI must be trained in advance, typically on a large dataset with much computational power needed; once fielded, the AI is frozen: It is unable to use its experience to operate if environment proves outside its training or to improve its expertise; and worse, since datasets cannot cover all possible real-world situations, systems with such frozen intelligent control are likely to fail. Lifelong Learning is the cutting edge of artificial intelligence - encompassing computational methods that allow systems to learn in runtime and incorporate learning for application in new, unanticipated situations. Until recently, this sort of computation has been found exclusively in nature; thus, Lifelong Learning looks to nature, and in particular neuroscience, for its underlying principles and mechanisms and then translates them to this new technology. Our presentation will introduce a number of state-of-the-art approaches to achieve AI adaptive learning, including from the DARPA’s L2M program and subsequent developments. Many environments are affected by temporal changes, such as the time of day, week, season, etc. A way to create adaptive systems which are both small and robust is by making them aware of time and able to comprehend temporal patterns in the environment. We will describe our current research in temporal AI, while also considering power constraints.

Associative memory of structured knowledge

A long standing challenge in biological and artificial intelligence is to understand how new knowledge can be constructed from known building blocks in a way that is amenable for computation by neuronal circuits. Here we focus on the task of storage and recall of structured knowledge in long-term memory. Specifically, we ask how recurrent neuronal networks can store and retrieve multiple knowledge structures. We model each structure as a set of binary relations between events and attributes (attributes may represent e.g., temporal order, spatial location, role in semantic structure), and map each structure to a distributed neuronal activity pattern using a vector symbolic architecture (VSA) scheme. We then use associative memory plasticity rules to store the binarized patterns as fixed points in a recurrent network. By a combination of signal-to-noise analysis and numerical simulations, we demonstrate that our model allows for efficient storage of these knowledge structures, such that the memorized structures as well as their individual building blocks (e.g., events and attributes) can be subsequently retrieved from partial retrieving cues. We show that long-term memory of structured knowledge relies on a new principle of computation beyond the memory basins. Finally, we show that our model can be extended to store sequences of memories as single attractors.

What do neurons want?

AI-assisted language learning: Assessing learners who memorize and reason by analogy

Vocabulary learning applications like Duolingo have millions of users around the world, but yet are based on very simple heuristics to choose teaching material to provide to their users. In this presentation, we will discuss the possibility to develop more advanced artificial teachers, which would be based on modeling of the learner’s inner characteristics. In the case of teaching vocabulary, understanding how the learner memorizes is enough. When it comes to picking grammar exercises, it becomes essential to assess how the learner reasons, in particular by analogy. This second application will illustrate how analogical and case-based reasoning can be employed in an alternative way in education: not as the teaching algorithm, but as a part of the learner’s model.

Learning static and dynamic mappings with local self-supervised plasticity

Animals exhibit remarkable learning capabilities with little direct supervision. Likewise, self-supervised learning is an emergent paradigm in artificial intelligence, closing the performance gap to supervised learning. In the context of biology, self-supervised learning corresponds to a setting where one sense or specific stimulus may serve as a supervisory signal for another. After learning, the latter can be used to predict the former. On the implementation level, it has been demonstrated that such predictive learning can occur at the single neuron level, in compartmentalized neurons that separate and associate information from different streams. We demonstrate the power such self-supervised learning over unsupervised (Hebb-like) learning rules, which depend heavily on stimulus statistics, in two examples: First, in the context of animal navigation where predictive learning can associate internal self-motion information always available to the animal with external visual landmark information, leading to accurate path-integration in the dark. We focus on the well-characterized fly head direction system and show that our setting learns a connectivity strikingly similar to the one reported in experiments. The mature network is a quasi-continuous attractor and reproduces key experiments in which optogenetic stimulation controls the internal representation of heading, and where the network remaps to integrate with different gains. Second, we show that incorporating global gating by reward prediction errors allows the same setting to learn conditioning at the neuronal level with mixed selectivity. At its core, conditioning entails associating a neural activity pattern induced by an unconditioned stimulus (US) with the pattern arising in response to a conditioned stimulus (CS). Solving the generic problem of pattern-to-pattern associations naturally leads to emergent cognitive phenomena like blocking, overshadowing, saliency effects, extinction, interstimulus interval effects etc. Surprisingly, we find that the same network offers a reductionist mechanism for causal inference by resolving the post hoc, ergo propter hoc fallacy.

