Gowers (op cit) has a brief section on "infinity" in which he shows there is no number system satisfying the ordinary axioms that also has a number such that . The argument is,
Thus it requires only associativity of , and a distinct number with and .
Here, of course, is considered as a number in the system.
Gowers also has a whole chapter entitled "Limits and Infinity". He says, "the concept of infinity is indispensible to mathematics, and yet is a very hard idea to make rigorous." The chapter discusses the possibility of a rigorous picture of infinity via the concept of limit.
So here, "infinity" is considered as a process rather than a number.
The distinction (or lack of distinction when one should be made) between infinity as a number and infinity as a process causes a great deal of confusion, aggravated by sloppy language such as "and so on, going to infinity", as if infinity is somewhere one can go, or the (sometimes justifiable) use of the symbol in calculations, for example,
What actually is a process in mathematics? To be a process requires some concept of time, and yet mathematics when done rigorously has no concept of time, unless time is a special variable letter such as , and then the mathematics is done for a fixed (but general) value of this .
Mathematicians do have an informal conventional language that involves time in some informal sense: we might say, "let , so varies with time ," but and hence are fixed in what follows. Or "consider the sequence as increases" suggesting the is the measure of "time".
A sequence is of course just a notation for a function . Thus processes are typically represented as functions such as or in mathematics and we study processes via the study of functions at specific and fixed but general values for their arguments or .
When we study functions in this way we typically want to abstract some property of these functions, such as their behaviour as the argument "approaches infinity". It seems to be natural to introduce a symbol for this "object" that the argument "approaches", and then natural to look at the limiting value notated or .
Unfortunately this does not make literal sense ( is not a possible argument for the function) nor could it make mathematical sense. For example, if is equal to or (or something else?).
The official solution is to say that the intuition is wrong and no number could behave like this. In its stead we must carefully analyse the concept of limit of a sequence. The definition is that a sequence has a limit if
What a pity we can't just consider a number and say that has a limit if is well defined.
That is the mainstream view. But there is a non-standard nonstandard view that you can do exactly this. This is Abraham Robinson's "nonstandard analysis" and is "nonstandard" in the technical sense that it is about such "nonstandard" numbers but it also non-standard in the literal sense that most mathematicians regard it as unnecessary, not standard, and a highly unusual discipline to follow, albeit one that is perfectly rigorous and acceptable in the sense of that word. I'd even go so far that most mathematicians (who typically are rather platonistic about their belief in numbers) would not even claim such nonstandard numbers to have (platonic) existence as numbers, though they do admit them as a possible fiction which (when necessary) can be seen as names for certain technical mathematical constructions involving ultrafilters. (Don't worry: I will explain what these are.)