Bruno Caldas Vianna application/pdf Text conference object book part Text Konferenzveröffentlichung Buchkapitel Media and Communication Sciences ÖFOS 2012 -- SOCIAL SCIENCES (5) -- Media and Communication Sciences (508) -- Media and Communication Sciences (5080) Information technology ÖFOS 2012 -- TECHNICAL SCIENCES (2) -- Electrical Engineering, Electronics, Information Engineering (202) -- Electrical Engineering, Electronics, Information Engineering (2020) -- Information technology (202022) Audiovisual media ÖFOS 2012 -- TECHNICAL SCIENCES (2) -- Electrical Engineering, Electronics, Information Engineering (202) -- Electrical Engineering, Electronics, Information Engineering (2020) -- Audiovisual media (202002) Arts ÖFOS 2012 -- HUMANITIES (6) -- Arts (604) -- Arts (6040) Generative AI Text-to-Image Models Philosophy of Language Computational Semantics Neural Networks eng St. Pölten University of Applied Sciences Medien- und Kommunikationswissenschaften ÖFOS 2012 -- SOZIALWISSENSCHAFTEN (5) -- Medien- und Kommunikationswissenschaften (508) -- Medien- und Kommunikationswissenschaften (5080) Informationstechnik ÖFOS 2012 -- TECHNISCHE WISSENSCHAFTEN (2) -- Elektrotechnik, Elektronik, Informationstechnik (202) -- Elektrotechnik, Elektronik, Informationstechnik (2020) -- Informationstechnik (202022) Audiovisuelle Medien ÖFOS 2012 -- TECHNISCHE WISSENSCHAFTEN (2) -- Elektrotechnik, Elektronik, Informationstechnik (202) -- Elektrotechnik, Elektronik, Informationstechnik (2020) -- Audiovisuelle Medien (202002) Kunstwissenschaften ÖFOS 2012 -- GEISTESWISSENSCHAFTEN (6) -- Kunstwissenschaften (604) -- Kunstwissenschaften (6040) http://creativecommons.org/licenses/by/4.0/ Testo Risultato di convegno Capitolo di libro Meaning and Language in Artificial Intelligence’s own Linguistic Turn open access 2024-11-27 This paper examines the rapid development of generative visual artificial intelligence, particularly text-to image models, through the lens of 20th-century philosophy of language. It contrasts the one-to-one symbolic representation reminiscent of early Wittgenstein and traditional AI with the subsymbolic, conceptual synthesis found in modern neural networks like CLIP and diffusion models. The paper traces the technological evolution from GANs to diffusion, analyzing the implications for creativity, originality, and the machine's capacity for metaphor. It argues that while current models can visually render complex concepts by learning from vast datasets, they struggle with true abductive reasoning and novel metaphor creation, thus highlighting the persistent limits of machine understanding when compared to human language. doi:10.60522/o:7213 https://phaidra.ustp.at/o:7213