Besides perhaps the mistaken notion that the inflation of national currencies is a bad thing,

the fixed money supply is something that a computer scientist would have chosen for ease of implementation.  An unbounded money supply would have been a pain to program; he could not have used a fixed block and transition layout with 64-bit long ints, he would have needed a variable-length multiword integer, in the layout and in the implementation.

That is only one of many details in the protocol that suggest that it was never meant to be more than a computer experiment.

Take, for example, the abrupt halving of the reward every 4 years: it is a source of instability and uncertainty, the 2016 halving is already having people worried now.  Why didn't he specify a more gradual decrease, that would give the same 0.5 factor in 4 years but in small monthly steps? The only reason I can think of is that it would have required another page or two of source code, and a table of 48 steps, properly rounded, and another paragraph in the white paper...