AGI is dramatically more difficult than the deep learning algorithms we have now, no matter how smart of human-like they might seem already today. It’d put money on AGI being at least 2 decades away. Let’s pray to our future basilisk overlords we’ll have solved the containment problem by then.