Rethinking Dual-Use Technology

Curated on Posted on by Stefaan Verhulst

Article by Artur Kluz and Stefaan Verhulst: “A new concept of “triple use” — where technology serves commercial, defense, and peacebuilding purposes — may offer a breakthrough solution for founders, investors and society to explore….

As a result of the resurgence of geopolitical tensions, the debate about the applications of dual-use technology is intensifying. The core issue founders, tech entrepreneurs, venture capitalists (VCs), and limited partner investors (LPs) are examining is whether commercial technologies should increasingly be re-used for military purposes. Traditionally, the majority of  investors (including limited partners) have prohibited dual-use tech in their agreements. However, the rapidly growing dual-use market, with its substantial addressable size and growth potential, is compelling all stakeholders to reconsider this stance. The pressure for innovations, capital returns and Return On Investment (ROI) is driving the need for a solution. 

These discussions are fraught with moral complexity, but they also present an opportunity to rethink the dual-use paradigm and foster investment in technologies aimed at supporting peace. A new concept of “triple use”— where technology serves commercial, defense, and peacebuilding purposes — may offer an innovative and more positive avenue for founders, investors and society to explore. This additional re-use, which remains in an incipient state, is increasingly being referred to as PeaceTech. By integrating terms dedicated to PeaceTech in new and existing investment and LP agreements, tech companies, founders and venture capital investors can be also required to apply their technology for peacebuilding purposes. This approach can expand the applications of emerging technologies to also include conflict prevention, reconstruction or any humanitarian aspects.

However, current efforts to use technologies for peacebuilding are impeded by various obstacles, including a lack of awareness within the tech sector and among investors, limited commercial interest, disparities in technical capacity, privacy concerns, international relations and political complexities. In the below we examine some of these challenges, while also exploring certain avenues for overcoming them — including approaching technologies for peace as a “triple use” application. We will especially try to identify examples of how tech companies, tech entrepreneurs, accelerators, and tech investors including VCs and LPs can commercially benefit and support “triple use” technologies. Ultimately, we argue, the vast potential — largely untapped — of “triple use” technologies calls for a new wave of tech ecosystem transformation and public and private investments as well as the development of a new field of research…(More)”.

Training LLMs to Draft Replies to Parliamentary Questions

Curated on Posted on by Stefaan Verhulst

Blog by Watson Chua: “In Singapore, the government is answerable to Parliament and Members of Parliament (MPs) may raise queries to any Minister on any matter in his portfolio. These questions can be answered orally during the Parliament sitting or through a written reply. Regardless of the medium, public servants in the ministries must gather materials to answer the question and prepare a response.

Generative AI and Large Language Models (LLMs) have already been applied to help public servants do this more effectively and efficiently. For example, Pair Search (publicly accessible) and the Hansard Analysis Tool (only accessible to public servants) help public servants search for relevant information in past Parliamentary Sittings relevant to the question and synthesise a response to it.

The existing systems draft the responses using prompt engineering and Retrieval Augmented Generation (RAG). To recap, RAG consists of two main parts:

  • Retriever: A search engine that finds documents relevant to the question
  • Generator: A text generation model (LLM) that takes in the instruction, the question, and the search results from the retriever to respond to the question
A typical RAG system. Illustration by Hrishi Olickel, taken from here.

Using a pre-trained instruction-tuned LLM like GPT-4o, the generator can usually generate a good response. However, it might not be exactly what is desired in terms of verbosity, style and writing prose, and additional human post-processing might be needed. Extensive prompt engineering or few-shot learning can be done to mold the response at the expense of incurring higher costs from using additional tokens in the prompt…(More)”

Increasing The “Policy Readiness” Of Ideas

Curated on Posted on by Stefaan Verhulst

Article by Tom Kalil: “NASA and the Defense Department have developed an analytical framework called the “technology readiness level” for assessing the maturity of a technology – from basic research to a technology that is ready to be deployed.  

A policy entrepreneur (anyone with an idea for a policy solution that will drive positive change) needs to realize that it is also possible to increase the “policy readiness” level of an idea by taking steps to increase the chances that a policy idea is successful, if adopted and implemented.  Given that policy-makers are often time constrained, they are more likely to consider ideas where more thought has been given to the core questions that they may need to answer as part of the policy process.

