Come on, we now have systems that can believably answer arbitrary questions in human language. This is literally what I dreamed of when I got into computing like 25 years ago, and would be considered science fiction only 5 years ago. As a side effect, entire tasks as important as machine translation and summarization have pretty much been solved for major languages.
Regardless of whether you buy the full hype or you think they're just stochastic parrots, I think it more than qualifies to make the second list (and probably the first, but I get that there's no perspective to be so sure about that).
The paper itself (as a paper, i.e. an explanation of the underlying results) is quite bad, by the way. It's better to learn about Transformers from one of the many good blog posts. But that doesn't detract from its influence.
Regardless of whether you buy the full hype or you think they're just stochastic parrots, I think it more than qualifies to make the second list (and probably the first, but I get that there's no perspective to be so sure about that).
The paper itself (as a paper, i.e. an explanation of the underlying results) is quite bad, by the way. It's better to learn about Transformers from one of the many good blog posts. But that doesn't detract from its influence.