Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

is comp.lang.lisp dead ?

161 views
Skip to first unread message

arnuld

unread,
Mar 15, 2018, 2:56:28 AM3/15/18
to
6 or some years ago, I used to post Common Lisp code on this newsgroup
and I remember people replying with better code and better ideas. At that
time I used to be on comp.object too because I was interested in CLOPS
and MOP. Today I see comp.lang.lisp seems dead :( . No good post in this
new year.

comp.object died several years ago because people moved onto better
things, newsgroup to forums and IRC, OO to something else. AI and
Robotics boom (2nd boom to be exact) has finally come after a long gap,
Machine Learning, Deep Learning and Neural Networks are the buzzwords
these days. Lisp is the language of AI, Uncle John McCarthy created it to
solve AI problems in first place and I was expecting this newsgroup to
be much busier now. But what happened ?

Pascal J. Bourguignon

unread,
Mar 15, 2018, 3:46:32 AM3/15/18
to
Gavino himself and Robert L.

--
__Pascal J. Bourguignon
http://www.informatimago.com

Bigos

unread,
Mar 15, 2018, 7:22:03 AM3/15/18
to
:-D

Spiros Bousbouras

unread,
Mar 15, 2018, 7:22:38 PM3/15/18
to
On Thu, 15 Mar 2018 06:56:21 +0000 (UTC)
arnuld <sun...@invalid.address> wrote:
> 6 or some years ago, I used to post Common Lisp code on this newsgroup
> and I remember people replying with better code and better ideas. At that
> time I used to be on comp.object too because I was interested in CLOPS
> and MOP. Today I see comp.lang.lisp seems dead :( . No good post in this
> new year.

I consider good any post from which I learn new stuff or makes me consider
[reasonable] view points I hadn't considered before. With this criterion ,
comp.lang.lisp had plenty of those although the overall activity is lower
than what it used to be in years past. Also , any on topic post will receive
replies so I think it's absurd to call the group dead.

For a very recent example , what do you think of <lz4llti...@gnu.org> ?
I think it's a very good post. It is mainly a general algorithmic one rather
than Lisp specific but who cares , you saw it here first.

> comp.object died several years ago because people moved onto better
> things,

Yeah , functional is the new black.

> newsgroup to forums and IRC, OO to something else. AI and
> Robotics boom (2nd boom to be exact) has finally come after a long gap,
> Machine Learning, Deep Learning and Neural Networks are the buzzwords
> these days. Lisp is the language of AI, Uncle John McCarthy created it to
> solve AI problems in first place and I was expecting this newsgroup to
> be much busier now. But what happened ?

Are you sure that's why he created it ? I thought he was just doing theoretical
work and someone else [a student of his ?] used that to write an actual working
Lisp interpreter.

Anyway , as has been pointed out on this group many times , many of the
aspects which made Lisp languages special , other languages now have them
too. To my surprise , many people [including Peter Norvig apparently] aren't
that bothered about powerful metaprogramming facilities. The latter was one
of the main things which attracted me to Lisp.

I note finally , that if the activity on a discussion forum is not influenced
by which are the latest buzzwords then that's probably a good thing. The
buzzwords may be such for good reasons , namely refer to things which will
have lasting value , but some people will be drawn to them just because they
are buzzwords and not because they are interested in the underlying depth [if
such exists] .

Helmut Eller

unread,
Mar 15, 2018, 7:36:01 PM3/15/18
to
On Thu, Mar 15 2018, arnuld wrote:

> Lisp is the language of AI, Uncle John McCarthy created it to
> solve AI problems in first place and I was expecting this newsgroup to
> be much busier now.

Lisp was intended for symbolic AI, i.e. problems that involve symbols
and logic. The kind of AI that is in fashion today is sub-symbolic,
i.e. mostly works with numbers and statistics. Lisp is not particularly
good for statistical problems.

Helmut

Kaz Kylheku

unread,
Mar 15, 2018, 8:13:50 PM3/15/18
to
Right; this domain screams for ... Python coupled with C libraries.

Pascal J. Bourguignon

unread,
Mar 16, 2018, 3:03:04 AM3/16/18
to
You can generate cuda code with a lisp program much better than with
python and C…

Robert Munyer

unread,
Mar 16, 2018, 6:17:35 AM3/16/18
to
arnuld wrote:

> 6 or some years ago, I used to post Common Lisp code on this newsgroup
> and I remember people replying with better code and better ideas.

You can still post code here and get good feedback. Try it!

[...]
> AI and
> Robotics boom (2nd boom to be exact) has finally come after a long gap,
> Machine Learning, Deep Learning and Neural Networks are the buzzwords
> these days. Lisp is the language of AI, Uncle John McCarthy created it to
> solve AI problems in first place and I was expecting this newsgroup to
> be much busier now. But what happened ?

McCarthy created Lisp because he wanted a concise, flexible way to
express algorithms that could be understood by people and evaluated
by machines.

The technologies that you mentioned above are focused on the goal of
implementing desired behavior _without_ expressing it as an algorithm.

When you use those technologies, you usually use existing floating-
point number-crunching algorithms that are not specialized for your
desired behavior, and instead of designing your own algorithm that
implements your desired behavior, you crunch floating-point numbers
to get the desired behavior emergently from the existing algorithms.

Look at Joel Grus's 2016 article about Fizz Buzz in Tensorflow.
I think he posted 100% of his source code, but if you look in his
source code for the algorithm that implements the fizzing and the
buzzing, you won't find it... there really isn't any algorithm
there, just lots of floating-point numbers.

