Ni is more generative, more creative, is looking to build new frameworks, find new connections, push the boundaries, look at things from as many different perspectives as possible to find the one that will be most useful in a particular situation, and more.
Si is less generative, less creative, is looking to use an existing framework or frameworks, establish how old connections fit the current situation, keep the boundaries where they are, look at things from the same perspective (or set of perspectives -- I won't say Si users have only one perspective, cuz that would be wrong) because that perspective seems to have always worked in the past, and so on.
This is really good, but like others in this thread, you seem to be implying that Ni is somehow inherently better than Si (more generative, more creative, etc.). Considering that Si continually gets a bad rap for being "boring, dull, and robotic" while Ni is viewed as this desirable, mystical function that could bring to light answers to the universe's greatest questions, I think a discussion on these two functions would be better if we could open our minds to the positive and negative characteristics of both. Otherwise I see this turning into yet another Si bash vs. Ni idolatry fest.
Rather than comparing Si to Ni through the positive skillsets that Ni definitively allows for (creatively organizing external information according to novel/unique internal constructs), it would be better to examine the two through some other lens, set apart by an equal distance from both Ni and Si, that way there's no comparative bias.
While Ni may be "more creative" than is Si (according to your Ni perspective on generative and creative, at least), Si is "more" of other positive skill sets than is Ni. Si rigidly molds external data according to a more organized, practical, and structured framework, while Ni loosely molds external data according to a more conditional, experimental, and flexible framework. Both have their merits, and both come in handy depending on the context.
Pardon me if this all seems nit-picky. I'm just not a fan of using positive Ni characteristics to desribe Si, as this description is set in terms that implicitly humbles Si to Ni.
