I'm a grad student in philosophy interested in AI. I was considering
taking a course in either Lisp or Prolog. Now, even though I'm asking
this in comp.lang.lisp, which should I learn and why?
Also, considering the academic market, I may need to turn my
programming skills into a private-sector job later on. To this end, I
believe Lisp is much more widely used than Prolog-- is this correct?
Much thanks in advance for any feedback!
Jason
Jason Fossen asks:
>Lisp or Prolog, which should I learn and why?
>Also, considering the academic market, I may need to turn my
>programming skills into a private-sector job later on. To this end, I
>believe Lisp is much more widely used than Prolog-- is this correct?
Take Lisp over Prolog now (you can write Prolog in Lisp if you need to
much easier than Lisp in Prolog), there is simply no better vehicle
for forming and manipulating symbolic abstractions than Lisp.
For the job market, learn Smalltalk (seriously, no one uses Lisp
anymore, unfortunately).
Regards,
Raj
[Rajendra Wall]
| For the job market, learn Smalltalk (seriously, no one uses Lisp
| anymore, unfortunately).
subscribe to the AI-JOBS and LISP-JOBS mailing lists off of CS.CMU.EDU for
a second opinion, from evidence of actual openings.
#<Erik 3017263583>
--
the difference between business and standardization is that businessmen
want to avoid beating dead horses, while standards prescribe the precise
means to deliver corporal punishment to deceased equines.
In article <················@naggum.no> ····@naggum.no "Erik Naggum" writes:
> subscribe to the AI-JOBS and LISP-JOBS mailing lists off of CS.CMU.EDU for
> a second opinion, from evidence of actual openings.
I'm on that mailing list, and I agree that there appear to be jobs
available for Lisp programmers. I don't know how many applicants
there are for each job advertised, but that's a different question.
Of course, many of the Lisp and AI jobs are in the US, which is worth
remembering. There may be good reasons for this, like the funding that
AI has received in the US.
Meanwhile, in the UK, there's a distinct _lack_ of funding, and some
say this is coz of a comment that Marvin Minsky made to a certain Prime
Minister, concerning AI keeping humans as pets. Here's a small tip for
any AI gurus visiting Britain: try to avoid making contraversial statements
to any politicians! They may decide, as Mrs Thatcher perhaps did, that
AI research is not a healthy endevour.
Alternately, any comments about the military applications of AI will
be very welcome. ;-) If the military can find a use for these ideas,
perhaps the commercial world can, also...
Of course, I may have misheard all of this, so perhaps Marvin Minsky
didn't say anything to Mrs Thatcher to make her kill British AI.
I'd appreciate any information on this matter. Perhaps misinformation
is being spread by people with a grudge against Minsky and/or Thatcher.
Perhaps I've also been mislead by a BBC documentary in which it was
suggested that AI has been funded in the US in order to develop smart
weapons, as a response to the failure of the CIA's "brainwashing"
research.
I dunno. However, I do like the idea of using "smart weapons" research
as an argument in favour of Lisp programming. ;-) It could at least
make some "advocacy" debates a little more interesting! I'd welcome
a change from the old "bit twiddling" arguments.
Could I use such an argument to justify programming in Lisp to a
potential employer? I dunno, but if I have to, I may try it, just
to see what the reaction might be.
--
<URL:http://www.demon.co.uk/community/index.html>
<URL:http://www.enrapture.com/cybes/namaste.html>
"You can never browse enough."
Yes, AI in the US (and thus Lisp use) was almost exclusively DARPA
(aka Militarily) funded. While it was good while it lasted, it left a
big headache. Being at the public trough encouraged non-commercially
viable solutions, pushing the technology away from delivering
solutions that could survive on their own in the market.
We don't have a good "Visual Lisp" because we were off building
Pilot's Associates, Robotic Air Vehicles, Smart Weapons, etc. Remember
Lisp Machines? Now *there's* a commercial product (not). But "Lisp
Machine on a Chip" made sense for the military "market".
My advice is give up on trying to "get funding" and instead
concentrate on building something that real people will want to buy.
