Ambient space is the space of high dimensional input data to an ML algorithm. In the case of input images this is also known as pixel space. I haven’t seen this term widely used in ML literature. Wikipedia has https://en.wikipedia.org/wiki/Ambient_space_(mathematics) as the space around a mathematical object under consideration - e.g. a 1-dimensional line may be in $\mathbb{R}^2$ or $\mathbb{H}^2$.

See - https://towardsdatascience.com/understanding-high-dimensional-spaces-in-machine-learning-4c5c38930b6a