[yepyoukno]

This isn’t a very practical consideration.

An intelligent system would not “figure out physics” from a picture. There may be some relevance to a video, though in actuality an embodiment with sensors and actuators would be more realistic.

The difficulty with such “thought experiments” also involves dynamics of needs. Life forms need to feed, protect from the environment and predators, etc. Incentive (necessity) after all is the mother of invention!