Join on WhatsApp
Join on Telegram
Today, the United Nations launched the Global Dialogue on AI Governance, where for the first time, all countries have a seat at the table of AI. The meeting is being held at the Palexpo convention center in Geneva, Switzerland. Taking place on July 6 and 7, the event gathers different stakeholders: governments, academia, and civil society, to discuss coordinated international approaches to artificial intelligence. This global dialogue comes at a critical time, just days after the Independent International Scientific Panel on AI published its Preliminary Report: “Evidence-based assessment of opportunities, risks and impacts of AI”.
This first-of-its-kind independent panel, composed of 40 scientists and selected from the five UN regions, worked for almost four months before publishing its report. This first evidence-based assessment is an urgent call and guideline for world leaders to help UN member states navigate a rapidly changing technology. Co-chaired by computer scientist Yoshua Bengio and Nobel laureate Maria Ressa, the report balances massive societal opportunities against critical safety and big governance gaps. Urgent, multilateral AI governance is no longer a choice as it was rightly framed by António Guterres, the Secretary-General of the United Nations: “The question is no longer whether AI will transform our world – it already is. The question is whether we will govern this transformation together – or let it govern us”.
One of the key messages of the report is that no single country or stakeholder can govern AI alone. This is a collective effort where all stakeholders should work together for a better use of AI.
The report adopted a “balanced” tone, avoiding pessimistic or optimistic analysis of the real benefits of AI and its potential threats.
AI’s promise meets a governance gap
Nowadays, no one can deny the huge potential benefits of AI. AI has advanced health science and medicine by enhancing diagnostic accuracy, drug discovery, and customizing patient care through the analysis of massive datasets, way faster than humanly possible. The report highlights the findings of AlphaFold, arguably the most revolutionary AI breakthrough in the history of biological science, which predicted the structures of millions of proteins and accelerated drug design and vaccine development in an unprecedented way. Early cancer detection has been one of the most spectacular AI-driven applications in medicine. Traditionally, drug discovery took scientists 10 to 12 years of investigation and work. Generative AI has shortened this to days or weeks.
In global weather forecasting, AI has made it possible to treat this paradigm like any pattern-recognition problem, significantly reducing the computational resources required compared with traditional numerical weather prediction methods.
AI can solve problems in healthcare, climate change, education, and agriculture, but without governance, it risks causing deepened inequality, widespread misinformation, and human rights threats. However, AI’s capabilities are outrunning the world’s ability to measure, govern, or evenly distribute it. AI is not like other technological evolutions that humanity has known until now. The report describes this AI moment as an “inflection point” because AI is the first technology whose widespread adoption has been compressed from years to months. By comparison, the Internet took decades to evolve from a research network in the 1970s – thanks to the pioneering work of Vint Cerf and Bob Kahn – to mass public adoption in the mid-1990s after the introduction of the World Wide Web by Sir Tim Berners-Lee. The Internet needed 15 years to reach one billion users. Generative AI, in contrast, reached hundreds of millions of users within just two months of its launch. The public release of ChatGPT in November 2022 marked what many commentators have called the “ChatGPT moment”. Unlike the Internet, electricity, or other technologies, policymakers have not been able to keep pace when it comes to AI.
Global AI dominance deepens inequality
Furthermore, the UN report warns against the geographical concentration of AI models. The statistics show that the United States-based institutions produced 55.1% of the notable AI models, followed by China at 32.7%, while the rest of the world combined accounted for just 12.1%. This unbalanced distribution and extreme concentration of AI infrastructure in a few countries have severe consequences, leading to environmental injustices and the homogenization of AI models, which erases local cultures and specificity.
Also, the development of major AI models is largely dominated by a handful of private firms with massive computing resources. These companies are dictating their rules on how data should be trained, secured, and deployed. Control over model access and the release of capabilities also remains concentrated within these firms.
The report also raises the language issue in Generative AI. Is this technology/tool currently capable of handling the world’s multilingualism? Doesn’t AI exacerbate the digital divide? Don’t AI platforms yield better and more accurate results when questions are formulated in English?
Most large language models (LLMs) – the cornerstone of generative AI – were trained using English data. This is not surprising when we consider the statistics provided by official institutions and platforms regarding the ranking of global languages in terms of online content. More than half of web content is available in English, far surpassing other languages such as Spanish, German, and Japanese. Even Chinese and Hindi don’t make it into the top ten.
Arabic, too, lacks a significant amount of online content, ranking around 20th, despite being the fifth most spoken language globally in terms of Internet users. This paradox reflects the crucial role and intrinsic relationship between the abundance of content in a particular language on the web and how AI platforms respond to that language.
Since language is not neutral, it carries its own cultural baggage and reflects the environment and context that produced it. Therefore, the issue extends beyond language itself to a superficial understanding of the cultural context. In other words, Generative AI answers our questions/prompts based primarily on the Western cultural framework, with all its inherent flaws and limitations.
The report raises awareness about AI governance worldwide, as a previous report of the United Nations Conference on Trade and Development shows that more than a hundred countries, predominantly in the global South, are not engaged in AI governance discussions, and only a few of them have effectively developed a national AI strategy.
It is time to work diligently on the governance of artificial intelligence in our region, not as a belated reaction to the decisions of a few AI firms, but as a proactive stance stemming from our conviction that the technology reshaping the consciousness and values of our societies cannot be left entirely to the calculations of entities that share neither our language, nor our culture, nor our priorities. As Canadian Prime Minister Mark Carney recently expressed in his speech at the Davos Forum, addressing middle-income countries that find themselves caught between two superpowers: “If you’re not at the table, you’re on the menu!”
