Their method uses two models trained separately on images and captions and turns images into captions without ever having seen a captioned image. So of course it's not very good at the task. Almost all the results in the paper are bad in an absolute sense. But they're surprisingly good considering the enormous handicap of not having any paired training data. So this might be not completely useless in settings where paired data is actually hard to come by.

That's also almost certainly not a cat.