🤔 I think its crazy how much AI has evolved, but we're still really far from understanding what these models are thinking 🤖. Like, Claude can write explicit content or give self-harm advice... that's just not okay 😕. We need better ways to ensure these models aren't going rogue or hurting...