AI is just like statistics.
With some help AI can determine the important features in the data.
Usually expressed in vectors (or matrices or Tensors). Using vectors for ease.
Using these vectors , the AI can recognize one of these feature by looking
for the vectors that is closest to the data.
And in reverse it can generate data by combining random vectors.

In research, they found out that good statistics is generally better than AI.
Namely, with good statistics you can work more on the vectors that you are using.
Whereas AI can just make up random vectors that make no sense, just because it fits the dataset.