Comment by snemvalts

The approach they show in the end - it's still machine learning though? Exploring a space and finding parameters to optimize for a loss function (speed around the track), just not deep learning with neural nets.