La génération d’images au sein du Google Research Lab. Jason Baldridge en entretien avec Nicolas Malevé et Katrina Sluis

In this interview, conducted on 14, 2023, Nicolas Malevé and Katrina Sluis enter into a wide-ranging discussion with Jason Baldridge, a research scientist at Google, on the historical, technical, and cultural implications of text-to-image generators. They discuss Baldridge’s background in computational linguistics and his involvement in the development of Google’s Pathways Autoregressive Text-to-Image model (Parti), which was publicly announced in June 2022. In teasing out the different paradigms of image generation models, the conversation turns to the dynamics of language, translation and semantics, and the sociocultural implications of developing generative models based on textual inputs. The resulting discussion underscores the way researchers apprehend the relations between technology and culture, text and image, model and training data.