On: Stable Diffusion 3

[futureshock]

At this point, the next thing that will blow me away is AGI at human expert level or a Gaussian Splat diffusion model that can build any arbitrary 3D scene from text or a single image. High bar, but the technology world is already full of dark magic.

[inference-lord → futureshock]

Will ask it for immortality, endless wealth, and still get bored.

[consumer451 → futureshock]

I would be a big fan of solid infographics or presentation slides. That would be very useful.

[attilakun → futureshock]

Is there a Guassian splat model that works without the "Structure from Motion" step to extract the point cloud? That feels a bit unsatisfying to me.