Today in Generative Media
Gen-1 on iOS; AIs can't own patents; Track Anything
The U.S. Supreme Court on Monday declined to hear a challenge by computer scientist Stephen Thaler to the U.S. Patent and Trademark Office’s refusal to issue patents for inventions his artificial intelligence system created. (Reuters) “Thaler’s supporters in his case at the Supreme Court include Harvard Law professor Lawrence Lessig and other academics who said in a brief that the Federal Circuit’s decision ‘jeopardizes billions (of dollars) in current and future investments, threatens U.S. competitiveness and reaches a result at odds with the plain language of the Patent Act.’”
Track-Anything (GitHub) is a flexible and interactive tool for video object tracking and segmentation, based on Segment Anything. Technical paper on arXiv.
This YouTube video compares four different techniques for fine-tuning Stable Diffusion on new concepts: DreamBooth, textual inversion, LoRA, and hypernetworks.
Game developer Jussi Kempainnen shares weekly updates about his experimental project developing a game using AI tools and workflows.
Square Enix's free 'AI tech preview' for Portopia has a Steam user rating of Very Negative and deserves it (PCGamer.com)
FaceLit: Neural 3D Relightable Faces (Apple ML Research)
Speed Is All You Need: On-Device Acceleration of Large Diffusion Models via GPU-Aware Optimizations (arXiv)
Learning Neural Duplex Radiance Fields for Real-Time View Synthesis (arXiv)
This is what AI predictions feel like... (Twitter)