A Framework for a Conscious AI: Viewing Consciousness through a Theoretical Computer Science Lens

We examine consciousness from the perspective of theoretical computer science (TCS), a branch of mathematics concerned with understanding the underlying principles of computation and complexity, including the implications and surprising consequences of resource limitations. We propose a formal TCS model, the Conscious Turing Machine (CTM). The CTM is influenced by Alan Turing's simple yet powerful model of computation, the Turing machine (TM), and by the global workspace theory (GWT) of consciousness originated by cognitive neuroscientist Bernard Baars and further developed by him, Stanislas Dehaene, Jean-Pierre Changeux, George Mashour, and others. However, the CTM is not a standard Turing Machine. It’s not the input-output map that gives the CTM its feeling of consciousness, but what’s under the hood. Nor is the CTM a standard GW model. In addition to its architecture, what gives the CTM its feeling of consciousness is its predictive dynamics (cycles of prediction, feedback and learning), its internal multi-modal language Brainish, and certain special Long Term Memory (LTM) processors, including its Inner Speech and Model of the World processors. Phenomena generally associated with consciousness, such as blindsight, inattentional blindness, change blindness, dream creation, and free will, are considered. Explanations derived from the model draw confirmation from consistencies at a high level, well above the level of neurons, with the cognitive neuroscience literature. Reference. L. Blum and M. Blum, "A theory of consciousness from a theoretical computer science perspective: Insights from the Conscious Turing Machine," PNAS, vol. 119, no. 21, 24 May 2022. https://www.pnas.org/doi/epdf/10.1073/pnas.2115934119

Feedforward and feedback processes in visual recognition

Progress in deep learning has spawned great successes in many engineering applications. As a prime example, convolutional neural networks, a type of feedforward neural networks, are now approaching – and sometimes even surpassing – human accuracy on a variety of visual recognition tasks. In this talk, however, I will show that these neural networks and their recent extensions exhibit a limited ability to solve seemingly simple visual reasoning problems involving incremental grouping, similarity, and spatial relation judgments. Our group has developed a recurrent network model of classical and extra-classical receptive field circuits that is constrained by the anatomy and physiology of the visual cortex. The model was shown to account for diverse visual illusions providing computational evidence for a novel canonical circuit that is shared across visual modalities. I will show that this computational neuroscience model can be turned into a modern end-to-end trainable deep recurrent network architecture that addresses some of the shortcomings exhibited by state-of-the-art feedforward networks for solving complex visual reasoning tasks. This suggests that neuroscience may contribute powerful new ideas and approaches to computer science and artificial intelligence.

Careers for neuroscience in Artificial Intelligence

The purpose of this event is twofold: to raise awareness of careers in AI to neuroscience postgraduate and Early Career Researchers (ECRs), and to give the chance for commercial organisations to acquire and diversify their talent pool.  We know that our early career members are highly motivated and interested in different career pathways, and wish to help them fulfil their ambitions. This will be a hybrid event held in person at Arca Blanca, Covent Garden, London and also available online. FREE for BNA members!

Faking emotions and a therapeutic role for robots and chatbots: Ethics of using AI in psychotherapy

In recent years, there has been a proliferation of social robots and chatbots that are designed so that users make an emotional attachment with them. This talk will start by presenting the first such chatbot, a program called Eliza designed by Joseph Weizenbaum in the mid 1960s. Then we will look at some recent robots and chatbots with Eliza-like interfaces and examine their benefits as well as various ethical issues raised by deploying such systems.

Forensic use of face recognition systems for investigation

With the increasing development of automatic systems and artificial intelligence, face recognition is becoming increasingly important in forensic and civil contexts. However, face recognition has yet to be thoroughly empirically studied to provide an adequate scientific and legal framework for investigative and court purposes. This observation sets the foundation for the research. We focus on issues related to face images and the use of automatic systems. Our objective is to validate a likelihood ratio computation methodology for interpreting comparison scores from automatic face recognition systems (score-based likelihood ratio, SLR). We collected three types of traces: portraits (ID), video surveillance footage recorded by ATM and by a wide-angle camera (CCTV). The performance of two automatic face recognition systems is compared: the commercial IDEMIA Morphoface (MFE) system and the open source FaceNet algorithm.

Understanding Natural Language: Insights From Cognitive Science, Cognitive Neuroscience, and Artificial Intelligence

Artificial Intelligence and Racism – What are the implications for scientific research?

As questions of race and justice have risen to the fore across the sciences, the ALBA Network has invited Dr Shakir Mohamed (Senior Research Scientist at DeepMind, UK) to provide a keynote speech on Artificial Intelligence and racism, and the implications for scientific research, that will be followed by a discussion chaired by Dr Konrad Kording (Department of Neuroscience at University of Pennsylvania, US - neuromatch co-founder)

Interdisciplinary College

The Interdisciplinary College is an annual spring school which offers a dense state-of-the-art course program in neurobiology, neural computation, cognitive science/psychology, artificial intelligence, machine learning, robotics and philosophy. It is aimed at students, postgraduates and researchers from academia and industry. This year's focus theme "Flexibility" covers (but not be limited to) the nervous system, the mind, communication, and AI & robotics. All this will be packed into a rich, interdisciplinary program of single- and multi-lecture courses, and less traditional formats.

