VLA = vision-language-action, a kind of a machine learning model