Polemics Coverage of the Dialogue of Continents Conference, organized by the Reinventing Bretton Woods Committee, December 4-5, 2023.

As part of the Dialogue of Continents conference, hosted at the Diplomatic Academy in December 2023, several experts in the field of artificial intelligence were invited to share their insight into an increasingly controversial field. Impressively, all three of the invitees were women, providing clear visibility in what is an otherwise male-dominated field. These women were: Doctor Loubna Boufara, a specialist in AI system development; Marta Sabou, who works in applying AI to educational purposes; and Vivienne Ming, who has extensive work with AI deep neural nets.

The panel kicked off with Sabou providing a brief overview of AI development’s three waves. Initially, AI was only applied to closed environments with strict rules where it could only provide definitions and clearly codified knowledge. This was mainly seen in the medical community where systems would provide definitions to experts, for example. Then in the early 2000s, AI entered its second-wave, characterized by machine learning. Here, AI was asked to learn quickly from patterns without providing insight or relevant knowledge. However since the third-wave of AI has been focused on bridging the gap between the two. Regardless of its capabilities though, the capacity of AI to outwit human intelligence has always been called into question.

In this regard, Dr. Boufara posited that current models of artificial intelligence do not need to pass the Turing test any longer. In other words, AI has already surpassed a level where the average person can detect if they are engaging with another human or AI. However, Vivienne Ming retorted to this assertion and instead maintains that no model of AI has managed to beat the Turing test. Instead, Ming posits that while AI has managed to surpass human capacity for coordination and statistical reasoning, its capacity to actually produce desired answers depends “very much” on the interrogator. In other words, according to Ming, AI knows “everything and understands nothing,” a distinction which could have profound effects on AI applications.

Ming continued her line of reasoning by delving into applications of AI in economic terms. In her view, AI is fully capable of completing routine labor, or labor which follows a known set of operations. Where AI struggles is creative labor, or the ability to explore the unknown and draw from influence. In Ming’s view, creative labor is where human capacity can trump AI and should be the focal point for inclusion in the economy moving forward. Ming ended her perspective by warning against the dangers of lowering wages of workers through the use of AI.

At this point, the panel moved into a discussion of the safety of AI in the modern world. Dr. Boufara kicked off the discussion with an analogy of AI to human children. According to her, we raise children through instilling certain values within them in order for them to become law-abiding, functional members of society, and we should treat AI no differently. Dr. Boufara posited that we should instill a universal set of values within AI models in order for them to learn and then apply them. While Dr. Boufara maintained this positive outlook on relations with AI models, but Vivienne Ming remained skeptical.

In Ming’s view, there are no predictions we can make today and instead, we as a society are confronted with a series of choices to make. These choices and the people who make them will shape the impact of AI on human life, according to Ming. Therefore, the problem is not AI itself but rather conditioning users and developers to implement moral and ethical considerations.

At this point, Sabou interjected with new concerns in the use of AI, notably regarding sustainability. For Sabou, efficiency is not the only factor worth considering in AI applications, as sustainability and its benefits to all involved are also major factors. She highlighted the impact digitization had on EU funding, where high-level EU politics became increasingly considerate of the long-term impacts technology could have based on the decisions made. To this end, Sabou advocated for a digital humanism where technology was applied through human values, impacts, and inclusions.

Wrapping up the panel, Ming made a final point on AI and democratization. In her view, there is currently a high risk for the monopolization of AI resources. The best example of this is Amazon’s control of servers globally. However, Ming posited, not every use of AI requires such large models which need massive server space or support. Instead, Ming advocated for further democratization and libertarianism in the world of AI, highlighting the prevalence of similar ways of thinking in the digital world already.

Furthermore, Ming also advocated for individual responsibility in the control of data with measures such as data trusts as a clear example. A data trust is an arrangement where a digital entity agrees to store its data collected from its services within an independent organization, thereby reducing any risk of foul play. Ultimately, Ming highlighted the necessity for moral action when dealing with data and AI usage in the modern world.

In sum, the panel on AI provided new perspectives and considerations on the use of new technologies. With AI becoming increasingly common in the modern world, considerations such as morality, ethics, user responsibility and data handling will only also become more important.

Written by Reed McIntire, Edited by Anna Riggs

Photo Credit: Anne Roos Ververs