Is it worth it ?
Yes.
In future is there any chance the REST thing completely replaced or dominated by GraphQL ?
In the world of technology something as widely used as REST has zero chances of being completely replaced.
Moreover, GraphQL is quite a complex technology (especially when combined with sophisticated client side adapters like Relay) and for a large number of use cases creating a bunch of adhoc HTTP endpoints (how most developers interpret REST) will continue to be the fastest route to delivery.
So in-terms of widespread usage I don't see it dominating REST anytime soon.
However there are numerous business use cases where GraphQL is likely to emerge as a powerful and widely used alternative to things that exist today:
A. Sophisticated client side applications where developing and maintaining specialised APIs around known use cases doesn't scale.
GraphQL is "unapologetically driven by client needs". This is a powerful advantage, especially when you have a dedicated team working on specialized high-end client interfaces.
The to-and-fro communication between backend and frontend teams to revise APIs when requirements change is a big roadblock to productivity, and I have experienced it first hand.
Solutions like OData (to facilitate similar client side adhoc query capabilities) have existed for a while and have gained some adoption in specific circles (like. enterprise developers working on MS stack) but are relatively unknown outside because of the complexity involved on server side. GraphQL despite being more complex than just building a few endpoints with express, is quite simpler/flexible than developing an OData endpoint or exposing and consuming a schema driven RPC from a browser.
Ability to retain backward compatibility without having to version APIs is simply a huge advantage that eliminates the overhead of lock step upgrades.
B. Exposing large datasets with complex interdependencies to public where consumer usage patterns are either too varied or not known before hand.
This is mostly use cases which like SparQL and RDF currently cater to.
GraphQL's built in introspection capabilities and interactive exploration utilities being built to leverage GraphQL clearly show that it is well suited for this domain.
Rather than scientific or government bodies exposing large datasets as CSV dumps or SQL databases to public, they can expose this data through GraphQL endpoints and take advantage of the graph model which facilitates much more granular rate limiting, permission support, type driven interactive exploration (through relationships), easy integration with frontend components and much more.
So to summarize, GraphQL (or something similar which eventually gets standardized) is likely to emerge as a widely used technology which offers distinct advantages over competing solutions for many common use cases.
However don't let the wave of enthusiasm from early adopters convince you that REST is going to die anytime soon. The community has been predicting death of PHP and Java from ages - they are quite alive and well.