Cognitive Maps

Ample evidence suggests that the brain generates internal simulations of the outside world to guide our thoughts and actions. These mental representations, or cognitive maps, are thought to be essential for our very comprehension of reality. I will discuss what is known about the informational structure of cognitive maps, their neural underpinnings, and how they relate to behavior, evolution, disease, and the current revolution in artificial intelligence.

Implementing structure mapping as a prior in deep learning models for abstract reasoning

Building conceptual abstractions from sensory information and then reasoning about them is central to human intelligence. Abstract reasoning both relies on, and is facilitated by, our ability to make analogies about concepts from known domains to novel domains. Structure Mapping Theory of human analogical reasoning posits that analogical mappings rely on (higher-order) relations and not on the sensory content of the domain. This enables humans to reason systematically about novel domains, a problem with which machine learning (ML) models tend to struggle. We introduce a two-stage neural net framework, which we label Neural Structure Mapping (NSM), to learn visual analogies from Raven's Progressive Matrices, an abstract visual reasoning test of fluid intelligence. Our framework uses (1) a multi-task visual relationship encoder to extract constituent concepts from raw visual input in the source domain, and (2) a neural module net analogy inference engine to reason compositionally about the inferred relation in the target domain. Our NSM approach (a) isolates the relational structure from the source domain with high accuracy, and (b) successfully utilizes this structure for analogical reasoning in the target domain.

Analogical Reasoning with Neuro-Symbolic AI

Knowledge discovery with computers requires a huge amount of search. Analogical reasoning is effective for efficient knowledge discovery. Therefore, we proposed analogical reasoning systems based on first-order predicate logic using Neuro-Symbolic AI. Neuro-Symbolic AI is a combination of Symbolic AI and artificial neural networks and has features that are easy for human interpretation and robust against data ambiguity and errors. We have implemented analogical reasoning systems by Neuro-symbolic AI models with word embedding which can represent similarity between words. Using the proposed systems, we efficiently extracted unknown rules from knowledge bases described in Prolog. The proposed method is the first case of analogical reasoning based on the first-order predicate logic using deep learning.

Human-like scene interpretation by a brain-inspired model

From single cell to population coding during defensive behaviors in prefrontal circuits

Coping with threatening situations requires both identifying stimuli predicting danger and selecting adaptive behavioral responses in order to survive. The dorso medial prefrontal cortex (dmPFC) is a critical structure involved in the regulation of threat-related behaviour, yet it is still largely unclear how threat-predicting stimuli and defensive behaviours are associated within prefrontal networks in order to successfully drive adaptive responses. Over the past years, we used a combination we used a combination of extracellular recordings, neuronal decoding approaches, and state of the art optogenetic manipulations to identify key neuronal elements and mechanisms controlling defensive fear responses. I will present an overview of our recent work ranging from analyses of dedicated neuronal types and oscillatory and synchronization mechanisms to artificial intelligence approaches used to decode the activity or large population of neurons. Ultimately these analyses allowed the identification of high dimensional representations of defensive behavior unfolding within prefrontal networks.

Towards a More Authentic Vision of the (multi)Coding Potential of RNA

Ten of thousands of open reading frames (ORFs) are hidden within transcripts. They have eluded annotations because they are either small or within unsuspected locations. These are named alternative ORFs (altORFs) or small ORFs and have recently been highlighted by innovative proteogenomic approaches, such as our OpenProt resource, revealing their existence and implications in biological functions. Due to the absence of altORFs from annotations, pathogenic mutations within these are being ignored. I will discuss our latest progress on the re-analysis of large-scale proteomics datasets to improve our knowledge of proteomic diversity, and the functional characterization of a second protein coded by the FUS gene. Finally, I will explain the need to map the coding potential of the transcriptome using artificial intelligence rather than with conventional annotations that do not capture the full translational activity of ribosomes.