A good first step is to ask questions about the policy landscape surrounding a particular idea:

1. What is a clear description of the problem or opportunity?  What is the case for policymakers to devote time, energy, and political capital to the problem?

2. Is there a credible rationale for government involvement or policy change?  

Economists have developed frameworks for both market failure (such as public goods, positive and negative externalities, information asymmetries, and monopolies) and government failure (such as regulatory capture, the role of interest groups in supporting policies that have concentrated benefits and diffuse costs, limited state capacity, and the inherent difficulty of aggregating timely, relevant information to make and implement policy decisions.)

3. Is there a root cause analysis of the problem? …(More)”.

AI: a transformative force in maternal healthcare

Curated on Posted on by Stefaan Verhulst

Article by Afifa Waheed: “Artificial intelligence (AI) and robotics have enormous potential in healthcare and are quickly shifting the landscape – emerging as a transformative force. They offer a new dimension to the way healthcare professionals approach disease diagnosis, treatment and monitoring. AI is being used in healthcare to help diagnose patients, for drug discovery and development, to improve physician-patient communication, to transcribe voluminous medical documents, and to analyse genomics and genetics. Labs are conducting research work faster than ever before, work that otherwise would have taken decades without the assistance of AI. AI-driven research in life sciences has included applications looking to address broad-based areas, such as diabetes, cancer, chronic kidney disease and maternal health.

In addition to increasing the knowledge of access to postnatal and neonatal care, AI can predict the risk of adverse events in antenatal and postnatal women and their neonatal care. It can be trained to identify those at risk of adverse events, by using patients’ health information such as nutrition status, age, existing health conditions and lifestyle factors. 

AI can further be used to improve access to women located in rural areas with a lack of trained professionals – AI-enabled ultrasound can assist front-line workers with image interpretation for a comprehensive set of obstetrics measurements, increasing quality access to early foetal ultrasound scans. The use of AI assistants and chatbots can also improve pregnant mothers’ experience by helping them find available physicians, schedule appointments and even answer some patient questions…

Many healthcare professionals I have spoken to emphasised that pre-existing conditions such as high blood pressure that leads to preeclampsia, iron deficiency, cardiovascular disease, age-related issues for those over 35, various other existing health conditions, and failure in the progress of labour that might lead to Caesarean (C-section), could all cause maternal deaths. Training AI models to detect these diseases early on and accurately for women could prove to be beneficial. AI algorithms can leverage advanced algorithms, machine learning (ML) techniques, and predictive models to enhance decision-making, optimise healthcare delivery, and ultimately improve patient outcomes in foeto-maternal health…(More)”.

How to build a Collective Mind that speaks for humanity in real-time

Curated on Posted on by Stefaan Verhulst

Blog by Louis Rosenberg: “This begs the question — could large human groups deliberate in real-time with the efficiency of fish schools and quickly reach optimized decisions?

For years this goal seemed impossible. That’s because conversational deliberations have been shown to be most productive in small groups of 4 to 7 people and quickly degrade as groups grow larger. This is because the “airtime per person” gets progressively squeezed and the wait-time to respond to others steadily increases. By 12 to 15 people, the conversational dynamics change from thoughtful debate to a series of monologues that become increasingly disjointed. By 20 people, the dialog ceases to be a conversation at all. This problem seemed impenetrable until recent advances in Generative AI opened up new solutions.

The resulting technology is called Conversational Swarm Intelligence and it promises to allow groups of almost any size (200, 2000, or even 2 million people) to discuss complex problems in real-time and quickly converge on solutions with significantly amplified intelligence. The first step is to divide the population into small subgroups, each sized for thoughtful dialog. For example, a 1000-person group could be divided into 200 subgroups of 5, each routed into their own chat room or video conferencing session. Of course, this does not create a single unified conversation — it creates 200 parallel conversations…(More)”.

Doing science backwards

Curated on Posted on by Stefaan Verhulst

Article by Stuart Ritchie: “…Usually, the process of publishing such a study would look like this: you run the study; you write it up as a paper; you submit it to a journal; the journal gets some other scientists to peer-review it; it gets published – or if it doesn’t, you either discard it, or send it off to a different journal and the whole process starts again.

