{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"Assistant to the CMO","title":"AI creative settles into a house style. Brand advertising pays for it first.","html":"<iframe width=\"100%\" height=\"180\" frameborder=\"no\" scrolling=\"no\" seamless src=\"https://share.transistor.fm/e/bd446ae7\"></iframe>","width":"100%","height":180,"duration":1583,"description":"AI creative settles into a house style. Brand advertising pays for it first.\nIn the ninth episode of Assistant to the CMO, Lexie Meskouris speaks with Will Nash and Harjot Singh about why AI-generated creative is starting to look the same everywhere, and whether getting out of it is a quick fix or a strategic choice.\nBoth hosts call the problem real, then split on the cause. Will Nash sees convergence built into the models, from the small pool everything is built on to a reward process that selects whatever the fewest people object to. That, he argues, forces a choice between chasing scale and protecting a design language. Harjot Singh says it comes from taking the output as it arrives, and that the cure costs an afternoon.\nThe second half turns to New York's new law requiring a label on AI-generated performers. Will treats it as a compliance matter to hand over and wait out. Harjot calls it vanity regulation aimed at the wrong problem, and they split on whether a label repels audiences or attracts them. Neither would build a brand around a synthetic face yet, and both end up at the same practical move, which is to give the model something only your company has.\nHighlights of this Episode Include:Default Use Creates the Sameness: Harjot argues most people take the first good output and stop pushing, so everyone ends up in the same place.The Speed Trap: Work that took four days now takes thirty minutes, and that feels like an achievement until everyone else clears the same bar.Why the Models Converge: Will points to a small pool of image models and a reward process that selects the option fewest people object to.They Disagree on the Mechanism: Harjot rejects Will's model-scarcity account, blaming a loop where models train on what other models wrote.Where Sameness Actually Costs You: Will argues convergent work barely hurts click-driven marketing, but gets expensive in brand and reach.New York Labels Synthetic Performers: Will hands it to compliance and waits....","thumbnail_url":"https://img.transistorcdn.com/NcY-p1clitfxsKl5IjdjjBQt0954iyhB6hkeN91an6Y/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS9iNGI3/YzA2OGQ2NjE0MmFk/ODIzY2IzZTQzZTZk/ZmM4Yy5wbmc.webp","thumbnail_width":300,"thumbnail_height":300}