Quantum computers a real candidate once they get off the ground. They might help solve a few problems in chemistry and condensed matter physics, but on the other hand they definitely will make a lot of encryption we heavily rely on obsolete, and the replacements are noticeably inferior. And that's about it, because quantum algorithms are hard to design. So, that seems like a net negative to me.
Deep neural nets are powerful, but the fact we fundamentally don't understand how they work is a bit nerve-wracking.