Gridcoin (the cryptocurrency you're referring to) doesn't actually use the scientific computation as proof-of-work. It's built on proof-of-stake. It just also happens to have minting set up so new coins go to people doing scientific computation, in a process that superficially resembles proof-of-work mining. The paying-people-to-do-scientific-computation part is purely something extra and isn't strictly necessary in its design to make it work as a cryptocurrency.

(It's kind of splitting hairs, but I see gridcoin brought up a lot by people arguing stuff like "Proof of stake isn't the answer. Proof of work is still good, we just need more useful proof of work" which is a big misunderstanding of it. Useful scientific computation generally doesn't work as proof-of-work and is not an example of it even in the context of Gridcoin. Only specific kinds of computation can work for proof-of-work, and the computation being useful in other contexts actually harms its usefulness for proof-of-work because it can allow the computation for 51% attacks to be subsidized by the other uses of the computation.)

Point taken. Duplication of effort could be argued is proof of work but perhaps that is splitting hairs.