Today in Generative Media

GenAI propaganda; Bing AI image generator fails to block prompts; generative D&D

• AI Chatbots Are Learning to Spout Authoritarian Propaganda (Wired)

• Bing’s AI image generator tries to block ‘twin towers’ prompts, but it’s not working (The Verge)

• I can use a procedural dungeon map generator, give the resulting map to GPT-4V, and ask it to create an adventure around it. It provided lore, tables, descriptions etc. Then it does well simulating the entire adventure as DM. (X)

• Kandinsky: an Improved Text-to-Image Synthesis with Image Prior and Latent Diffusion (arXiv)

• Aligning Text-to-Image Diffusion Models with Reward Backpropagation (project page)

• God-tier prompt injection: using the knowledge cutoff date against ChatGPT (X)