Regards,
From: Scott Nettles
Subject: Re: Smart Weapons [was Re: Lisp and AI: Newbie Question]
Date:
Message-ID: <40mgm7$4q8@netnews.upenn.edu>
In article <··········@fohnix.metronet.com>,
Rajendra Wall <·····@fohnix.metronet.com> wrote:
>big headache. Being at the public trough encouraged non-commercially
>viable solutions, pushing the technology away from delivering
>My advice is give up on trying to "get funding" and instead
>concentrate on building something that real people will want to buy.
I'm not sure that I buy this argument.
The area of computer architecture provides some good counter examples. For
example, once there was this defense funded project at Stanford that developed
a workstation and another that developed a microprocessor. These days those
projects still have basically the same names, SUN and MIPS. And SGI, well Jim
Clarke developed some chips at Stanford and even had DARPA help getting them
fabbed. And that SPARC on your desk, looks a lot like a Berkeley RISC to me.
Furthermore, none of that work was "something real people wanted to buy" at the
time it was done. In fact a lot of it flew directly in the face of the current
conventional wisdom.
I think a good case can be made that computer science in the US is what it is
because of DOD money. Now that might be good or bad, but it's hard to call it
commercially unsuccessful. Your simple economics would suggest that it would
be a failure.
So maybe Lisp's problems aren't explained by such simple economic notions.
Personally, I like the approach the architects have taken, which is basically
showing that you can't afford to do it any other way. Of course, that requires
a lot of hard work, and funding...
Scott
·······@cis.upenn.edu
Scott Nettles <·······@saul.cis.upenn.edu> wrote:
>Rajendra Wall <·····@fohnix.metronet.com> wrote:
>>My advice is give up on trying to "get funding" and instead
>>concentrate on building something that real people will want to buy.
>
>I'm not sure that I buy this argument.
>
>The area of computer architecture provides some good counter examples.
Wrong. The success of Sun was due to the people involved, not previous
DARPA funding.
>I think a good case can be made that computer science in the US is what it is
>because of DOD money. Now that might be good or bad, but it's hard to call it
>commercially unsuccessful. Your simple economics would suggest that it would
>be a failure.
And your "complex" economics would have us believe "Visual Ada" is the
best selling development environment, wouldn't it?
>So maybe Lisp's problems aren't explained by such simple economic notions.
>Personally, I like the approach the architects have taken, which is basically
>showing that you can't afford to do it any other way. Of course, that requires
>a lot of hard work, and funding...
Oh, yes, the architects have had great success: $100MM later and the
Lisp Machine On A Chip is a big hit. Even small appliances have them
now, don't they. And as for not being able to do it any other way,
that's why Intel is such a loser in the processor market, isn't it.
Let's get this back on track, since my "simple economics" are too hard
for you to understand. The question was, didn't the military fund AI,
with things like Smart Weapons? My answer was, yes, and because DARPA
took the technology off into such screwy directions contact was lost
with the real world market. When DARPA funding was lost the military
systems had no commercial value. (How many Smart Weapons do you have in
your garage?)
I am not arguing "you'll never get lucky and make something useful feeding
at the public trough", but "did military funding help wreck AI in the
'80s"? The answer is clearly yes. When the AI-hype crashed it took
Lisp with it. Clearly unfair, but that's what happened. Lisp would
have been better served by "symbolic spreadsheets" or "presentation
assistants" or something that had value after the Cold War ended.
Regards,
Raj
From: Scott Nettles
Subject: Re: Smart Weapons [was Re: Lisp and AI: Newbie Question]
Date:
Message-ID: <40ol16$r58@netnews.upenn.edu>
In article <··········@fohnix.metronet.com>,
Rajendra Wall <·····@fohnix.metronet.com> wrote:
>Wrong. The success of Sun was due to the people involved, not previous
>DARPA funding.
But you told me the reason Lisp failed WAS a direct result of having DARPA
funding. You really can't have it both ways. Personally I think the issues
are more complicated than your simple DOD funding is bad, explanation. Now it
seems you agree. Thanks.
>And your "complex" economics would have us believe "Visual Ada" is the
>best selling development environment, wouldn't it?
