Pathways to equitable and safe development of AGI

Event report

Globally, artificial intelligence (AI) has a ubiquitous presence in our day-to-day practices, increasing economic growth, resolving societal issues, taking on hazardous work, and lifting the living standards for thousands, if not millions of people.

As promising as this development seems, we must also recognise that many benefits of AI algorithms are concentrated in the Global North. At the same time, the Global South continues to be the consumer and testing ground for the former’s wealth. We must address the exacerbated inequalities, especially the gap between the haves and have-nots driven by rapid technological advancement. With artificial general intelligence (AGI) on the horizon, the panellists shed light on the ethical principles they wished to see embodied in AI design, as well as some practical steps we could take to achieve them.

A few challenges lie ahead as the world moves forward with AI advancement. One repeatedly highlighted by the panellists is the representativeness of stakeholder groups in AI technical development, its relevant policymaking and regulation, and its deployment. Countries are drafting respective national strategies and frameworks for protecting the rights of people, but the UNESCO-launched Artificial Intelligence Needs Assessment Survey in Africa discovered that many governments in the Global South still need a comprehensive understanding of the full impacts of AI in local contexts, and which exact policy instruments will be the most effective.

Due to this lack of an adequate environment for legislation and policymaking in the Global South, much of the regulations on AI technology have been led by the Global North. Speakers commented that the countries lagging in terms of AI development and regulation are increasingly aware of the need for them to assert their positions and ensure their voices are heard.

Critically relevant is the idea of ensuring digital sovereignty. Speakers discussed ‘intelligent devices’ supported by AI technologies. From their creation to their mass production, and from their consumption to their regulation, intelligent devices are mostly in control of Global North countries, not those of the Global South. The latter’s interpretation of ethical principles and rights to be protected is, therefore, seldom reflected in global discussions. Speakers recognised that, as the Global South takes up the majority of the world’s population, it is paramount for these countries to enhance their capacities and allocate resources to concoct their own solutions.

One way will be to reconsider governments’ usual role as customers to Global North companies. Speakers proposed that governments set aside funds and resources to foster innovation and technological advancement within the public sector, becoming the producers of public goods instead of staying mere consumers of private companies’ products. This way, they would be more capable of pursuing qualities that public goods should enjoy, such as transparency of product design and fairness in outcomes. Another will be to pry open the money-driven models that private tech companies share, and understand where ethical principles stand against their profits. Private companies’ conduct in AI design has practical implications for customers in the Global South, especially when profit-maximising decision-making processes cannot account for the exclusion of the most marginalised, as well as biases in algorithmic outcomes.

Speakers proposed participatory design as a methodology that could challenge business practices and provide insight into how best to implement ethical principles. By having the end users partake in the very beginning of the product design phase, participatory design methods empower the users to comprehend the technology behind the product, flag issues, and have their concerns mitigated throughout various iterations of the product. Such a methodology would contextualise the principle of inclusion for all stakeholders and increase transparency in design. This also might inform regulators on what the best industrial practices would be.

To close the session, panellists agreed on joint responsibility for stakeholders, for ensuring a diverse, inclusive, and democratic digital transformation process across the world. Long-term commitments and strong leadership from public institutions must be coupled with accountability and transparency for better AI governance.

By Yung-Hsuan Wu

The session in keywords