I'm still puzzed by what type of memory would allow for such behavior:

The girl disappeared for several seconds after full occlusion (sec. 13-14), however appeared once more at the end of the video. So it is clearly not LSTM style memory.

Any ideas?

Happy to organize #CoRL2025 Workshop! If you are interested in world models or learning simulators for robotics, consider joining!

simulatingrobotworlds.github.io

More info 👇

Anyway, I still hope and wait for the time of real modularity in self-specified "set" of information topics that can enter the brain. E.g., imagine a super adaptable blue sky that has filters for specific topics only, but now for all the devices and apps that you are using

Who would be "Japan" to have an example?

arxiv.org/abs/2410.16770

This reminded me recent paper that tries to use program generation for compositional scene generation