Nope, I didn't propose any "complex" economics, you made that up. I only
suggested that your claim that getting government money leads to a lack of
commercial viability was incorrect. But you've conceded this point already, so
there is no need to argue.
>Oh, yes, the architects have had great success: $100MM later and the
I wasn't talking about the lisp machine architects, they were a dismal failure.
But those other guys, the ones I mentioned explicitly in my post, they are
doing pretty well.
>that's why Intel is such a loser in the processor market, isn't it.
I never said that NOT having DOD money caused you to fail, just that having it
DOESN'T cause you to fail. It doesn't cause you to succeed either, but if you
never get a chance to try out your ideas success is pretty hard to achieve.
>for you to understand. The question was, didn't the military fund AI,
Lisp is not equal to AI, there are a lot of us that are big Lisp fans without
being big AI fans. As for the rest of this, maybe the fall of AI brought Lisp
down, maybe not. Lisp has been declared dead more than once. I do think that
Lisp is more likely to gain widespread use if it isn't so tied to AI.
>I am not arguing "you'll never get lucky and make something useful feeding
I'm arguing something different. Despite the failures of AI, the funding that
D(ARPA) provided to computer science and related fields has created something
of great lasting value, and not just by "getting lucky". I would suggest that
the failure of AI is the exception rather than the rule.
BTW, I'm not disagreeing with the other person who suggested that times may be
different now. But that wasn't the claim I was countering.
Scott
·······@cis.upenn.edu
From: Brian Leverich
Subject: Re: Smart Weapons [was Re: Lisp and AI: Newbie Question]
Date:
Message-ID: <40onjq$ab0@rand.org>
Rajendra Wall (·····@fohnix.metronet.com) wrote:
: I am not arguing "you'll never get lucky and make something useful
: feeding at the public trough", but "did military funding help wreck AI
: in the '80s"? The answer is clearly yes. When the AI-hype crashed it
: took Lisp with it. Clearly unfair, but that's what happened. Lisp
: would have been better served by "symbolic spreadsheets" or
: "presentation assistants" or something that had value after the Cold
: War ended.
I'm not sure I'd blame DoD dollars or the AI-hype crash for the death
of Lisp. That lets the community off the hook too easily.
We're fundamentally a dying breed because we can't get our collective
act together and reasonably support MS-Windows (I know, I hate it too,
but any fool can see that Unix workstations have all the long-run
viability of MVS mainframes ... ) and because even on Unix boxes we
can't deliver the quality and stability of compilers, development
environments, and toolkits that other programming communities have
developed.
We're losing to a large degree because we're fiddling as Rome burns.
I think if Lisp is going to have a future other than simply as an
academic curiosity, the community needs to find niches where we have a
very strong competitive edge and then get serious about building good,
stable, and reasonably-priced environments that support real
applications.
WRT applications you might have been pointing in the right direction
with "symbolic spreadsheets" and "presentation assistants", but I
think there may be even better applications in heavily symbolic areas
like cgi-bin servers, branches of numerical analysis where efficient
manipulation of list structures is crucial, and in areas that require
both symbolic and list manipulation such as simulation modeling. -B
--
Dr. Brian Leverich
Information Systems Scientist, The RAND Corporation
········@rand.org
Brian Leverich wrote:
>I'm not sure I'd blame DoD dollars or the AI-hype crash for the death
>of Lisp. That lets the community off the hook too easily.
I understand your point. It wasn't just the DoD dollars, but they led
to a deadly embrace with corporate R&D dollars as they both went down
the military "product" rat hole. As far as the AI-hype's effect on
Lisp goes, all I can tell you is as the military dollars (for all
things) dried up the long knives came out to eviserate anything having
to do with AI (this at Texas Instruments). Lisp, Lisp machines,
commercial application possibilities, everything. A purely emotional
backlash.
>We're fundamentally a dying breed because we can't get our collective
>act together and reasonably support MS-Windows (I know, I hate it too,
>but any fool can see that Unix workstations have all the long-run
>viability of MVS mainframes ... ) and because even on Unix boxes we
>can't deliver the quality and stability of compilers, development
>environments, and toolkits that other programming communities have
>developed.
