What you are actually encountering here is implicit typing in mathematics, which is prevalent in almost every area of mathematics but is rarely taught in mathematical pedagogy. $\mathbb{N}$ is the type of natural numbers, and if you have members of $\mathbb{N}$ then all you can do on them are operations that require natural numbers as inputs. For example we define various exponentiation operations:
$S^n$ is the set of $n$-tuples from $S$, for any set $S$ and $n∈\mathbb{N}$.
$x^0 = 1_M$ and $x^{n+1} = x^n·x$ for any monoid $(M,·,1_M)$ and $x∈M$ and $n∈\mathbb{N}$.
$x^y = \exp(y·\ln(x))$ for any $x,y∈\mathbb{R}$ such that $x>0$.
$S^T$ is the cardinality of the set of functions from $T$ to $S$, for any sets $S,T$.
$k^m$ is the cardinality of the set of functions from $m$ to $k$, for any cardinals $k,m$.
$k^\varnothing=\{\varnothing\}$ and $k^m = \sup\{ k^p·q : p∈m ∧ q∈k \} $, for any ordinals $k,m$.
Implementation in set theory is very much irrelevant to the intrinsic mathematics. For example, $\mathbb{N}$ is implemented by $ω$ in modern set theory based on ZFC, and hence $0$ is implemented by $\varnothing$, but we never think of $0^2$ as $\varnothing^2$. Why? Because these operations are overloaded but disambiguated by the implicit input types. $0$ has implicit type $\mathbb{N}$, while $\varnothing$ has implicit type "set".
Types are implicitly introduced by mathematical definitions. For example, $\aleph_k$ is defined as the $k$-th cardinal, and this implicitly defines $\aleph_k$ to have implicit type "cardinal". In contrast, $ω$ is defined as the first infinite ordinal, so its implicit type is "ordinal". That is why $ω^ω$ is a countable ordinal while ${\aleph_0}^{\aleph_0}$ is an uncountable cardinal, despite $\aleph_0$ often being implemented as $ω$ in modern set theory.
Your three examples would all be understood by mathematicians who know the individual terms, due to implicit type coercion. That is, when none of the (overloaded) defined operations have matching type signature, we would pick the closest one that is compatible with the inputs according to their actual implementations. Since $\aleph_0$ is a cardinal and cardinals are usually implemented as ordinals, and ordinals are sets at the bottom of it all, we have an available implicit type coercion of $\aleph_0$ from "cardinal" to "ordinal" to "set".
Specifically, $\mathbb{R}^{\aleph_0}$ would be automatically interpreted as exponentiation of sets, which yields the set of functions from $\aleph_0$ to $\mathbb{R}$. $\mathbb{R}^\mathbb{N}$ needs no type coercion. $\mathbb{R}^ω$ yields the set of functions from $ω$ to $\mathbb{R}$, which is of course no different from $\mathbb{R}^{\aleph_0}$ at the implementation level, but at the communication level conveys that the input is an ordinal index, which corresponds to the notion that an ordinary infinite sequence of reals is an $ω$-sequence. Recursive definition of sequences only works for indices from a well-ordering and ordinals are canonical well-orderings (see this post), so it should not be surprising to see ordinals used as exponents to indicate the length of sequences from a set. Another example along this line is $S^{<ω}$, which denotes the set of all finite sequences from $S$, where "$<ω$" is suggestive of the meaning since ordinals less than $ω$ are finite.
elementary-set-theory
andset-theory
before asking the question. Also, I have read this question regarding the difference between these two tags. While I do consider my question should be "elementary" to set theorists, – Apr 19 '20 at 01:33elementary-set-theory
is appropriate here: Topics include intersections and unions, differences and complements, De Morgan's laws, Venn diagrams, relations and functions, countability and uncountability, etc. More advanced topics should use (set-theory) instead. But I will not argue with the editor for the moment. – Apr 19 '20 at 01:33