If you could time-travel to 1966, hand a printout of Grus's article
to John McCarthy, and say "I want this technology working ASAP", I
imagine McCarthy might read the article and say, "You want FORTRAN,
not LISP. And you want Seymour Cray to design your hardware."

--
-- Robert Munyer code below generates e-mail address

(format nil "~(~{~a~^ ~}~)" (reverse `(com dot munyer at ,(* 175811 53922))))

arnuld

unread,
Mar 16, 2018, 6:44:03 AM3/16/18
to
> On Thu, 15 Mar 2018 08:46:27 +0100, Pascal J. Bourguignon wrote:
> Gavino himself and Robert L.

Ahaa, Pascal, I was waiting for you to reply, so nice to see group is
alive and well. Long live Lisp :)

arnuld

unread,
Mar 16, 2018, 6:51:02 AM3/16/18
to
> On Fri, 16 Mar 2018 00:35:51 +0100, Helmut Eller wrote:

> Lisp was intended for symbolic AI, i.e. problems that involve symbols
> and logic. The kind of AI that is in fashion today is sub-symbolic,
> i.e. mostly works with numbers and statistics. Lisp is not particularly
> good for statistical problems.

That is a very good insight.

arnuld

unread,
Mar 16, 2018, 6:55:13 AM3/16/18
to
> On Fri, 16 Mar 2018 08:02:58 +0100, Pascal J. Bourguignon wrote:

> You can generate cuda code with a lisp program much better than with
> python and C…


Do you mean, as a language Lisp is much better in terms of abstraction or
performance or quality of output ? Evan Miller won't agree with you I
think:

http://www.evanmiller.org/mathematical-hacker.html


I remember reading a paper explaining that how Common Lisp can be used as
basis to create a new programming language required for high performance
analysis of Big Data (it was written after Python was already there for
many years). I lost the link. It read more like prophecy of Julia
language than of need of Lisp based language or libraries. I know GSLL is
there (https://common-lisp.net/project/gsll/) but I don't know its
performance specifically for Data Analysis or Machine Learning

Bigos

unread,
Mar 16, 2018, 9:43:03 AM3/16/18
to
Yes, but is it possible to make lisp better for that using superior
meta-programming and well designed libraries?

I remember trying to solve some AI puzzle on Hackerrank that involved
statistics. I had to write 4 functions or so, to solve most of the
puzzles I tried.

I did not try to make Lisp better for statistics because I was not
interested in it. But can someone combine interest in statistics with
Lisp programming to make Lisp better for statistics?

Stefan Monnier

unread,
Mar 16, 2018, 11:39:35 AM3/16/18
to
> I did not try to make Lisp better for statistics because I was not
> interested in it. But can someone combine interest in statistics with Lisp
> programming to make Lisp better for statistics?

Of course: Lisp is not inherently worse than, say, Python at this game.

But for better or worse, the existing tooling/libraries is much better if
you use Python than if you use Lisp.


Stefan

Christian Lynbech

unread,
Apr 10, 2018, 10:59:19 AM4/10/18
to
Interestingly enough, somebody recently linked to some old posts back
when reddit switched from lisp to python (I probably saw it on Planet
Lisp).

The first post in which the CEO (or something like that) explained the
move, citing in part how python had all these amazing libraries where
lisp typically, if any, would hav only a single option.

Then another post (from Aaron Schwartz) went into some details and there
it was stated that allthough python did in fact had numerous web
frameworks available, none of them were any good and they ended up
writing their own.

Admittedly, the reddit change was done a long time ago and many may have
changed, but me thinks that one should be carefull not to be blinded by
the *number* of supporting libraries, even if everybody else seems to be
going a certain route. If that is so inherently wonderfull, everybody
would be coding in Java, no matter how silly that language is.

After all, as the saying goes, it is hard to get ahead if you are
walking in somebody elses footsteps :-)

------------------------+-----------------------------------------------------
Christian Lynbech | christian #\@ defun #\. dk
------------------------+-----------------------------------------------------
Hit the philistines three times over the head with the Elisp reference manual.
- pet...@hal.com (Michael A. Petonic)

a.daniel...@gmail.com

unread,
Apr 10, 2018, 11:24:22 AM4/10/18
to
On Thursday, March 15, 2018 at 7:56:28 AM UTC+1, arnuld wrote:
> .. I was expecting this newsgroup to
> be much busier now. But what happened ?

I kind of think that there are actually a lot of LISP coders now, more than 15 years ago for instance, as it's getting pretty easy to find LISP libraries on the web. LISP might be becoming a more mainstream language, so maybe folks don't feel they need to join a group to learn and use it. Also I think folks communicate on the newer coding oriented chat/message boards, so 'comp.lang.lisp' has gotten to be less in the middle of it all. (I still hang here, though, at least I do when I hang).

I also think that while there are a lot of LISP coders, there are a lot who have moved to recent dialects (Clojure, for one) and other languages (Haskell, for one).

The population of computer programmers has had exponential growth, like the population in general. It's HUGE. That means a lot of coders for every language.

Even the "traditional" C language family now has a super-complex C++ which can do a lot with static typing that in the old days was best done with dynamic typing in LISP (or Python). So even though I think LISP coding is growing, everything else is growing too.

But I think the "comp.lang.lisp" group - though still alive and kicking! - might not be the main locus.

Lars Brinkhoff

unread,
Apr 10, 2018, 11:44:31 AM4/10/18
to
arnuld wrote:
> I was expecting this newsgroup to be much busier now. But what
> happened ?

One contributing factor of perhaps many is, I think, that Usenet is
falling out of favour. I see a similar trend in other newsgroups.
Maybe there's some Facebook group that's the hotspot.
0 new messages