If you compare the actual frames, you'd see that the 4k variant has insane blocking, which destroys all the information. 1080p just blurs these sections, but still looks way better because it doesn't create 8x8 blocks.
Youtube can only encode to 4k because they use an inferior entropy coding and no proper deblocking. Otherwise 4k wouldn't run on any but the fastest CPUs, even without the flash overhead.
As it is, the 4k variant is inferior to 1080p and 720p because of the massive blocking, I don't see why Youtube wastes space and bandwidth on this unless they are just experimenting.