Eh, people have always predicted this kind of technological revolution is mere decades away. It's probably no more true now than it was a hundred years ago. But this sort of thing does reach tipping points along the way for sure.
The timelines have been inaccurate... but the point is ultimately still valid.
Once automation (i.e. AI) tech is generalizable enough and cheap enough, it obviates the need for human labour.
The general conception of how this will go down is simply; 'lower skill' (i.e. easier to automate) jobs will go away first, until even high skill jobs have been replaced, rendering the labour of people unnecessary.
That said, 30-50 years from our current stand point is a fairly educated guess when accounting for technology acceleration. I know that's something that tech optimists say a lot... but we are are closing in on multiple vectors of a versatile, generalizable form of AI - either from a brute force emulation perspective, or several million cycles of complex evolutionary algorithms evolving AI, or via neural network modeling, or any other possible avenue from which to reach intelligence.
Obviously, the more optimistic the estimates are, the larger the margin of error - saying that cheap effective AI will be available within 100 years is almost an absolute definitive.
100 years of actual accelerated progress should encompass some 10,000+ current years of tech progress. It's difficult to imagine that we can't come up with hard AI in that time span when we're already coming up with headlines about Google search AI 'dreaming of cats' from the sampling of youtube videos.