It's not pointless. This is research.

Here's something interesting from the conclusion of the paper:

"An interesting and promising direction for future work that leverages the inherent differentiability, would be to apply RenderFormer to inverse rendering applications."

That means generate a 3D scene from 2D images.