That’s standard operating procedure. But it shouldn’t be. Think about the job of the peer-reviewer: when they start their work, they’re handed a full-fledged paper, reporting on a study and a statistical analysis that happened at some point in the past. It’s all now done and, if not fully dusted, then in a pretty final-looking form.

What can the reviewer do? They can check the analysis makes sense, sure; they can recommend new analyses are done; they can even, in extreme cases, make the original authors go off and collect some entirely new data in a further study – maybe the data the authors originally presented just aren’t convincing or don’t represent a proper test of the hypothesis.

Ronald Fisher described the study-first, review-later process in 1938:

To consult the statistician [or, in our case, peer-reviewer] after an experiment is finished is often merely to ask him to conduct a post mortem examination. He can perhaps say what the experiment died of.

Clearly this isn’t the optimal, most efficient way to do science. Why don’t we review the statistics and design of a study right at the beginning of the process, rather than at the end?

This is where Registered Reports come in. They’re a new (well, new-ish) way of publishing papers where, before you go to the lab, or wherever you’re collecting data, you write down your plan for your study and send it off for peer-review. The reviewers can then give you genuinely constructive criticism – you can literally construct your experiment differently depending on their suggestions. You build consensus—between you, the reviewers, and the journal editor—on the method of the study. And then, once everyone agrees on what a good study of this question would look like, you go off and do it. The key part is that, at this point, the journal agrees to publish your study, regardless of what the results might eventually look like…(More)”.

AI-Ready FAIR Data: Accelerating Science through Responsible AI and Data Stewardship

Curated on Posted on by Stefaan Verhulst

Article by Sean Hill: “Imagine a future where scientific discovery is unbound by the limitations of data accessibility and interoperability. In this future, researchers across all disciplines — from biology and chemistry to astronomy and social sciences — can seamlessly access, integrate, and analyze vast datasets with the assistance of advanced artificial intelligence (AI). This world is one where AI-ready data empowers scientists to unravel complex problems at unprecedented speeds, leading to breakthroughs in medicine, environmental conservation, technology, and more. The vision of a truly FAIR (Findable, Accessible, Interoperable, Reusable) and AI-ready data ecosystem, underpinned by Responsible AI (RAI) practices and the pivotal role of data stewards, promises to revolutionize the way science is conducted, fostering an era of rapid innovation and global collaboration…(More)”.

Real Chaos, Today! Are Randomized Controlled Trials a good way to do economics?

Curated on Posted on by Stefaan Verhulst

Article by Maia Mindel: “A few weeks back, there was much social media drama about this a paper titled: “Social Media and Job Market Success: A Field Experiment on Twitter” (2024) by Jingyi Qiu, Yan Chen, Alain Cohn, and Alvin Roth (recipient of the 2012 Nobel Prize in Economics). The study posted job market papers by economics PhDs, and then assigned prominent economists (who had volunteered) to randomly promote half of them on their profiles(more detail on this paper in a bit).

The “drama” in question was generally: “it is immoral to throw dice around on the most important aspect of a young economist’s career”, versus “no it’s not”. This, of course, awakened interest in a broader subject: Randomized Controlled Trials, or RCTs.

R.C.T. T.O. G.O.

Let’s go back to the 1600s – bloodletting was a common way to cure diseases. Did it work? Well, doctor Joan Baptista van Helmont had an idea: randomly divvy up a few hundred invalids into two groups, one of which got bloodletting applied, and another one that didn’t.

While it’s not clear this experiment ever happened, it sets up the basic principle of the randomized control trial: the idea here is that, to study the effects of a treatment, (in a medical context, a medicine; in an economics context, a policy), a sample group is divided between two: the control group, which does not receive any treatment, and the treatment group, which does. The modern randomized controlled (or control) trial has three “legs”: it’s randomized because who’s in each group gets chosen at random, it’s controlled because there’s a group that doesn’t get the treatment to serve as a counterfactual, and it’s a trial because you’re not developing “at scale” just yet.

