-=> Aaron Thomas wrote to Mike Powell <=-

 AT> The phrase "AI" is ambiguous.

Oh, very much so.  Go back to the 1980's and the ELIZA program was touted as
"an AI".

 AT> I did a Google search of "Which language
 AT> is AI written in" and it said "Python," but is that accurate? It seems
 AT> like there are better programming languages to use for tasks like
 AT> emulating human thoughts.

It's not the input.  It's the establishing the relationships between things.

Unlike a simple automated task, AI's learn.  They do so by looking at their
input over and over again and establishing relationships between things that
occur.  Ex: Feed in a thousand recipes that have been proven to be popular.The
AI will figure out what combination of flavors are good - and come up with some
combinations that we didn't think of.

IHMO: They should be using Julia, not Python.  Julia is Python-like, but
FORTRAN performance.

 AT> I get what you're saying, and it's a good guess, but to get to the
 AT> bottom of this I'll have to (someday) learn why AI programs are more
 AT> resource dependent than others.

Think of a the fastest gaming rig you can think of.  Now multiply that by 100
and that's an AI server.

Remember what I said above: The AI is constantly reviewing its inputs to look
for relationships.  New input (which is happening all the time) means it needs
to review its data for relationships.  It strengthens some, lessens others, and
creates new.  It's constantly using all those GPUs (Yes, they usually use
graphics processing units) as much as possible.


... Would it help if I got out and pushed?
___ MultiMail/Linux v0.52

--- Mystic BBS/QWK v1.12 A47 2021/12/25 (Windows/32)