Stefaan Verhulst

Report by Epoch: “What will happen if AI scaling persists to 2030? We are releasing a report that examines what this scale-up would involve in terms of compute, investment, data, hardware, and energy. We further examine the future AI capabilities this scaling will enable, particularly in scientific R&D, which is a focus for leading AI developers. We argue that AI scaling is likely to continue through 2030, despite requiring unprecedented infrastructure, and will deliver transformative capabilities across science and beyond.

Scaling is likely to continue until 2030: On current trends, frontier AI models in 2030 will require investments of hundreds of billions of dollars, and gigawatts of electrical power. Although these are daunting challenges, they are surmountable. Such investments will be justified if AI can generate corresponding economic returns by increasing productivity. If AI lab revenues keep growing at their current rate, they would generate returns that justify hundred-billion-dollar investments in scaling.

Scaling will lead to valuable AI capabilities: By 2030, AI will be able to implement complex scientific software from natural language, assist mathematicians formalising proof sketches, and answer open-ended questions about biology protocols. All of these examples are taken from existing AI benchmarks showing progress, where simple extrapolation suggests they will be solved by 2030. We expect AI capabilities will be transformative across several scientific fields, although it may take longer than 2030 to see them deployed to full effect…(More)”.

Paper by Jed Sundwall: “The COVID-19 pandemic revealed a striking lack of global data coordination among public institutions. While national governments and the World Health Organization struggled to collect and share timely information, a small team at Johns Hopkins University built a public dashboard that became the world’s most trusted source of COVID-19 data. Their work showed that planetary-scale data infrastructure can emerge quickly when practitioners act, even without formal authority. It also exposed a deeper truth: we cannot build global coordination on data without shared standards, yet it is unclear who gets to define what those standards might be.

This paper examines how the World Wide Web has created the conditions for useful data standards to emerge in the absence of any clear authority. It begins from the premise that standards are not just technical artefacts but the “connective tissue” that enables collaboration across institutions. They extend language itself, allowing people and systems to describe the world in compatible terms. Using the history of the web, the paper traces how small, loosely organised groups have repeatedly developed standards that now underpin global information exchange…(More)”.

Report by the Friedrich Naumann Foundation: “Despite the rapid global advancement of artificial intelligence, a significant disparity exists in its accessibility and benefits, disproportionately affecting low- and middle-income countries facing distinct socioeconomic hurdles. This policy paper compellingly argues for immediate, concerted international and national initiatives focused on strategic investments in infrastructure, capacity development, and ethical governance to ensure these nations can leverage AI as a powerful catalyst for equitable and sustainable progress…(More)”.

Blog by Beth Noveck: “Officials in Hamburg had long struggled with the fact that while citizens submitted thousands of comments on planning projects, only a fraction could realistically be read and processed. Making sense of feedback from a single engagement could once occupy five full-time employees for more than a week and chill any desire to do a follow-up conversation. Learn about how Hamburg built its own open source artificial intelligence to make sense of citizen feedback on a scale and speed that was once unimaginable…The Digital Participation System (DIPAS) is Hamburg, Germany’s integrated digital participation platform, designed to let residents contribute ideas, comments, and feedback on urban development projects online or in workshops. It combines mapping, document sharing, and discussion tools so that citizens can engage directly with concrete plans for their neighborhoods. 

City officials had long struggled with the fact that while citizens submitted thousands of comments on planning projects, only a fraction could realistically be read and processed.

“We take the promise of participation seriously,” explained Claudius Lieven, one of DIPAS’s creators in Hamburg’s Ministry of Urban Development and Housing. “If people contribute, their collective intelligence must count. But with so many inputs, we simply couldn’t keep up.” 

Making sense of feedback from a single engagement could once occupy five full-time employees for more than a week and chill any desire to do a follow-up conversation. 

As a result, Lieven and his team spent three years integrating AI into the open-source system to make the new analytics toolbox more useful and the government more responsive. They combined the fine-tuning of Facebook’s advanced open-source language models LLaMA and RoBERTa with topic modeling and geodata integration. 

