[Editor's Note: This is part one of a two-part series this week on the future of AI.]
GENEVA (AN) — How do we determine “good” and "bad" uses for artificial intelligence as a global technology, and who decides what to do about it?
The short answer: Ethical questions are concentrated in the hands of "privileged groups," technology policy and governance expert Nanjira Sambuli tells Arete News, while governments, organizations, and regulators act as decision-makers. Consequently, most AI policy focuses on national priorities, financing, and incentives for public research and private innovation. A review of 1,920 AI policy initiatives among 69 countries, territories, and the European Union finds that just 4% tackle regulatory oversight and ethical advice.
"Indeed, ‘for good’ is subjective to the unarticulated assumptions of what is good or bad, and who determines it," says Sambuli, a Kenya-based researcher, policy analyst and strategist of information and communications technology. "The technological is political, and the politics of a globalized world are evident in how we design, deploy and deliberate on tech governance."
American computer scientist Norbert Wiener warned as far back as 1960 that machines would be able to act so quickly and irrevocably, automated tools would be needed to steer the machines by embodying humanity. "Because the action is so fast and irrevocable that we have not the data to intervene before the action is complete," he wrote in Science, "then we had better be quite sure that the purpose put into the machine is the purpose which we really desire and not merely a colorful imitation of it."
The U.N.'s oldest agency, the International Telecommunication Union, says it is working with 40 U.N. partners and Switzerland to do just that. ITU, created in 1865 to regulate telegraphs, says its 2024 AI for Good Global Summit promotes AI to advance health, climate, gender, inclusive prosperity, sustainable infrastructure, and other development priorities. Ahead of the AI summit, ITU and Switzerland are co-hosting the WSIS+20 Forum High-Level Event 2024 to mark the two decades that have passsed since the first World Summit on the Information Society gathered at Geneva in 2003 to agree on principles for global digital cooperation.
"One thing which is clear, and this is something which affects several industries: ethical AI and governance are something that are stressing people a lot," KPMG Switzerland's AI Director Mattia Ferrini told Applied Machine Learning Days at Lausanne in March. "We can see that there's a common theme around fairness, there's clearly a concern around explainability, there are concerns around the use of data – data lineage and quality, including the need to be able to disclose which data was used to train the AI system, the same way you have ingredients on food labels." He lists security, accountability and privacy among other concerns.
A 'common language' open to interpretation
ITU's fifth such AI summit in Geneva opens this week focused on boosting the U.N.'s 17 Sustainable Development Goals for 2030. The High-level Political Forum on Sustainable Development in July and the Summit of the Future in September also will focus on the embattled SDGs, a blueprint for humanity and the planet that the 193-nation U.N. General Assembly unanimously adopted in 2015. In November, the World Data Forum will examine how the SDGs are tracked, monitored, and reported while the COP29 climate summit will seek to limit warming affecting all of the SDGs. But trying to use the SDGs as an exact substitute for what's "good" isn't so simple, according to Sambuli, who studies how the adoption of information and communications technology produces gendered impacts on governance, media, entrepreneurship, and culture.
"The SDGs gave us a common language, but the point of departure has always been in the interpretation and praxis. We are squarely in a world where we can use the same terminology and refer to very different things that become politically fraught to unpack," she says. "At the risk of oversummarizing, this is why the SDGs are so off track, and why there are ever new interpretations on how to deliver on them, such as the role of digital technologies, and more recently AI, in delivering on each goal!"
In April, the U.N.'s Department of Economic and Social Affairs reported that nothing short of a “surge of financing” – an estimated US$4.2 trillion annually – plus “reform” of the international financial system can rescue the SDGs. With six years remaining to achieve the SDGs, the report says some hard-won development gains are backsliding, particularly in debt-saddled countries spending upwards of 12% of their revenue on interest payments alone. If the trends continue, the U.N. says, more than a half-billion people will still live in extreme poverty in 2030 and beyond.
The U.N. concluded last year that only 15% of the SDGs were on track, with progress on nearly half of the goals labeled as “weak and insufficient.” Experts still hope AI can brace the SDGs and lower poverty and inequality, protect the planet, and spread greater health, justice and prosperity.
Getting to more transparency and 'explainability'
In March, the Organization for Economic Cooperation and Development, an intergovernmental forum with 38 member countries, revised its definition of what constitutes an "AI system" from a technical standpoint:
"An AI system is a machine-based system that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments. Different AI systems vary in their levels of autonomy and adaptiveness after deployment."
AI algorithms and tools often produce biased results that perpetuate human biases, including historical and social inequalities, so determining who monitors and makes decisions about our assumptions is key to creating inclusive AI governance systems that serve diverse populations.
"For now," says Sambuli, "the sad reality is that the decisions on assumptions, views on addressing discrimination and bias are increasingly concentrated in a very unrepresentative sample of actors — the AI startup founders, their investors, and other privileged groups around the resulting ecosystem — with the rest of us relegated to holding banners outside the ivory tower demanding transparency, explainability and pointing out the real and possible risks and harms resulting from all this."
[Next in the series: 'Everything has changed' since last year's AI for Good Global Summit.]
