I wonder, if given a big set of time1, time2 pairs (with relatively short dt) of the positions of the planets in our solar system, as seen in the night sky on e

I wonder, if given a big set of time1, time2 pairs (with relatively short dt) of the positions of the planets in our solar system, as seen in the night sky on earth, a machine learning system would be able to come up with good way to predict where the planets are going, given their positions. Surely a machine learning system would need to be able to solve this 18th century physics problem, before we throw modern physics problems at it.