With AI, DIPAS can cluster and summarize thousands of comments and distinguish between a “major position” about the current situation (for example, “The bike path is unsafe”) and an “idea” proposing what should be done (“The bike path should have better lighting”)…(More)”.

Article by Aivin Solatorio: “As artificial intelligence (AI) becomes a new gateway to development data, a quiet but significant risk has emerged. Large language models (LLMs) can now summarize reports, answer data queries, and interpret indicators in seconds. But while these tools promise convenience, they also raise a fundamental question: How can we ensure that the numbers they produce remain true to the official data they claim to represent?

AI access does not equal data integrity

Many AI systems today use retrieval-augmented generation (RAG), a technique that feeds models with content from trusted sources or databases. While it is widely viewed as a safeguard against hallucinations, it does not eliminate them. Even when an AI model retrieves the correct data, it may still generate outputs that deviate from it. It might round numbers to sound natural, merge disaggregated values, or restate statistics in ways that subtly alter their meaning. These deviations often go unnoticed because the AI still appears confident and precise to the end user.

Developers often measure such errors through evaluation experiments (or “evals”), reporting aggregate accuracy rates. But those averages mean little to a policymaker, journalist, or citizen interacting with an AI tool. What matters is not whether the model is usually correct, but whether the specific number it just produced is faithful to the official data. 

Where Proof-Carrying Numbers come in

Proof-Carrying Numbers (PCN), a novel trust protocol developed by the AI for Data – Data for AI team, addresses this gap. It introduces a mechanism for verifying numerical faithfulness — that is, how closely the AI’s numbers match the trusted data they are based on — in real time.

Here’s how it works:

• The data passed to the LLM must include a claim identifier and a policy that defines acceptable behavior (e.g., exact match required, rounding allowed, etc.). 
• The model is instructed to follow the PCN protocol when generating numbers based on that data.
• Each numeric output is checked against the reference data on which it was conditioned.
• If the result satisfies the policy, PCN marks it as verified [✓].
• If it deviates, PCN flags it for review [⚠️]. 
• Any numbers produced without explicit marks are assumed unverified and should be treated cautiously.

This is a fail-closed mechanism, a built-in safeguard that errs on the side of caution. When faithfulness cannot be proven, PCN does not assume correctness; instead, it makes that failure visible. This feature changes how users interact with AI: instead of trusting responses blindly, they can immediately see whether a number aligns with official data…(More)”.

Paper by Margaret Hughes, et al: “Communities frequently report sending feedback “into a void” during community engagement processes like neighborhood planning, creating a critical disconnect between public input and decision-making. Voice to Vision addresses this gap with a sociotechnical system that comprises three integrated components: a flexible data architecture linking community input to planning outputs, a sensemaking interface for planners to analyze and synthesize feedback, and a community-facing platform that makes the entire engagement process transparent. By creating a shared information space between stakeholders, our system demonstrates how structured data and specialized interfaces can foster cooperation across stakeholder groups, while addressing tensions in accessibility and trust formation. Our CSCW demonstration will showcase this system’s ability to transform opaque civic decision-making processes into collaborative exchanges, inviting feedback on its potential applications beyond urban planning…(More)”.

Book by Maximilian Kasy: “AI is inescapable, from its mundane uses online to its increasingly consequential decision-making in courtrooms, job interviews, and wars. The ubiquity of AI is so great that it might produce public resignation—a sense that the technology is our shared fate.
 
As economist Maximilian Kasy shows in The Means of Prediction, artificial intelligence, far from being an unstoppable force, is irrevocably shaped by human decisions—choices made to date by the ownership class that steers its development and deployment. Kasy shows that the technology of AI is ultimately not that complex. It is insidious, however, in its capacity to steer results to its owners’ wants and ends. Kasy clearly and accessibly explains the fundamental principles on which AI works, and, in doing so, reveals that the real conflict isn’t between humans and machines, but between those who control the machines and the rest of us.
 
