The Loop Facility is a controversial extra language that lives
inside lisp. Specifically Zetalisp and Common Lisp's, but the same
idea is true (maybe more true) of the interlisp tradition's for do
what I mean.
I was reading one of the papers shared by Jose A Alonso, one of the
infinite repetitions of asking GPT forked chatbots to try answering
CODING CHALLENGE job-screening-data-mining websites (leet coder dot
com or something in this case).
Aside: Oh, cool application. You could trash job screening data
businesses by using well-motivated chatbots, who are famously hard
to electronically screen against.
Anyway, "
ChatGPT [] gained significant attention upon its release [] due to
its advanced natural language processing and code-writing
capabilities.
". I suppose that business management staff are not aware that
these two things together or at all pre-date Thiele's marketing by
many decades.
This is what Lisp's controversial Loop Facility is and does. Many
programming problems implement nicely using an approach something
like-
Curryingly collect and maybe destructure all these different
states; curry so that every already collected state can be
maybe-used in collecting later states followed by an explosion of
conditional behaviours depending on what those states are, looping,
aggregating, summing various states in passing, and either
returning different conjugations of those under various conditions,
but often CONTINUEing with some automatic notion of logically next
early states until termination.
That last paragraph has high entropy. It is hard to boil that
problem solving strategy - if one even exists in there somewhere -
down to a clear set of behaviours, and the implied requirements are
quite unreasonable- the code that does that is going to be large
and intricately tangled.
Perhaps mutual recursion and some kind of von Neumann state is more
beautiful in lots of cases, and can be cooked on its own?
But in lisp we have the LOOP facility. It just does all those
things. It speaks a special and highly mnemonic language that is
both easy to think problems into and thunk answers out of.
And it does this by accepting an essay in this formally structured
but highly mnemonic language, and deterministically generating
efficient code that does it. The kind of code being asked for is
long, filled with explicit special cases and difficult to write
without this facility.
So the difference is that LISP's LOOP facility speaks its own
formal language, not a naieve but big probabilistic record of
groups of words in inputs,
and LOOP's code generation is deterministic, not a big trained
probabilistic blend of likely conjunctions of output words.
FURTHERMORE consider the cultural distinctions: LOOP is a part of
the lisp community, produced by it, produced for it, and with
authors for and against it littered through the language and
language's history. I believe ANSI Common Lisp 2E's LOOP was due to
Moon, and the Chinual 4E LOOP description is Burke's memo TM-169.
In contrast, GPT created chat-bots are being found to match
cultural aspects of questions and answers. Receiving the highest
quality answers requires culturally matching the highest quality
scraped programming code authors' communities: And tricks like
asking for names to namedrop while rephrasing a code request are
being discovered to strongly increase chat-bot performance.
So far many programmers, especially web design / general purpose
programmers are seeking to detach the requirement of tickling a
chat-bot code request into 1337 language to get a 1337 answer and
avoid problematic Answer Website Company garbage by protecting
human users from the Natural Language question phrasing requirement
by placing it underneath a computer-friendly formalisation.
