I read an article about a subject. I will forget some of it. I will misunderstand some of it. I will not understand some of it. (These two are different because in misunderstanding I think I understand but I am wrong. In simply not understanding the information I can not make heads or tails of that portion)

Just because you're worse at comprehension or have worse memory doesn't make you any more real. And AIs also "forget" things, they also get stuff imperfectly, because they don't store any actual "full length texts" or anything. It's just separete words (more or less) and the likelyhood of what should come next.

Another, I as a natural intelligence know what I can quote, and what I should not due to copyrights, social mores, and law. AI regurgitates everything that might match regardless of source.

Except you don't not perfectly. You can be absolutely sure that you often say something someone else has said or written, which means they technically have a copyright to it... But noone cares for the most part.

And it goes the other way too - you can quote something imperfectly.

Both actually can/do happen already with AIs, though it would be great if we could train them with proper attribution - at least for the clear cut cases.

The third issue: The AI does not understand even with copious training data. It does not know that dogs bark, it does not have a concept of a dog.

A sufficiently advanced artificial intelligence would be indistinguishible from natural intelligence. What sets them apart then?

You can look at animals, too. They also have intelligence, and yet there are many concepts that are incomprehensible to them.

The thing is though, how can you actually tell that you don't work the exact same way? Sure the AI is more primitive, has less inputs - text only, no other outside stimuli - but the basis isn't all that different.