Maths, AI and Neuroscience meeting

To understand brain function and develop artificial general intelligence it has become abundantly clear that there should be a close interaction among Neuroscience, machine learning and mathematics. There is a general hope that understanding the brain function will provide us with more powerful machine learning algorithms. On the other hand advances in machine learning are now providing the much needed tools to not only analyse brain activity data but also to design better experiments to expose brain function. Both neuroscience and machine learning explicitly or implicitly deal with high dimensional data and systems. Mathematics can provide powerful new tools to understand and quantify the dynamics of biological and artificial systems as they generate behavior that may be perceived as intelligent. In this meeting we bring together experts from Mathematics, Artificial Intelligence and Neuroscience for a three day long hybrid meeting. We will have talks on mathematical tools in particular Topology to understand high dimensional data, explainable AI, how AI can help neuroscience and to what extent the brain may be using algorithms similar to the ones used in modern machine learning. Finally we will wrap up with a discussion on some aspects of neural hardware that may not have been considered in machine learning.

Artificial Intelligence towards Autonomous Manufacturing

In this talk, Frans Cronje will be speaking about the journey towards autonomous manufacturing. He will demonstrate how artificial intelligence (AI) can be implemented to achieve a reduction in poor quality costs in manufacturing. The talk will showcase the power of applied AI.

Causal Reasoning: Its role in the architecture and development of the mind

The seminar will first outline the architecture of the human mind, specifying general and domain-specific mental processes. The place of causal reasoning and its relations with the other processes will be specified. Experimental, psychometric, developmental, and brain-based evidence will be summarized. The main message of the talk is that causal thought involves domain-specific core processes rooted in perception and served by special brain networks which capture interactions between objects. With development, causal reasoning is increasingly associated with a general abstraction system which generates general principles underlying inductive, analogical, and deductive reasoning and also heuristics for specifying causal relations. These associations are discussed in some detail. Possible implications for artificial intelligence and educational implications are also discussed.

Embodied Artificial Intelligence: Building brain and body together in bio-inspired robots

TBC

AI UPtake: Panel discussion on collaborative research

Artificial intelligence (AI) and machine learning (ML) can facilitate new paradigms and solutions in almost every research field. Collaboration is essential to achieve tangible and concrete progress in impactful and meaningful AI and ML research, due to its transdisciplinary nature. Come and meet University of Pretoria (UP) academics that are embracing and exploring the opportunities that AI and ML offer to transcend the conventional boundaries of their disciplines. Join the discussion to debate this new frontier of opportunities and challenges that may enable you to look beyond the obvious, and discover new directions and opportunities that we may offer for tomorrow — together!

Career in Data Science Webinar

What does an executive at a South African Bank, a machine learning lead, and a CEO of an AI company have in common? They all will be on a panel talking about careers in Data Science, Machine Learning and Artificial Intelligence

Seeing things clearly: Image understanding through hard-attention and reasoning with structured knowledges

In this talk, Jonathan aims to frame the current challenges of explainability and understanding in ML-driven approaches to image processing, and their potential solution through explicit inference techniques.

Can connectomics help us understand the brain and sustain the revolution in AI?

3 short talks and a panel discussion on the topic of "Can connectomics help us understand the brain and sustain the revolution in AI?" Expect beautiful connectomics data, provocative dreaming, realistic critiques and everything in between. Students & post-docs, stay on to meet our 3 amazing speakers. Moderator: Dr Greg Jefferis https://www2.mrc-lmb.cam.ac.uk/group-leaders/h-to-m/gregory-jefferis/

Playing StarCraft and saving the world using multi-agent reinforcement learning!

This is my C-14 Impaler gauss rifle! There are many like it, but this one is mine!" - A terran marine If you have never heard of a terran marine before, then you have probably missed out on playing the very engaging and entertaining strategy computer game, StarCraft. However, don’t despair, because what we have in store might be even more exciting! In this interactive session, we will take you through, step-by-step, on how to train a team of terran marines to defeat a team of marines controlled by the built-in game AI in StarCraft II. How will we achieve this? Using multi-agent reinforcement learning (MARL). MARL is a useful framework for building distributed intelligent systems. In MARL, multiple agents are trained to act as individual decision-makers of some larger system, while learning to work as a team. We will show you how to use Mava (https://github.com/instadeepai/Mava), a newly released research framework for MARL to build a multi-agent learning system for StarCraft II. We will provide the necessary guidance, tools and background to understand the key concepts behind MARL, how to use Mava building blocks to build systems and how to train a system from scratch. We will conclude the session by briefly sharing various exciting real-world application areas for MARL at InstaDeep, such as large-scale autonomous train navigation and circuit board routing. These are problems that become exponentially more difficult to solve as they scale. Finally, we will argue that many of humanity’s most important practical problems are reminiscent of the ones just described. These include, for example, the need for sustainable management of distributed resources under the pressures of climate change, or efficient inventory control and supply routing in critical distribution networks, or robotic teams for rescue missions and exploration. We believe MARL has enormous potential to be applied in these areas and we hope to inspire you to get excited and interested in MARL and perhaps one day contribute to the field!