Ask HN: Are you using machine learning to do UI testing?

Web and mobile UIs are constantly changing slightly, and testing frameworks tend to be brittle; has anyone had luck applying machine learning to make them more robust — either the inputs or the outputs?

For example, "Here's what I want; see if you can still make it happen without too much of a change in your routine."

Or "Here's what I want; tell me if it changes fundamentally, but not if it changes a little."