The rapid advancement of artificial intelligence (AI) technology has spurred an urgent conversation about the need for robust regulations. Professor Paul Dumouchel from the University of Quebec at Montreal argues that the common misunderstanding of AI’s nature has significant consequences for how it should be regulated. This analysis is detailed in his essay “AI and Regulations,” published in the journal AI.
Professor Dumouchel emphasizes that AI should not be viewed as a singular, monolithic entity. Instead, it comprises diverse cognitive technologies with varying characteristics and applications. This misconception often leads to inefficient regulatory approaches. He states, “Viewing AI as something that exists in itself, rather than as a set of cognitive technologies…leads to inefficient approaches to regulation.” This perspective underscores the importance of understanding AI’s multifaceted nature to develop effective oversight mechanisms.
The urgency for AI regulation has increased with the emergence of advanced AI models like ChatGPT. Questions arise: Should AI be regulated? If so, how and by whom? Professor Dumouchel explores these questions, noting that the diversity of AI technologies complicates the regulatory landscape. He explains that AI encompasses a wide array of devices, algorithms, and applications, each requiring tailored regulatory approaches.
Professor Dumouchel traces the historical roots of AI regulation, noting that early AI research was predicated on the assumption that human thought could be precisely replicated by machines. This led to the belief that human and artificial intelligence were fundamentally similar, which has influenced regulatory thinking. However, Professor Dumouchel argues that this view is flawed. He writes, “The claim that there is no difference between human thought and the performance of an artificial cognitive system thus rested on the possibility of creating a machine able to duplicate the results of diverse intellectual operations.”
He critiques two prevalent approaches to AI regulation. The first is the proposal for a moratorium on AI research, which he describes as a contradictory and ultimately ineffective measure. Professor Dumouchel asserts, “A moratorium, however, is not a form of regulation. It is more like a double bind, a contradictory injunction that is doomed to fail.” The second approach involves embedding ethical considerations into AI systems, which he argues overlooks the fundamental differences between human and artificial agents.
Instead, Professor Dumouchel advocates for a regulatory framework that focuses on the unique characteristics of AI systems. He highlights the importance of understanding the specific cognitive domains and limitations of AI agents. Unlike humans, AI systems are confined to the data and algorithms that define their functions. This distinction is crucial for developing regulations that address the actual capabilities and risks of AI.
He also emphasizes the importance of transparency and accountability in AI systems. Since AI agents operate invisibly and ubiquitously, it is essential to ensure that their actions are traceable and understandable. Professor Dumouchel notes, “We never encounter the agent itself for it ultimately is a mathematical function. We are only exposed to some consequences of its functioning.” Therefore, regulations should mandate clear reporting and accountability mechanisms for AI systems.
Professor Dumouchel concludes by stressing that the primary challenges of AI regulation are political rather than ethical or metaphysical. The deployment of AI technologies alters power dynamics between different social actors, necessitating a regulatory approach that considers these shifts. He writes, “Artificial cognitive systems and associated information and communication technologies have been central in the unprecedented concentration of wealth and power that we are witnessing.”
In summary, Professor Dumouchel’s study underscores the need for a nuanced and informed approach to AI regulation. By recognizing AI as a diverse set of cognitive technologies, policymakers can develop more effective oversight mechanisms that address the specific risks and capabilities of AI systems. Transparency, accountability, and an understanding of AI’s limitations are essential components of a robust regulatory framework.
Journal Reference
Dumouchel, Paul. “AI and Regulations.” AI, vol. 4, no. 4, 2023, pp. 1023-1035. DOI: https://doi.org/10.3390/ai4040052
About the Author
Paul Dumouchel is Canadian and until recently professor at the Graduate School of Core Ethics and Frontier Sciences, Ritsumeikan University, Kyoto, Japan, where he thought political philosophy and philosophy of science. He is the author of Emotions (Seuil, 1999) The Ambivalence of Scarcity and Other Essays (2014) and The Barren Sacrifice (2015) both at Michigan State University Press. With Reiko Gotoh he edited Against Injustice: The New Economics of Amartya Sen (Cambridge University Press, 2009) and Social Bonds as Freedom (Berghahn Books, 2015). His most recent book, with Luisa Damiano, is Vivre avec les robots (Seuil, 2016) The English translation Living with Robots (Harvard University Press) came out in 2017 and the Italian and Korean translation in 2019. He is presently Associate Professor at the department of philosophy of the Université du Québec à Montréal, Canada.
Paul Dumouchel est Canadien, de 2003 à 2021 il était professeur de philosophie politique et de philosophie des sciences à la Graduate School of Core Ethics and Frontier Sciences de l’Université Ritsumeikan à Kyoto au Japon. Il est avec Jean-Pierre Dupuy co-auteur de L’Enfer des choses (Seuil, 1979). Il est aussi auteur de Émotions (Seuil, 1999) ainsi que de Le Sacrifice inutile (Flammarion, 2011) de même que The Ambivalence of Scarcity and Other Essays (2014). Avec Reiko Gotoh il est co-directeur de collection de Against Injustice: The New Economics of Amartya Sen (Cambridge University Press, 2009) ainsi que de Social Bonds as Freedom (Berghahn Books, 2015). Son livre le plus récent avec Luisa Damiano, est Vivre avec les robots (Seuil, 2016) a été traduit en anglais, en italien et en coréen. Il est présentement professeur associé au département de philosophie de l’Université du Québec à Montréal, Canada.
