Today in Generative Media

GenAI does math homework; Gemini creates people again; Spotify's AI band problem

News

• Generative AI Transformed English Homework. Math Is Next (Wired)

• Google Gemini will let you create AI-generated people again (The Verge)

• Spotify Has a Fake-Band Problem. It’s a Sign of Things to Come. (Slate)

• Integrating Runway in Art Education at RISD with Daniel Lefcourt (Runway ML)

Software

• Anthropic launches Claude Artifacts generally for all users, mobile (VentureBeat)

• Basic apps can literally be programmed from scratch in English on your phone. Wow (X)

• Google Meet ‘Take notes for me’ rolling out to Workspace (9to5google)

• YouTube Music AI conversational radio sees wider testing (9to5google)

• Introducing Cerebras Inference ‣ Llama3.1-70B at 450 tokens/s – 20x faster than GPUs ‣ 60c per M tokens – a fifth the price of hyperscalers (X)

Research

• Diffusion Models Are Real-Time Game Engines (project page)

• Generative Inbetweening: Adapting Image-to-Video Models for Keyframe Interpolation (project page)

• GenWarp: Single Image to Novel Views with Semantic-Preserving Generative Warping (project page)

• Build-A-Scene: Interactive 3D Layout Control for Diffusion-Based Image Generation (project page)

• TranSplat: Generalizable 3D Gaussian Splatting from Sparse Multi-View Images with Transformers (project page)

• Drone-assisted Road Gaussian Splatting with Cross-view Uncertainty (project page)

• GenCA: A Text-conditioned Generative Model for Realistic and Drivable Codec Avatars (arXiv)

Misc

• Parent tip: you can upload a photo of your kids printed paper school calendar to ChatGPT and ask it to generate an .ics calendar file that you can directly import (X)