Why could it be important to randomly select people for economic studies? Well, you want the only difference, on average, between the two groups to be whether or not they get the treatment. Consider military service: it’s regularly trotted out that drafting kids would reduce crime rates. Is this true? Well, the average person who is exempted from the draft could be, systematically, different than the average person who isn’t – for example, people who volunteer could be from wealthier families who are more patriotic, or poorer families who need certain benefits; or they could have physical disabilities that impede their labor market participation, or wealthier university students who get a deferral. But because many countries use lotteries to allocate draftees versus non draftees, you can get a group of people who are randomly assigned to the draft, and who on average should be similar enough to each other. One study in particular, about Argentina’s mandatory military service in pretty much all of the 20th century, finds that being conscripted raises the crime rate relative to people who didn’t get drafted through the lottery. This doesn’t mean that soldiers have higher crime rates than non soldiers, because of selection issues – but it does provide pretty good evidence that getting drafted is not good for your non-criminal prospects…(More)”.

Connecting the dots: AI is eating the web that enabled it

Curated on Posted on by Stefaan Verhulst

Article by Tom Wheeler: “The large language models (LLMs) of generative AI that scraped their training data from websites are now using that data to eliminate the need to go to many of those same websites. Respected digital commentator Casey Newton concluded, “the web is entering a state of managed decline.” The Washington Post headline was more dire: “Web publishers brace for carnage as Google adds AI answers.”…

Created by Sir Tim Berners-Lee in 1989, the World Wide Web redefined the nature of the internet into a user-friendly linkage of diverse information repositories. “The first decade of the web…was decentralized with a long-tail of content and options,” Berners-Lee wrote this year on the occasion of its 35th anniversary.  Over the intervening decades, that vision of distributed sources of information has faced multiple challenges. The dilution of decentralization began with powerful centralized hubs such as Facebook and Google that directed user traffic. Now comes the ultimate disintegration of Berners-Lee’s vision as generative AI reduces traffic to websites by recasting their information.

The web’s open access to the world’s information trained the large language models (LLMs) of generative AI. Now, those generative AI models are coming for their progenitor.

The web allowed users to discover diverse sources of information from which to draw conclusions. AI cuts out the intellectual middleman to go directly to conclusions from a centralized source.

The AI paradigm of cutting out the middleman appears to have been further advanced in Apple’s recent announcement that it will incorporate OpenAI to enable its Siri app to provide ChatGPT-like answers. With this new deal, Apple becomes an AI-based disintermediator, not only eliminating the need to go to websites, but also potentially disintermediating the need for the Google search engine for which Apple has been paying $20 billion annually.

The Atlantic, University of Toronto, and Gartner studies suggest the Pew research on website mortality could be just the beginning. Generative AI’s ability to deliver conclusions cannibalizes traffic to individual websites threatening the raison d’être of all websites, especially those that are commercially supported…(More)” 

Why policy failure is a prerequisite for innovation in the public sector

Curated on Posted on by Stefaan Verhulst

Blog by Philipp Trein and Thenia Vagionaki: “In our article entitled, “Why policy failure is a prerequisite for innovation in the public sector,” we explore the relationship between policy failure and innovation within public governance. Drawing inspiration from the “Innovator’s Dilemma,”—a theory from the management literature—we argue that the very nature of policymaking, characterized by myopia of voters, blame avoidance by decisionmakers, and the complexity (ill-structuredness) of societal challenges, has an inherent tendency to react with innovation only after failure of existing policies.  

Our analysis implies that we need to be more critical of what the policy process can achieve in terms of public sector innovation. Cognitive limitations tend to lead to a misperception of problems and inaccurate assessment of risks by decision makers according to the “Innovator’s Dilemma”.  This problem implies that true innovation (non-trivial policy changes) are unlikely to happen before an existing policy has failed visibly. However, our perspective does not want to paint a gloomy picture for public policy making but rather offers a more realistic interpretation of what public sector innovation can achieve. As a consequence, learning from experts in the policy process should be expected to correct failures in public sector problem-solving during the political process, rather than raise expectations beyond what is possible. 

The potential impact of our findings is profound. For practitioners and policymakers, this insight offers a new lens through which to evaluate the failure and success of public policies. Our work advocates a paradigm shift in how we perceive, manage, and learn from policy failures in the public sector, and for the expectations we have towards learning and the use of evidence in policymaking. By embracing the limitations of innovation in public policy, we can better manage expectations and structure the narrative regarding the capacity of public policy to address collective problems…(More)”.