Agenda Day 3: AI for the Digital Economy

Abstract:
The talk starts with a quick exposition of the modern monetary circuit model (MMCT), considered an integral part of the general dynamic input-output model. Then, it analyzes money creation by the banking system and highlights central banks’ role as an anchor that keeps the financial system together. Particular emphasis is placed on interactions among different systems and foreign exchange. Second, the monetary system’s pain points and frictions are discussed, and it is argued that distributed ledger technology (DLT) can remove some of these frictions. The most recent developments and their potential applications are reviewed. Third, Digital twins (DTs)—virtual models that mirror physical assets, systems, or processes—are discussed in the context of supply chains (SC).

Abstract:
In a world where authenticity and traceability are increasingly critical, establishing reliable links between physical objects and their digital identities has become a foundational challenge. Digital product passports (DPPs) aim to address this need by embedding persistent, verifiable records throughout a product’s lifecycle. Recent advances in AI-based image matching—particularly detector-free, transformer-based models—offer a tag-free solution to this problem by extracting distinctive visual “fingerprints” directly from an object’s surface. These methods enable scalable visual verification without barcodes, chips, or serial numbers, supporting applications in provenance tracking, anti-counterfeiting, and circular economy systems.

Abstract:
his presentation presents findings from a large-scale household survey of over 25,000 U.S. residents designed to disentangle the causal contributions of the channels of monetary policy transmission. Leveraging a randomized controlled trial with information treatments and hypothetical scenarios, the talk shows that monetary tightening reduces household consumption—both in terms of intentions and self-reported behavior. However, the transmission mechanisms differ from those typically embedded in standard macroeconomic models. Specifically, households expect consumer and housing prices to rise following a tightening, anticipating increases in borrowing costs, and higher inflation leads households to reduce consumption, contrary to the intertemporal substitution mechanism implied by the Euler equation.

Abstract:
This talk explores the challenges and solutions for building more reliable AI systems. It first examines key threats and mitigation strategies for achieving robustness in machine learning classification and regression models, then highlights the barriers to aligning large language models when relying solely on simple input–output filtering.

Abstract:
Accurate inflation forecasts support sound policy and stable markets. This presentation adapts the hedged random forest (HRF) to inflation forecasting. Unlike standard random forests, HRF learns how much to trust each tree—including allowing for negative weights—to reduce forecast error. The talk introduces a data-driven procedure that uses past forecast errors to set these weights while accounting for correlations across trees. In an extensive empirical analysis, HRF consistently outperforms the conventional random forest, providing a practical tool for both analysts and policymakers.

Abstract:
The rapid expansion of digital assets and decentralized finance (DeFi) presents unique challenges for risk assessment, demanding models that capture both dynamic on-chain behaviours and contextual signals from external sources. We propose a framework leveraging graph neural networks for transaction analysis and data distillation for sharing of suspicious transactions. Empirical validation is conducted via backtesting on historical adverse events in on-chain behaviour, benchmarking predictive performance, fairness, and interpretability against traditional finance.

Abstract:
Privacy primitives like BaseSAP restore unlinkability, but the digital economy can fail when block producers bias or delay inclusion. This talk introduces a neutrality-aware, agentic AI stack that detects censorship risk in post-Merge PBS, adapts execution through relay diversification, inclusion lists, and local-build fallback, and feeds latency and inclusion signals into trading and payments agents. The talk quantifies how such turbulence compresses anonymity sets, distorts prices, and erodes PnL, and then demonstrates controls that restore autonomy and performance. The outcome is a deployable playbook for censorship-resilient, high-throughput on-chain agentic AI operations.

Abstract:
The talk analyses how quantum computing is poised to break current computational barriers, threatening traditional cryptography through algorithms such as Shor’s, while also providing unprecedented solutions in optimization and cybersecurity. It explores how AI, particularly Quantum Machine Learning (QML), opens new frontiers for threat detection and enhanced data processing efficiency.

Abstract:
The talk presents a system in which nonhuman species, which sustain the global economy but hold no capital, can hold wallets, express interests through AI, and issue payments for verified conservation tasks. AI models infer these interests from ecological and behavioral data, anchored in digital identities and verified via distributed ledgers. This “Interspecies Money” approach introduces a new category of financial instrument: capital issued on behalf of other life. It enables a circular system in which new capital funds frontline ecological data, generates species-level knowledge, distributes equity efficiently across the species divide, mitigates risk, and delivers high-yield returns for species, communities, and investors.

Abstract:
The presentation proposes a new data anonymisation method based on the concept of a quantum feature map. This approach combines a high degree of security with the ability to perform classification tasks directly on anonymised (encrypted) data, achieving equal or even higher accuracy than working with the original plain-text data. This enables important applications in fields such as medicine and finance, where anonymised datasets from different organisations can be combined to improve machine learning outcomes using the aggregated data.

Abstract:
The recent EU regulation, Markets in Crypto Assets Regulation (MiCA), represents significant progress in establishing a multi-jurisdictional framework for crypto-assets, enabling greater consumer participation in the digital assets industry. However, the nascent tokenized assets industry faces several challenges related to standardizing interfaces, functions, and semantic definitions of digitizable assets. Standards are needed to digitally represent off-chain assets that are tokenized, such that there is semantic compatibility among tokens independent of the blockchains upon which the tokens are instantiated. Decentralized registries of metadata information regarding tokenized assets are also needed in order to allow trading parties to obtain assurance of the legal status of the asset prior to the transaction.