Google AI winner uses lisp
Alec Battles
week, though strangely ZDnet -- which, on principle, I refuse to link
to -- is one of the only places to write it up.
http://pr-usa.net/index.php?option=com_content&task=view&id=560484&Itemid=99999999
Nice news to read before bedtime.
A
Ken Wesson
> I'm sure a few people have read this news already. It's been up for a
> week, though strangely ZDnet -- which, on principle, I refuse to link
> to
Why?
> -- is one of the only places to write it up.
>
> http://pr-usa.net/index.php?option=com_content&task=view&id=560484&Itemid=99999999
>
> Nice news to read before bedtime.
Yes, but was it Clojure, or another Lisp?
javajosh
http://quotenil.com/
On Dec 9, 7:09 pm, Ken Wesson <kwess...@gmail.com> wrote:
> > I'm sure a few people have read this news already. It's been up for a
> > week, though strangely ZDnet -- which, on principle, I refuse to link
> > to
>
> Why?
>
> > -- is one of the only places to write it up.
>
Ken Wesson
javajosh
On Dec 9, 9:07 pm, Ken Wesson <kwess...@gmail.com> wrote:
> > Common Lisp.
>
> It figures. :)
reading Gabor's post (http://quotenil.com/Planet-Wars-Post-
Mortem.html) and it sounds like something I might enjoy next time.
Perhaps I'll give it a shot with clojure.
The exciting thing is that a) the winning program was Lisp, b) it won
by a lot, and c) 99 out of the top 100 were not Lisp. Now, this could
just mean that Gabor was unusually obsessed with the contest, and the
language just "got out of his way". Personally, I'm ok with that.
Ken Wesson
That is indeed one of Lisp's strengths -- once you're reasonably
proficient with it, the language frequently does just "get out of the
way", in much the way that Java doesn't.
javajosh
aether, alas) that, basically, you can make *any* reasonable solution
sketch work in a Lisp. If you want to keep thinking in Java, fine. You
can do that. If you want get all functional, cool. In the end, all
real programming work is just ordering the execution of imperative,
side-effect generating functions. The ordering and parameterization of
those calls is arbitrary (and fun!).
For Clojure there's the added question of "what level of side-effect
functions will you want?" - e.g. with javax.Swing or
java.awt.Graphics? I don't know anything about Common Lisp but I'm
sure they have the same decisions to make WRT C libraries.
It does beg the question, though: what is a reasonable bare minimum
function set that a real-life lisp would require?
Ken Wesson
> It does beg the question, though: what is a reasonable bare minimum
> function set that a real-life lisp would require?
I think different people might give different answers to that.
The academic computer scientist is likely to consider lambda, the
function-call operator, nil, if, and eval to suffice. After all those
suffice for anything you can express in lambda calculus.
A Lisper interested in Lisp hacking as an end in itself will want to
add cons, car/cdr or first/rest, =, cond, etc.
The application programmer is going to additionally require FFI with,
particularly, GUI libraries, not to mention various forms of disk,
networking, keyboard, and mouse I/O.
The systems programmer is going to additionally require being able to
get at the bare metal and run tight, efficient, non-interpreted code
in kernel mode.
Ultimately I suppose it hinges on what someone means by a "real-life"
Lisp. I'd say the application programmer's needs are probably th best
answer there, since Lisp hacking for its own sake and academic
computer science are more mathematics than "real-world", and systems
programming can be accomodated by a combination of C and a Lisp with
some kind of FFI. (Clojure as the Lisp requires Java as well, and
Clojure and C calling each other indirectly via Java and JNI as a
go-between.)
Jon Seltzer
On Dec 9, 7:09 pm, Ken Wesson <kwess...@gmail.com> wrote:
> > I'm sure a few people have read this news already. It's been up for a
> > week, though strangely ZDnet -- which, on principle, I refuse to link
> > to
>
> Why?
>
> > -- is one of the only places to write it up.
>
javajosh
Wilson MacGyver
C/C++ libs.
> --
> You received this message because you are subscribed to the Google
> Groups "Clojure" group.
> To post to this group, send email to clo...@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+u...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/clojure?hl=en
Alec Battles
>> I'm sure a few people have read this news already. It's been up for a
>> week, though strangely ZDnet -- which, on principle, I refuse to link
>> to
>
> Why?
Because of the website's close ties to Washington.
This blog post is exemplary:
http://www.zdnet.com/blog/government/is-wikileaks-julian-assange-worse-than-osama-bin-laden/9578