Keep on to blur preview images; turn off to show them clearly
You won't believe it, I was just thinking of «can't Chyna build cheap inflatable space stations once they get better launch capacity?» and lo and behold, they've indeed been looking into that! Very 2D culture, very predictable. Just going through sane ideas.
We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1
RT @a16z: David Sacks says the biggest risk of AI was described not by James Cameron in The Terminator but by George Orwell in 1984. “I al…
AI Optimist. Empiricist, not 'rationalist'. Anti world government.
> they don't know we uploaded Yud 2.5 years ago
We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1
开荒
I built a Cursor-like experience plugin for Neovim: avante.nvim
Paper: https://t.co/NEBweZNnwG
Your guide to radiance fields | Host of the podcast @ViewDependent | Founder and CEO of https://t.co/5MjtfpwEU3 | FTP: 279 | discord: https://t.co/lrl64WGvlD
4D Neural Voxel Splatting: Dynamic Scene Rendering with Voxelized Guassian Splatting Contributions: • Unified 4D Voxel Architecture: We extend 3D voxel grids to 4D by treating time as an additional dimension in the voxel feature space. This enables temporal-aware Gaussian generation that adapts both spatially and temporally. Unlike Scaffold-GS, which generates static Gaussians, our voxels produce time-varying Gaussians through learned temporal features. • Selective Deformation Strategy: Through extensive experimentation, we identified that deforming all Gaussian properties leads to training instability. We introduce a selective approach that only deforms geometric properties (position, scale, rotation) while keeping appearance properties (color, opacity) fixed, significantly improving convergence and quality. • View-Adaptive Refinement: We propose a novel refinement mechanism that identifies and selectively improves underperforming viewpoints through adaptive densification, addressing temporal inconsistencies without global overhead. • Memory-Efficient Design: Our framework achieves O(fV + F) memory complexity instead of O(N · T), making dynamic scene rendering feasible on consumer GPUs.
Paper: https://t.co/NEBweZNnwG