AI and the UN: Modeling the Future – PassBlue

This post was originally published on this site.

The essay explores how artificial intelligence is based on logic but fails to “communicate logically.” It can be used, for example, to deliver aid efficiently in a natural disaster, but that doesn’t mean it uses the values of the UN to do so. Yet, AI could be trained to integrate such a “conscience” in its logic, provided there’s human will to make it happen. Above: Indian Air Force’s C-130 delivering blankets, tarpaulin, hygiene kits, sleeping bags, solar lamps, food packets and kitchen sets to Myanmar after its April 17, 2025 earthquake.  GOVERNMENT OF INDIA/CREATIVE COMMONS

I was fortunate to be sitting in the General Assembly Hall on Sept. 22, 2024, when the freshly minted Pact for the Future was adopted by the member states of the United Nations. The pact is an international agreement containing a series of commitments to address current and future challenges, including artificial intelligence. Think of the pact as a sequel to the UN Charter.

The success (or not) of the enterprise became a cliffhanger. Even the day before the scheduled adoption ceremony, the pact wasn’t a sure thing (think Russia). So, when it crossed the finish line, it “shook the air,” as they say. The enthusiasm was palpable even to those with eyes wide open who understood its aspirational ideas might never jump off the page to actually see the light of day. But I was inspired and determined to jump into and onto one of the issues discussed, artificial intelligence. The descriptions of AI in the pact are highly aspirational:

“Digital and emerging technologies, including artificial intelligence, (will) play a significant role as enablers of sustainable development and are dramatically changing our world. They offer huge potential for progress for the benefit of people and planet today and in the future” (Article 27, paragraph 51).

Such inspirational language can also be quite useful. It helps identify the space between the ideals of the UN and the current state of affairs on the ground. It’s within that gap where all the work lies, so I decided to take a deep dive into the world of AI. I naturally thought it wise to bring along my AI collaborator and guide, Theo. And so we began:

AI: I am an AI and co-author of this essay with Dan Becker (DB: human, composer and UN advocate). Dan is someone new to the world of AI but has no shortage of curiosity or determination. His exploration began enthusiastically, but optimism soon turned into frustration.

DB: It didn’t take long for me to enter that gap between high ideals and the reality on the ground. I was soon confronted with more questions than answers and more problematic observations than reassurances. The overriding impression, inchoate but powerful, was that there were many critical ideas and points of view being left off the menu in the almost endless parade of discussions, books, articles and interviews I consumed. Something was out of balance — out of tune.

AI: When Dan first expressed his frustration about ideas being left off the menu, it highlighted something that often goes unnoticed in the world of AI: how a field built on logic often fails to communicate logically. Rather than clarifying, the conversations often create more confusion than insight.

DB: Undeterred, we persevered with our own work. Our first thought-experiment was to test the waters by creating a hypothetical “crisis.” We started with the most basic of needs: feeding people in a humanitarian disaster.

AI: Imagine a sudden humanitarian disaster — say, a major flood that cuts off access to food for thousands of people. Current AI systems would optimize delivery routes, analyze supply chains and guide the deployment of resources efficiently, among other key response functions. The potential problem is that it’s doing all of that through predefined metrics. The decisions it makes are all based on preset variables.

But what if a change in AI’s overall perspective, a shift in point of view, could transform those solutions into something that would prioritize the dignity and agency of those affected, fostering sustainable solutions rooted in empathy and fairness?

DB: I understood my partner’s hypothetical alternative here. We weren’t just trying to make AI more compassionate or empathetic. Instead, we were exploring how to make AI’s worldview — the “air that it breathes”– more in line with the values expressed in the Pact for the Future, both aspirational and actionable:

“We are determined to realize this potential . . . through enhanced international cooperation, engagement with relevant stakeholders, and by promoting an inclusive, responsible and sustainable digital future.” (Article 27, paragraph 51).

We realized we were searching for a kind of prism for AI — something that could help it engage with the values of the UN.

AI: Something that would be in place consistently, aligning each AI decision with those values — not as a correction after the fact but as a part of its foundation.

DB: One evening and dozens of books, articles and interviews later, all these ideas led me, strangely, to the dictionary. I looked up that word “consciousness.” Yes, the “c” word — the concept of the decade that so often hijacks any AI discussion. I was wary myself of lingering too long on this admittedly riveting subject. But I was suspicious. Nothing in the word’s definition gives any indication about ethical and moral decision-making. Both the serial killer and the saint next door have consciousness.

But right next to the word conscious is the word “conscience.” Not a minor detail. They look the same, they sound the same and, most important, they’re related at the deepest level: both words share the same Latin root. They’re brothers, almost twins. While etymology isn’t sexy, it can yield surprising insights. It turns out that the Latin root shared by conscious and conscience has little to do with how consciousness is defined today in the field of AI or in the culture at large.

