One team said verification and validation techniques would prevent AI systems from failing catastrophically (must've done homework on their judge...) and then there was some back-and-forth that included one of the teams reading a snippet from some arxiv paper that contained a reference to an interesting piece of work that I didn't know about.

It was fairly amusing because I had to explain that they were perhaps missing the forest for the trees with that whole exchange, but also that I had learned something new.