This weekend in Generative Media
The AI feedback loop; Adobe covers AI liability; TechCrunch reviews Gen-2
The AI feedback loop: Researchers warn of ‘model collapse’ as AI trains on AI-generated content (VentureBeat)
Adobe says it’ll cover legal bills for its AI tool. (Futurism.com)
Runway’s Gen-2 shows the limitations of today’s text-to-video tech (TechCrunch)
GitHub accused of varying Copilot output to avoid copyright allegations (The Register)
Introducing Whisper Web: ML-powered speech recognition directly in your browser! 🚀 (Twitter) Demo (HuggingFace)
We derived a model from scratch with our own graphics-people intuitions. Simple derivation, simple implementation, SOTA quality. (Twitter) Project page, SIGGRAPH 2023
Realistic Saliency Guided Image Enhancement (CVPR 2023)
BOOT🥾: Data-free Distillation of Denoising Diffusion Models with Bootstrapping (project page)
GANeRF: Leveraging Discriminators to Optimize Neural Radiance Fields (arXiv)
Conditional Adapters: Parameter-efficient Transfer Learning with Fast Inference (arXiv) - an alternative to LoRA that can boost inference speeds up to 18x