Is it me or this makes no sense? Isn’t the goal of taking a picture of a black hole to extrapolate data we do not have yet?

How could any AI “know” anything that would allow reconstructing a low quality image?