Current Work

Gordon LaForge, Anne-Marie Slaughter, Simon Levin, Adam Day, Allison Stanger, Ann Kinzig, Stephanie Forrest, Bruce Schneier, Cristopher Moore, Kevin O’Neil, Moshe Vardi, Nazli Choucri, Robert Axelrod, Sihao Huang, Steve Crocker, Tina Eliassi-Rad, Nick Silitch, and Merle Weidt (2024): How Complexity Thinking Can Help the World Navigate AI - New America

In the coming years, advanced artificial intelligence (AI) systems are expected to bring significant benefits and risks for humanity. Many governments, companies, researchers, and civil society organizations are proposing, and in some cases, building global governance frameworks and institutions to promote AI safety and beneficial development. Complexity thinking, a way of viewing the world not just as discrete parts at the macro level but also in terms of bottom-up and interactive complex adaptive systems, can be a useful intellectual and scientific lens for shaping these endeavors. This paper details how insights from the science and theory of complexity can aid understanding of the challenges posed by AI and its potential impacts on society. Given the characteristics of complex adaptive systems, the paper recommends that global AI governance be based on providing a fit, adaptive response system that mitigates harmful outcomes of AI and enables positive aspects to flourish. The paper proposes components of such a system in three areas: access and power, international relations and global stability; and accountability and liability.

Lennart Heim, Tim Fist, Janet Egan, Sihao Huang, Stephen Zekany, Robert Trager, Michael A Osborne, Noa Zilberman (2024): Governing Through the Cloud: The Intermediary Role of Compute Providers in AI Regulation - Oxford Martin School

As jurisdictions around the world take their first steps toward regulating the most powerful AI systems, such as the EU AI Act and the US Executive Order 14110, there is a growing need for effective enforcement mechanisms that can verify compliance and respond to violations. We argue that compute providers should have legal obligations and ethical responsibilities associated with AI development and deployment, both to provide secure infrastructure and to serve as intermediaries for AI regulation. Compute providers can play an essential role in a regulatory ecosystem via four key capacities: as securers, safeguarding AI systems and critical infrastructure; as record keepers, enhancing visibility for policymakers; as verifiers of customer activities, ensuring oversight; and as enforcers, taking actions against rule violations. We analyze the technical feasibility of performing these functions in a targeted and privacy-conscious manner and present a range of technical instruments. In particular, we describe how non-confidential information, to which compute providers largely already have access, can provide two key governance-relevant properties of a computational workload: its type (e.g., large-scale training or inference) and the amount of compute it has consumed. Using AI Executive Order 14110 as a case study, we outline how the US is beginning to implement record keeping requirements for compute providers. We also explore how verification and enforcement roles could be added to establish a comprehensive AI compute oversight scheme. We argue that internationalization will be key to effective implementation, and highlight the critical challenge of balancing confidentiality and privacy with risk mitigation as the role of compute providers in AI regulation expands.

Eleanor Atkins, M. Taylor Fravel, Raymond Wang, Nick Ackert, Sihao Huang (2023): Two Paths: Why States Join or Avoid China’s Belt and Road Initiative - Global Studies Quarterly

Although China’s motives for developing the Belt and Road Initiative (BRI) have been well studied, scholars have yet to comprehensively examine why states seek to join the initiative. We fill this gap by examining how and why states join the BRI. Countries join by signing a Memorandum of Understanding (MOU) with China on cooperation under the BRI framework. These MOUs create few or no obligations for the states who sign them but increase the possibility of reaping future economic benefits. Thus, we argue that most states should join the BRI unless they view the costs of participation as higher. We hypothesize, and find support for, the argument that democracies are less likely to join because they view participating in a Chinese-led initiative as more costly than non-democracies. Our statistical analysis using a new dataset of BRI participants and paired case studies provides quantitative and qualitative support for this argument.

Sihao Huang, Alexander Siegenfeld, and Andrew Gelman (2022): How Democracies Polarize: A Multi-Level Perspective - arXiv:2211.01249

Democracies employ elections at various scales to select officials at the corresponding levels of administration. The geographical distribution of political opinion, the policy issues delegated to each level, and the multilevel interactions between elections can all greatly impact the makeup of these representative bodies. This perspective is not new: the adoption of federal systems has been motivated by the idea that they possess desirable traits not provided by democracies on a single scale. Yet most existing models of polarization do not capture how nested local and national elections interact with heterogeneous political geographies. We begin by developing a framework to describe the multilevel distribution of opinions and analyze the flow of variance among geographic scales, applying it to historical data in the United States from 1912 to 2020. We describe how unstable elections can arise due to the spatial distribution of opinions and how tradeoffs occur between national and local elections. We also examine multi-dimensional spaces of political opinion, for which we show that a decrease in local salience can constrain the dimensions along which elections occur, preventing a federal system from serving as an effective safeguard against polarization. These analyses, based on the interactions between elections and opinion distributions at various scales, offer insights into how democracies can be strengthened to mitigate polarization and increase electoral representation.

