On Sun, Aug 14, 2011 at 8:50 AM, Paulo Pinto <paulo.jpi...@gmail.com> wrote:
> I guess that nowadays many AI systems are mainly programmed in
> some kind of specialized DSL.
>
> Sure Lisp based languages are a perfect candidate for it, but the
> plain
> mention of Lisp brings up some issues that you cannot get rid of, like
> the parenthesis.
And what, exactly, is wrong with parentheses?
Put another way: the syntax tree of a program will have some sort of
particular structure. To the extent that parentheses DON'T determine
the sub-branches, something else WILL. What do other languages use for
that "something else"?
* Other delimiters, either characters as in C's { ... } blocks, whole
words as in "if ... fi" in some shell languages, or other things.
* Significant whitespace, most commonly in languages that use newlines
to delimit statements (BASIC) but sometimes going much further
(Python).
* Syntax rules where the positions of various things among various
keywords determines the structure.
* Precedence rules among math operators.
Drop the first one (Clojure also uses other delimiters) and what we're
left with is, mostly, icky. Significant whitespace means an editor
line-wrapping or, in some cases, even reindenting code might change
its semantics. Syntax rules and precedence rules burden the developer
with remembering them all, and beyond simple and fairly standardized
rules such as "* before +" precedence rules tend to be avoided by
defensive use of ... parentheses. The one language family with a small
and regular syntax other than Lisp seems to be Smalltalk, and it
avoids the parentheses, with devastating results: 1 + 3 * 4 comes out
as 16 instead of 13. Yikes! They just use strict left to right
evaluation to avoid having complex rules.
> To be honest, while I was at the university I always preferred Prolog
> to Lisp, due to the close relationship some our professors had with
> Edinburgh's university.
This seems to be a total non sequitur, like saying you always
preferred beef to chicken due to the US/Canada border being wiggly
instead of straight east of North Dakota.
In other words: How, exactly, are your former university's professors
and Edinburgh's university causally related to a personal preference
for Prolog? There's no obvious reason why that would be. Does
Edinburgh's university have a strong focus on Prolog for some reason,
which "rubbed off" via their professors and your university's
professors to you?
> I think it is better that the students learn AI than a new programming
> language
> with forces them to think in a different way. Learning multiple
> concepts at
> the same time is not easy and you might loose students along the way
> because of it.
>
> Who knows, maybe some of those students will eventually find their way
> to
> Clojure/Lisp.
If they go into AI, it's very likely that will expose them to Lisp at
some point, yes.
I doubt, on the other hand, that Prolog is a very good language to
use, odd though that may seem. The problem is that an AI you try to
develop in Prolog will probably be strongly influenced by the language
choice towards being a theorem-prover type of system with bells on,
and natural intelligence Does Not Work That Way. Natural intelligence
guesses and intuits and invents whole new concepts to create
simplified models that predict its past sense data, and tests them
against future sense data. Self-aware intelligence adds a simplified
model to predict its own past feelings and behaviors (and has
autobiographical memory so these can be "past sense data"), and tests
them against future sense data (and, potentially, also against
simulations spawned using the world models -- think about that as you
go to sleep tonight). Formal logical reasoning, Prolog style, is
actually something we had to invent, rather than something we had
innately. I doubt we'd be anywhere near as prone to various mistakes
and fallacies in reasoning if we were based on Prolog! On the other
hand we'd probably not be creative.
In simpler language, logic-based AI has been "done to death" and
doesn't seem to lead to anything fundamentally new in terms of
software capabilities (i.e., closer to what we can do, able to
automate more that we currently have to do ourselves). Of course it
can be done in principle, since Prolog is Turing-complete, but it may
be easier in some other language.
Lisp, of course, isn't just "some other language" but a kind of
language-mother due to the ease with which it spawns DSLs, so the best
shot probably actually still lies with Lisp. The most advanced AI we
ever made, able to reason in the most humanlike ways and to invent new
concepts, was Eurisko, and that was programmed in Lisp. In fact,
Eurisko was so promising, and we haven't even come close to equaling
that achievement in the many years since, that I still wonder if the
government or aliens (:)) or time travelers (:)) or someone put the
kibosh on that line of research out of fear of Skynet (:)) or
something.
--
Protege: What is this seething mass of parentheses?!
Master: Your father's Lisp REPL. This is the language of a true
hacker. Not as clumsy or random as C++; a language for a more
civilized age.
--
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