The Means of Prediction offers a powerful vision of the future of AI: a future not shaped by technology, but by the technology’s owners. Amid a deluge of debates about technical details, new possibilities, and social problems, Kasy cuts to the core issue: Who controls AI’s objectives, and how is this control maintained? The answer lies in what he calls “the means of prediction,” or the essential resources required for building AI systems: data, computing power, expertise, and energy. As Kasy shows, in a world already defined by inequality, one of humanity’s most consequential technologies has been and will be steered by those already in power.
 
Against those stakes, Kasy offers an elegant framework both for understanding AI’s capabilities and for designing its public control. He makes a compelling case for democratic control over AI objectives as the answer to mounting concerns about AI’s risks and harms. The Means of Prediction is a revelation, both an expert undressing of a technology that has masqueraded as more complicated and a compelling call for public oversight of this transformative technology…(More)”.

Paper by Rosa Vicari: “…Misinformation significantly challenges disaster risk management by increasing risks and complicating response efforts. This technical note introduces a methodology toolbox designed to help policy makers, decision makers, practitioners, and scientists systematically assess, prevent, and mitigate the risks and impacts of misinformation in disaster scenarios. The methodology consists of eight steps, each offering specific tools and strategies to help address misinformation effectively. The process begins with defining the communication context using PESTEL analysis and Berlo’s communication model to assess external factors and information flow. It then focuses on identifying misinformation patterns through data collection and analysis using advanced AI methods. The impact of misinformation on risk perceptions is assessed through established theoretical frameworks, guiding the development of targeted strategies. The methodology includes practical measures for mitigating misinformation, such as implementing AI tools for prebunking and debunking false information. Evaluating the effectiveness of these measures is crucial, and continuous monitoring is recommended to adapt strategies in real-time. Ethical considerations are outlined to ensure compliance with international laws and data privacy regulations. The final step emphasizes managerial aspects, including clear communication and public education, to build trust and promote reliable information sources. This structured approach provides practical insights for enhancing disaster response and reducing the risks associated with misinformation…(More)”.

Article by Victor Dominello: “Every form tells a story. It reveals how government sees itself, how it interacts with people, and who really holds the power.

When you fill out a form on paper, online, or through an app, you’re not just providing information. You’re entering a contract of trust. You’re revealing how much the system values your time, your data, and your dignity.

Forms are not paperwork. They’re the DNA of government. They show who controls the data. They show how much friction citizens face in accessing basic services. And as forms have evolved, so too has government, from paper-based bureaucracy to intelligent, anticipatory systems.

Government 0.0: The paper-form era

Before 1990, everything started with paper. Licences, registrations, applications: all handwritten, stamped, and filed away.

In my early years as a lawyer, I remember attending property settlements with clients clutching paper certificates of title and mortgage documents. They waited their turn in long queues at government offices. Every transaction required physical signatures, physical stamps, and physical handovers…Now we stand on the threshold of Government 4.0: the form-less age.

This is the era of mainstream AI agents. Intelligent systems that talk to each other, share verified data securely, and act on our behalf.

In this world, humans never see a form.

A birth automatically triggers relevant services. A job loss activates training and income support. A licence renewal happens invisibly in the background.

Forward-looking nations are already designing for this future. They’re building systems where services become proactive and anticipatory…(More)”.

Article by Steve MacFeely: “On 1 May 2025, the United Nations (UN) Commission for Science and Technology for Development (CSTD) held the first meeting of a new Multi-Stakeholder Working Group on Data Governance at All Levels in Geneva. Less than three weeks later, the UN Statistics Commission held the first (online) meeting of their new Working Group on Data Governance. Only a fortnight later, on 8 June 2025, the UNESCO Broadband Commission launched their new Data Governance Toolkit at the World Summit on the Information Society (tinyurl.com/fcnzu4ha).

What triggered this flurry of data governance activity at the UN? This article sets out some of the key events and work that led up to this eruption of activity and attempts to explain why it is important…(More)”.