The bits 11001010 01101101 don’t mean anything without a dictionary. That same
pattern could be used to represent text, numbers, part of an image, part of a
mp3 file, or an IP address.
Engineers (often long-ago engineers) defined standards for what bytes should
mean what. For example the ASCII standard (man ascii) says which 7-bit pattern
indicates which (majuscule or minuscule) English letter.
MPEG-4 is another standard.
SQLite (a common database) defines its standard [here].
If you don’t know what the bits will be used for---MPEG, PNG, MP3, sqlite3,
plain ascii, Big 5, unsigned integer, integer, float----then those bits have
come from any of those formats . (Unless you are a code
cracker; then you might do the work yourself of figuring out that ajelks isn’t
a likely word so the underlying bits must be of a different type.)
There’s nothing mathematical about these encodings---nothing spiritual, eternal, or necessary. You couldn’t infer the ASCII encoding from logic.
You declare int x to tell C how to interpret those bits.
How a language should treat types is debated in PLT.