Thursday, November 17, 2022
thomasglee wrote: Thu Nov 17, 2022 8:08 am
Hate speech will be whatever the
programmer of the AI decides is hate
speech. Which means it will still be up
to some form of interpretation. I can't
believe anyone, much less John, would
think this is a good idea. As a
programmer, when should know that
garbage in = garbage out.
That's not how it works. The code will
not make any decisions about hate
speech. If properly implemented, the
code will perform AI functions such as
natural language recognition and
processing, and will translate the
English into some kind of canonical
structure that can be processed by other
components. This actually is how
language translation program works. One
component translates the input language
into this canonical structure, and then
another component translates the
structure into the foreign
language. That way, the code doesn't
have to know how to translate one
language into another.
Separate from the code will be a rules
file, or a set of rules files. These
files will contain millions of rules
that can be read and understood by non
programmers . For example, one rule
might be something like : the N word is
hate speech. It's important to emphasize
that this decision Is not made by the
programmer. It's made by the person or
persons or committee that creates the
rules file. Furthermore, as the meaning
of hate speech changes over time, the
code remains unchanged.