Why isn't there the equivalent of Digitalk's MS-Windows Smalltalk for
Lisp?
>WRT applications you might have been pointing in the right direction
>with "symbolic spreadsheets" and "presentation assistants"
As the bombs rained down on our rubble-strewn AI/Lisp bunker we
desperately tried to develop the "super application" that would defeat
the rapidly advancing c++ army. However, so great was the fury
directed towards us that even after signing contracts with 4 European
airlines for symbolic spreadsheet applications the product was killed.
Regards,
Raj
From: Vassili Bykov
Subject: Re: Smart Weapons [was Re: Lisp and AI: Newbie Question]
Date:
Message-ID: <40rfvq$7mv@stratus.CAM.ORG>
·····@fohnix.metronet.com (Rajendra Wall) writes:
>Why isn't there the equivalent of Digitalk's MS-Windows Smalltalk for
>Lisp?
Actually, there's STk, which is rather close and can be brought to the
same level. But not in Windoze. (Feeling Schemish in spare time and doing
Smalltalk for living, I also often used to ask the same question. I was
really pleased to see STk).
--Vassili
In article <··········@fohnix.metronet.com>
·····@fohnix.metronet.com "Rajendra Wall" writes:
> I am not arguing "you'll never get lucky and make something useful feeding
> at the public trough", but "did military funding help wreck AI in the
> '80s"? The answer is clearly yes. When the AI-hype crashed it took
> Lisp with it. Clearly unfair, but that's what happened. Lisp would
> have been better served by "symbolic spreadsheets" or "presentation
> assistants" or something that had value after the Cold War ended.
Have you noticed how many apps today have so-called "intelligent"
features? I'm not that impressed by them, but perhaps they use the
word to mean "relatively intelligent"? I dunno.
Meanhile, market for a new style of app is opening up: agents and
groupware. The later has been getting publicity for some time now,
but as far as I'm aware, agents have only received coverage in the
mainstream magazines in the last few years.
Isn't it possible that some of this "new" software could be written
in Lisp? I sometimes suspect that the problem with Lisp today is a
marketting one (see my comments about marketing in earlier posts).
The push for Lisp appears to be coming from outside of "mainstream"
computing, like acedemia. This has little (or no) credibility with
the mainstream programmers that I know, and such people could be
determined to keep Lisp (and any other language they don't trust) out
of the mainstream.
As someone else has said, the computer industry can be a very
conservative one. It may be easier to add "Lisp"-like features to
existing languages than to get Lisp accepted by the people who
"know" what works and what doesn't.
I'd rather not tell such people that I use Lisp, and wait for them
to ask which language I use to write a particular program. I'd like
to suprise them. ;-) You never know, it may open their minds.
--
<URL:http://www.demon.co.uk/community/index.html>
<URL:http://www.enrapture.com/cybes/namaste.html>
"You can never browse enough."
Scott Nettles (·······@saul.cis.upenn.edu) wrote:
: In article <··········@fohnix.metronet.com>,
: Rajendra Wall <·····@fohnix.metronet.com> wrote:
: >big headache. Being at the public trough encouraged non-commercially
: >viable solutions, pushing the technology away from delivering
: >My advice is give up on trying to "get funding" and instead
: >concentrate on building something that real people will want to buy.
: I'm not sure that I buy this argument.
: The area of computer architecture provides some good counter examples. For
: example, once there was this defense funded project at Stanford that developed
: a workstation and another that developed a microprocessor. These days those
: projects still have basically the same names, SUN and MIPS. And SGI, well Jim
: Clarke developed some chips at Stanford and even had DARPA help getting them
: fabbed. And that SPARC on your desk, looks a lot like a Berkeley RISC to me.
: Furthermore, none of that work was "something real people wanted to buy" at the
: time it was done. In fact a lot of it flew directly in the face of the current
: conventional wisdom.
All of this is good history, but what it suggests to me is that academia
remains a good place for innovation because it exists apart from the quarter-
to-quarter commercial pressure to ship a product. The fact that these efforts
were to some extent DOD/DARPA funded seems an orthogonal observation.
