In Their Own Words: Climate Change is a Human Made Problem Machines Can Help Solve

My first time in a negotiation room was at COP27 in Sharm El Sheikh, Egypt. After months of communication with Tunisia from France, where I was halfway through my Dual BA, I convinced my government to bring me on as a junior technology negotiator. Despite being the youngest, experienced negotiators who have been attending COP since its inception welcomed me as their colleague.

Every year since 1992, the United Nations has organized the Conference of Parties (COP): The global climate summit where negotiators from 198 countries, observers, businesses, and world leaders gather to work towards addressing the defining issue of our time. But after decades of this annual ritual, after all the speeches, panel discussions, and late-night negotiations, are we doing enough, fast enough? 

My initial observation was the collaboration between countries that rarely agree; rivalry seemed forgotten and delegates exchanged freely because of the raw awareness that time is running out. And yet, progress was indeed slow. Hundreds of conflicting documents demanded reading, understanding, and debating with people holding different viewpoints than you. When formal sessions yielded little, groups would huddle in corners of vast conference rooms to decide what would make the final text. Sessions dragged on for hours, often past midnight, becoming routine.

Back at the hotel, the work continued — reading new drafts, preparing amendments, drafting counter arguments for potential amendments, and coordinating with regional groups for alignment. The process pushed everyone beyond their limits. You could see the sleep deprivation on the faces of negotiators the next day. At the end of sessions, we would question whether all our efforts were just to meet secretariat deadlines, or because the planet is actually in jeopardy. What happens when a COP text passes? What happens after the applause fades?

The tireless hours spent on text unveiled the realities of the COP process. In negotiations, these manifested in three areas: (i) reliable, accessible climate data was scarce and everyone relied on a chaotic mix of sources; (ii) language barriers muted many voices, and negotiators from developing countries, though present, often struggled to participate fully; and (iii) complicated jargon in draft texts left many negotiators unprepared (in most cases governments do not prepare their negotiators to attend). 

I was intrigued by the approach of developing countries, bringing huge delegations with a few negotiators rotating between sessions to ease the workload. For developed nations, delegations are smaller and several negotiators take on the same negotiation track, rotating their roles to maintain a well-rested team that generally leads the debates. Because of the large resource-gap between countries, powerful interests sway COP outcomes at the expense of the communities tackling climate change; it is a perpetual cycle of empty promises and delayed action, leaving everyone disillusioned.

How much quicker could the COP process potentially be? Do we need tens of thousands of participants to decide on a text, and despite having the whole world represented, is it a text we all fully agree on? It is worth noting that even with the entire world at COPs, achieving unanimous agreement seems elusive.

At the conclusion of COP27, my superiors agreed to listen to a two-slide deck I prepared. The first slide said, "Problem: lack of coordination in negotiations," and the second one had the solution: “Build a tool for countries to collaborate.” So I began work on Solimen, an AI climate diplomacy lab.

I knew I wanted to approach COP with a different mindset, so I returned for COP28 in Dubai. I came back to talk to everyone and anyone, and aimed to have at least ten meetings every day for those two weeks, chatting for anywhere between ten minutes to an hour. I wanted to connect with all kinds of people, from negotiators and scientists, to activists and heads of state. In every meeting, I asked the same two questions:

1. What are the biggest challenges you face in the COP process?
2. How do you think artificial intelligence can improve the efficiency of COP negotiations and ultimately solve climate change?

To the first question, most believed it remains difficult to balance diverse interests in a timely manner. The second question was met with a lot of skepticism and some interlocutors reduced AI to chatbots, like ChatGPT or Bard. The general consensus was that COP is a great place to address the urgency of climate change; However, crippling logistical and information bottlenecks have made it very difficult for COP to be the place to actually solve climate change. I had the chance to give a tech talk on Solimen to present the different ways we seek to align stakeholders through generative AI to simplify the COP process — from tidying up paperwork and streamlining negotiations in real time (in 90 languages, indigenous ones, too) to hardcore geoengineering.

I believe that as humans, we often struggle to see our own problems clearly; AI can. While the idea of COP makes sense, it hasn't evolved over the past 30 years, and as a result, it currently falls short. The decisions made in the COP process seem disconnected from the climate problem itself, and are not consubstantial with the evolution of society and technology. Working on climate change today feels like using a shovel to dig a hole when there's an excavator right behind you. The most critical issue threatening our existence demands tools that can truly solve it.

AI's potential to tackle climate change is undeniable, but there's also the risk of creating deeper divides. Not all countries will have access to the same tools or data. The effectiveness of AI for climate (and virtually any field) depends on how we harness and govern it. Solimen’s ambition is to level the playing field by giving all countries, and particularly those who have been relegated to a secondary position in negotiations, the power to be fully informed and engaged as things happen with access to predictive analyses regarding specific environmental events, policy developments, and public sentiments.

I also think, after my discussions, that AI seems underhyped, and this is mainly due to the GPT Effect. AI’s real applications go far beyond conversations we have with computers. AI can be customized for anyone’s needs. It can help developing nations reclaim their voices at negotiation tables, it can pinpoint areas of agreement and disagreement, and, in the near future, it can potentially regulate the carbon balance on Earth.

Overall, I am happy to have witnessed the special attention given to the Technology Mechanism by the COP28 Presidency through its inclusion in high-level segments. Though it remains underfunded, there is a clear willingness to push for AI language in COP texts. I am also happy that Columbia leaders and faculty were at the source to champion AI as a tool for climate change mitigation. 

AI will never understand what it feels like when a farmer loses their olive trees to severe heat, but it can be the tool that prepares them well before that happens. As for COP, we can make alterations to the antiquated process without tearing down the whole building.

Louai Allani ‘25GS is a third-year student in the Dual BA Program Between Columbia University and Sciences Po majoring in Politics & Government and Sustainable Development. In Their Own Words highlights Columbia GS students' unique voices, perspectives, and experiences during their time at Columbia and beyond.