Also, apart from inventing/discovering new algorithms, there's another part in inventing how to use them to help in particular "real-world" (or "non-quantum pro

Also, apart from inventing/discovering new algorithms, there's another part in inventing how to use them to help in particular "real-world" (or "non-quantum programmer world") problems. So, maybe one of the known algorithms could actually somehow help with "training a RNN", but probably nobody invented a way to rephrase ("refactor") the "training a RNN" problem in a way which could get boosted by one of the known quantum algos (the set of which can be looked at as "the quantum API") yet. Now that I think of it, the few most well known algorithms (like the Shor's one) are probably so because "real world" use cases were found for them, which are understandable to non-quantum engineers (i.e. "factorization of primes"), as opposed to "transforming a hamiltonian foobzdringle into a hilbertian mesoism in laplacian meta-subordinates" :P