It does not seem that quantum memory can be controlled or be deterministic. If it is not deterministic I wonder how it can be used for many of the current applications.

Probably for something like machine learning, image processing, or computer vision that things do not have to be exact it might work wonders.

The skull, etc. above, doesn’t “collapse” into one state — we can see both at once. This is known as sub-critical, or inefficient measurement on quantum systems. We’re just beginning to understand these thinsg.

Yes, parallel computing algorithms benefit from computing elements that can contain multiple states, whatever “states” means. A circularly-polarized radio wave can be picked up by horizontal (US) or vertical (Brit) antennas, because the signal is purposely given a complex state. However, not all the energy appears in each state, which is why Schrodinger’s Cat example is a poor fabrication.

As we study quantum-sized entities, we’ll gain more understanding of how to use them for computing, much as how we learned how to use vector processors decades ago to speed up certain algorithms, and now have entire chipsets devoted to graphics processing because these problems are amenable to parallelization. Only certain problems are. Using encoded DNA snippets also give us parallel computing abilities, if somewhat messy.
;]
There’s nothing magic about it and nothing approaching “a solution to all computing”, despite the “quantum” hype.