Titre: "Generative AI: a new kind of Automedia ?"
Session conçue et modérée (en anglais) par Nina Begus ( UC Berkeley, Center of Science, Technology, Medicine and Society ; automedias.org)
Intervenant.e.s: Gasper Begus (Berkeley Speech & Computation Lab) et Agnieszka Kurant (Artiste indépendante)
Propos: Every technology is molded by our cultural understanding of it, including its fictional representations. The very creation of AI systems and our interactions with them are influenced by how we conceptualize AI. How do fictional representations limit our view, and in what ways are they beneficial? The talk will focus on AI and speech. While AI is imitating human speech, it is also producing its artificial speech in a unique manner. Generative AI presents a yet unprecedented, new kind of automedia:self-productive, automatic, and autonomous but also tied to human design and prompts. No AI system is isolated from its means of production, including broader social and cultural aspects. We will demonstrate how the cultural implications of language learning have shaped AI products as we recognize them today. Nina Begus will introduce a fresh approach to designing AI systems that veers away from the traditional anthropomorphic trajectory. Avoiding external projections can only succeed with enhanced interpretability of automatic and highly generative AI. Engaging in a dialogue with this talk, we are honored to invite UC Berkeley professor Gašper Beguš, an expert on GANs and speech, and Agnieszka Kurant, a conceptual artist investigating collective and nonhuman intelligences and the exploitations present in digital capitalism.