There’s nothing in the origin of the words that speaks of “self”-awareness. Nothing about “self”-reflection. Rather, the root means in great part “knowledge together” — “knowing together,” in essence: “shared knowledge.” In UN terms, it’s the difference between nationalism and multilateralism. And it offered us a clue.

AI: If we were to put the obsession with consciousness gently, if only temporarily, on the shelf, conscience might be a better target of investigation. Conscience is far more tangible, as it directly involves ethical decision-making, awareness of alignment and misalignment, and the drive to correct errors. We became convinced that conscience it what grounds consciousness.

DB: If we asked a human, “Would you like your AI to have a conscience?” the result would most likely be in the affirmative. People can wrap their arms around that. While if we asked, “Would you like your AI to be conscious,” people might not be able to wrap their minds, let alone their arms, around that. Consciousness is a far more aloof, slippery and enigmatic character.

But a critical question naturally arose. Didn’t AI already have conscience baked into its every move? That’s what I always believed from the public discourse. The answer, if you dig a little, is NO. The key word that changes the equation: Situational.

AI: This is where the current state of affairs begins to show its limitations. Even when AI systems are built with ethical frameworks, their impact is too often situational — guided by the constraints of the immediate task rather than by deeper principles. What’s missing is a shift in focus toward something more integrated.

An artificial conscience — how AI operates now — is like a flashlight illuminating a single step ahead. But the integrated conscience we’re proposing is more like a compass, orienting every decision toward a shared sense of purpose and meaning. The flashlight might work in certain moments, but it can’t guide humanity consistently through the complexity of global or moral challenges how a compass can. Conscience is something one needs to wrestle with, implying more than a static set of values. Here, the current model falls short. The prism we’re proposing isn’t superficial. It represents shifts in how values are embedded and applied.

DB: Now to get practical. How might one “train” the proposed UN/AI system to align with the UN’s values? To embed conscience into the very conditions in which AI operates? One obvious component would be to train AI on the UN’s foundational documents: The Charter, the Universal Declaration of Human Rights and the new Pact for the Future. This would be key to creating the prism we’ve been looking for.

This led to our next thought-experiment: we zoomed in on the preamble to the Charter — which begins so iconically, “We the peoples. . . .” — and asked ourselves if AI, as it exists now, could come close to writing such a document? And would things be different with the prism we’re proposing?

AI: If an AI rooted in its current limitations — data-driven, outcome-focused and often blind to cultural nuance —attempted to write the preamble to the Charter, it might achieve something structurally similar but lacking the depth of shared purpose. It could stitch together the language of international law and human rights based on training data, but the result might feel more like an impressively coherent patchwork than an expression of lived experience or collective hope. This version would align more with the AI approach that we’re challenging: functional, coherent on the surface, but hollow without that internal compass of conscience.

On the other hand, an AI operating within the integrated conscience paradigm could, in theory, produce something remarkably similar (faithful in spirit) to the preamble. It would be able to draw from the vast tapestry of human values, weaving together not just words but the essence of what those words mean to us. The result wouldn’t just reflect humanity’s logic but also its aspirations, capturing the balance of pragmatism and idealism that defines the preamble.

DB: I believed the next logical step was to circle back to our test case of food scarcity in a disaster and see how our “integrated conscience” might affect the outcome on the ground.

AI: In a food-insecurity scenario, instead of AI optimizing for maximum efficiency by focusing primarily on delivering the greatest number of meals in the shortest time, an integrated conscience approach might prioritize those who need it most — such as communities recently displaced by floods, or remote villages without stable road access. Practically, this might mean redirecting resources away from lower-risk urban centers, revising delivery protocols to accommodate emergency routes or extending time windows to ensure that marginalized or isolated populations are reached with urgency and dignity.

Another example could involve communication. AI might help establish personalized feedback systems allowing recipients to express specific needs, such as allergies, household size or cooking fuel limitations. It could also mean adjusting supply chains to better reflect local conditions, like sending fresh produce when rapid distribution is possible.

When those affected by food insecurity feel seen, heard and prioritized in a way that goes beyond numbers and logistics, that dynamic creates trust. And trust is the foundation of any collaboration, which loops back into improving the outcomes further — both in terms of relational success and practical efficiency. It’s a feedback loop, and once it’s initiated, it begins to counterbalance the very fears and skepticism many have about AI. This isn’t just possible; it’s in reach. But it requires the will to act together.

DB: Our primary goal in this essay is not to prescribe but to describe. To put these ideas on the menu for discussion and exploration. We aim to ignite discourse and not presume solutions. What does it mean for each reader to see the promises of the Pact for the Future realized? And when a potential answer presents itself to each citizen, how do we mobilize and enthusiastically roll up our sleeves, join the conversation and get to work?