'Another bucket of AI gunk is dropped on to our faces' Castle Walls is the first fully AI-generated series to premiere on a major streaming platform. 'Another bucket of AI gunk is dropped on to our faces' Castle Walls is the first fully AI-generated series to premiere on a major streaming platform. Castle Walls review - Prime's first fully AI-generated show is here, and it's a Vaseline-smeared mess Billed as the first end-to-end AI-created series on a major streaming service, this Turkish show looks like the result of the prompt'make it cinematic' J ust as the world reels from Elon Musk's AI-powered The Odyssey remake - though "remake" feels rather generous for a seven-minute splat of synthetic sludge - another bucket of AI gunk is dropped on to our faces. For Prime Video's AI-generated TV series Castle Walls, produced by Turkish production studio Ay Yapım using a melange of generative video tools, has arrived, billed as the first end-to-end AI-created series to premiere on a major streaming platform. Set in the 11th century, the show revolves around the historic Alamut Castle, although it's entirely absent of any real sense of time or place.

While state-of-the-art diffusion models (DMs) excel in image generation, concerns regarding their security persist. Earlier research highlighted DMs' vulnerability to data poisoning attacks, but these studies placed stricter requirements than conventional methods like'BadNets' in image classification. This is because the art necessitates modifications to the diffusion training and sampling procedures. Unlike the prior work, we investigate whether BadNets-like data poisoning methods can directly degrade the generation by DMs. In other words, if only the training dataset is contaminated (without manipulating the diffusion process), how will this affect the performance of learned DMs?

While state-of-the-art diffusion models (DMs) excel in image generation, concerns regarding their security persist. Earlier research highlighted DMs' vulnerability to backdoor attacks, but these studies placed stricter requirements than conventional methods like 'BadNets' in image classification. This is because the former necessitates modifications to the diffusion sampling and training procedures. Unlike the prior work, we investigate whether generating backdoor attacks in DMs can be as simple as BadNets, i.e., by only contaminating the training dataset without tampering the original diffusion process. In this more realistic backdoor setting, we uncover bilateral backdoor effects that not only serve an adversarial purpose (compromising the functionality of DMs) but also offer a defensive advantage (which can be leveraged for backdoor defense). Specifically, we find that a BadNets-like backdoor attack remains effective in DMs for producing incorrect images (misaligned with the intended text conditions), and thereby yielding incorrect predictions when DMs are used as classifiers. Meanwhile, backdoored DMs exhibit an increased ratio of backdoor triggers, a phenomenon we refer to as `trigger amplification', among the generated images. We show that this latter insight can be used to enhance the detection of backdoor-poisoned training data. Even under a low backdoor poisoning ratio, studying the backdoor effects of DMs is also valuable for designing anti-backdoor image classifiers. Last but not least, we establish a meaningful linkage between backdoor attacks and the phenomenon of data replications by exploring DMs' inherent data memorization tendencies. The codes of our work are available at https://github.com/OPTML-Group/BiBadDiff.