The 3D Data Gap: Daniel Cremers on Generative 3D and Physical AI
The 3D Data Gap: Why Generative 3D Still Lags Behind
Generative AI has transformed text, images, and video. Diffusion models can generate photorealistic images of almost anything. Foundation models power language and multimodal systems at unprecedented scale.
But when it comes to 3D worlds, progress is significantly slower.
In the first episode of Inside ALLSIDES, we spoke with Prof. Daniel Cremers, Director of the Munich Center for ML and Professor at TU Munich, about one of the core bottlenecks in generative 3D and physical AI
Watch the Full Conversation
In this episode of Inside ALLSIDES, Daniel Cremers shares insights on:
> The limitations of current generative 3D approaches
> Why 2D training data cannot fully produce 3D understanding
> The future of physical AI and robotics
> The importance of measured reality for AI systems
If you are working in Generative 3D, Physical AI, robotics, or simulation, the future of your systems depends on the quality of your 3D data.
And that conversation is just getting started.