Predicting things like this is always a crapshoot. Look at the methodology of the paper CrabCake linked to; the researchers' idea was to identify a few skills they think machine learning/AI is hard to replicate, and then classify the "risk" of an occupation being replaced based off of how much of those skills you find in these occupations. The result is that low-prestige jobs like "craft artists", freelance therapists and nurses are at minor predicted risk of substitution, but jobs like models, referees and tour guides are.
It would be too presumptuous of me to say that referees won't be replaced by machines, ever. But economists on this matter have moved on to the more prevalent question of complementary computer aid, like when referees depend on instant replays. That does not mean referees will be replaced outright, but rather that they remain in order to interpret the scope and nature of their profession. For other jobs - where automation is possible but sounds strange - freelancing may become popular, where individuals service high-income customers.
Work for much of human civilisation has been an integral part of someone's identity. You meet someone new, you ask them what they do, they respond "I'm a ________ ". A good proportion of our surnames (including my own) are a reflection of our ancestor's jobs. The whole of civilisation is built on the division of labour in the form of careers. And now we have to recognise that the idea we've cherished so long - that of "being employed in order to make a living" - is impossible to keep up.
I'm Chinese and my last name doesn't reflect this at all. For more of human history, individuals have been struggling to survive and produce enough food to live past the winter. The people who are most scared by this computerization shift may be those who think they know about history but in fact only know a tiny sliver.