Julia Balla, Sihao Huang, Owen Dugan, Rumen Dangovski, and Marin Soljacic (2022): Machine Learning-Aided Discovery of Formal Models in Social Science - arXiv:2210.00563

In social science, formal and quantitative models, such as ones describing economic growth and collective action, are used to formulate mechanistic explanations, provide predictions, and uncover questions about observed phenomena. Here, we demonstrate the use of a machine learning system to aid the discovery of symbolic models that capture nonlinear and dynamical relationships in social science datasets. By extending neuro-symbolic methods to find compact functions and differential equations in noisy and longitudinal data, we show that our system can be used to discover interpretable models from real-world data in economics and sociology. Augmenting existing workflows with symbolic regression can help uncover novel relationships and explore counterfactual models during the scientific process. We propose that this AI-assisted framework can bridge parametric and non-parametric models commonly employed in social science research by systematically exploring the space of nonlinear models and enabling fine-grained control over expressivity and interpretability.

Sihao Huang (2022): Taking Stock of the Role of Complexity in Social Science - SocArXiv

The triumph of the natural sciences has inspired many attempts to transfer its methods to social inquiry. Although most have fallen short, there has been a handful of notable successes like empiricism and statistical theory. Complexity science is a new contender in this line of epistemologies, claiming to bring a new anti-reductionist framework for reasoning about social and political systems. There is, however, no lack of criticism over its lack of rigor and substantive novelty. Being a relatively new field, I argue that instead of debating what complexity currently is, it may be more productive to examine what social science needs to advance its capabilities in building robust theories. This essay considers three challenges to positive theory-building – data sparsity, complexity, and normativity – and discusses how a theory of complexity can be useful in addressing these issues.

Sihao Huang, Benjamin Lienhard, and the team at EQuS (2021): Microwave Package Design for Superconducting Quantum Processors - PRX Quantum 2, 020306

Solid-state qubits with transition frequencies in the microwave regime, such as superconducting qubits, are at the forefront of quantum information processing. However, high-fidelity, simultaneous control of superconducting qubits at even a moderate scale remains a challenge, partly due to the complexities of packaging these devices. Here, we present an approach to microwave package design focusing on material choices, signal line engineering, and spurious mode suppression. We describe design guidelines validated using simulations and measurements used to develop a 24-port microwave package. Analyzing the qubit environment reveals no spurious modes up to 11 GHz. The material and geometric design choices enable the package to support qubits with lifetimes exceeding 350μs. The microwave package design guidelines presented here address many issues relevant for near-term quantum processors.

Sihao Huang (2020): Towards Multicellular Biological Deep Neural Nets Based on Transcriptional Regulation - arXiv:1912.11423

Artificial neurons built on synthetic gene networks have potential applications ranging from engineering cellular decision-making to bioreactor regulation. Due to the high information throughput of molecular systems, they provides an interesting candidate for biologically-based supercomputing and analog simulations of traditionally intractable problems. In this paper, we propose an architecture for constructing multicellular neural networks and programmable nonlinear systems. We design an artificial neuron based on gene regulatory networks and optimize its dynamics for modularity. Using gene expression models, we simulate its ability to perform arbitrary linear classifications from multiple inputs. Finally, we construct a two-layer neural network to demonstrate scalability and nonlinear decision boundaries and discuss future directions for utilizing uncontrolled neurons in computational tasks.

Sihao Huang and Haowen Lin (2018): Fully Optical Spacecraft Communications: 8Mb/s LED Visible Light Downlink with Deep Learning Error Correction - IEEE Aerospace and Electronic Systems, 33 (4)

Free space optical communication techniques have seen significant advancements with multiple missions expected to fly in the near future. Existing methods require high pointing accuracies, drastically driving up overall system cost. Based on recent developments in LED-based visible light communication (VLC) and in-orbit experiments, we propose a new optical communication system utilizing a VLC downlink and a high throughput, omnidirectional photovoltaic cell receiver system. By performing error-correction via deep learning methods and by utilizing phase-delay interference, the system is able to deliver data rates that match those of traditional laser-based solutions. A prototype of the proposed system has been constructed, demonstrating the scheme to be a feasible alternative to laser-based methods. This creates an opportunity for the full scale development of optical communication techniques on small spacecraft as a backup telemetry beacon or as a high throughput link.