Dr. Alexander Popov (MON Fellow, Oct ‘20 – Jun ‘21) will give his second talk on the progress of his research at CAS on the topic:

Narrating AI. Speculating through Science and Fiction on the Future of Machine and Human Intelligence

13 May, 2021 at 16:30 h.

Moderated by Stanimir Panayotov

—

Abstract

This presentation will provide readings of several fictional texts in which artificial intelligence (AI) is a figure of central importance: as principal narrator, or as narrative agent whose being is determining of the textual structure. These critical readings attempt to bridge the field of AI research and the humanities, so that difficult (and so far intractable) questions like those of artificial consciousness and AI value alignment become (more) graspable from alternative perspectives. The objective of ensuring benevolent AI, which has recently become prominent in AI research, entails the need for machines that can learn the values of humans, and moreover – that can learn to amalgamate diverse value systems into a coherent and sustainable worldview. Learning to recognize and think about human values is therefore recast as a practical problem which must be stated in formal terms. But human values depend on an intuitive (“soft”) understanding of being in the world that is seemingly incompatible with full formalization. The presentation will therefore locate points of potential convergence between the formal, algorithmic approaches that seek to fix, explain, and
mathematize and humanistic inquiries whose goal is to provide rich, critically informed descriptions of being.

The methodology employed is that of close reading, filtered through several phenomenological and ontological concepts with particular relevance to AI development, as is demonstrated in the fictional texts. These concepts – of body, soul, world, care – are especially useful because of their practical meaning in the context of AI engineering. I hope to show that through sustained, narratively-mediated engagement with them, which pays attention both to algorithmic and to humanistic concerns, they can be reinterpreted and reintroduced in AI, so that seemingly ineffable issues become thinkable in pragmatic terms, and the need for strong entanglement between engineering and ethics, politics, affect studies, and even metaphysics is made visible as well. The presentation will outline several interpretations of these concepts, drawing on work in phenomenology, science and technology studies, anthropology, and algorithmic criticism, and will then use them as keys to reading three novels about AI: Richard Powers’ Galatea 2.2, Kazuo Ishiguro’s Klara and the Sun, and Kim Stanley Robinson’s Aurora.