I think the critical question here is does it have a concept of cattyness? This to me is the crux of a AGI: can it generalise concepts across domains?
Moreover, can it relate non-cat but cat-like objects to it's concept of cattyness? As in, this is like a cat because it has whiskers and pointy ears, but is not like a cat because all cats I know about are bigger than 10cm long. It also doesn't have much in the way of mouseyness: it's aspect ratio seems wrong.
I don't disagree with you, and I think that what you're saying is critical; but it feels more and more like we are shifting the goalposts. 5 years ago; recognizing a cat and describing a cat in an image would be incredible impressive. Now, the demands we are making and the expectations we keep pushing feel like they are growing as if we are running away from accepting that this might actually be the start of AGI.
Moreover, can it relate non-cat but cat-like objects to it's concept of cattyness? As in, this is like a cat because it has whiskers and pointy ears, but is not like a cat because all cats I know about are bigger than 10cm long. It also doesn't have much in the way of mouseyness: it's aspect ratio seems wrong.