It's also worth mentioning that the examples you cite are examples of general
computer architectures, whereas the example that the original poster cited were
of efforts tailored explicitly to the perceived needs of the DOD/DARPA market,
and this might have made a crucial difference in their long-term success.
: I think a good case can be made that computer science in the US is what it is
: because of DOD money. Now that might be good or bad, but it's hard to call it
: commercially unsuccessful. Your simple economics would suggest that it would
: be a failure.
The same logic lies in the statement that "microelectronics and medicine in the
US is what it is because of the space program." They're both true statements,
but they're also both historical. It's not at all clear that the kind of
military technology transfer that fueled the commercialization of computer
science roughly from the mid-'50s to the mid-'70s still pertains to today's
hardware or software markets.
: So maybe Lisp's problems aren't explained by such simple economic notions.
: Personally, I like the approach the architects have taken, which is basically
: showing that you can't afford to do it any other way. Of course, that requires
: a lot of hard work, and funding...
Not to mention a lot of "throw it out there in the market and see what
happens," which may be good capitalist economics, but most microcomputer
software-development companies are already far too conservative to embrace.
: Scott
: ·······@cis.upenn.edu
Paul Snively
Activision, Inc.
········@activision.com
Jason's article got a lot of responses but they didn't seem to me to
answer the question he posed. Seeing I was in exactly the same
position ten years ago, I've tried to address it more directly. I'm
posting my response because there may be others out there in a similar
position.
In article <··········@geraldo.cc.utexas.edu>,
Jason Fossen <·······@mail.utexas.edu> wrote:
>I'm a grad student in philosophy interested in AI. I was considering
>taking a course in either Lisp or Prolog. Now, even though I'm asking
>this in comp.lang.lisp, which should I learn and why?
I don't know to what extent you are a logician. Prolog claims to be an
implementation of First Order Predicate Calculus, but of course FOPC
isn't computable (Turing). So it is FOPC with a 'closed world
assumption' -- which is to say 'anything I don't know to be true, and
can't infer from other things I already know, is false'. Furthermore,
the resolution algorithm used to implement Prolog isn't very
efficient. Consequently, most real-world Prolog programs make heavy
use of the 'cut' operator, which defeats the concept both of a logic
language and of declarative coding.
In other words, for a logician Prolog is an ugly abortion. I *hate* it
-- not really because it's bad as because it's such a disappointment.
However, if you want to learn it, try:
Gibbins, P: _Logic with Prolog_: Oxford University Press, Oxford,
1988. ISBN: 0-19-859659-6
Lisp on the other hand I love. The first two or three weeks of
learning Lisp will probably tie your head in knots, and then it will
suddenly become clear. Lisp is a beautiful language for expressing and
exploring philosophico-logical problems. It is very clean and clear
and elegant.
It also keeps out of your way when you are working. You can
effectively ignore all the undergrowth of making the machine run, and
just describe the problem. You can work top down or bottom up. You can
run unfinished bits of code while you work out what the bits you
haven't tackled yet will do.
However, having said all that, Common LISP is a bit of an ugly
monster. <quavery voice on> Back in the good old days when I were a
young 'un we used to have a standing joke that InterLISP was like a
1950's Cadillac -- covered with chrome and glitz and with more bells
and whistles than you could shake a stick at, but if you wanted
something it was sure to be there; while Portable Standard LISP was
more like a vintage Bugatti -- clean and elegant, if a little spartan,
but by God it would get you there.
The modern comparison would be between Common LISP and Scheme. Scheme
is in my opinion the cleanest, most elegant Lisp yet (although it
doesn't support programming idioms which are dear to my antediluvian
heart), and it's where I would advise you to start.
For books, there's lots of good Lisp books, and for Scheme you must
not miss the classic
Abelman, H and Sussman, G J: _Structure and Interpretation of Computer
Programs_ : MIT Press, Cambridge, Massachusetts, 1985. ISBN: 0-262-01077-1
However that's a bit daunting for beginners, and I'd start with much
simpler things if I were you. Try:
Friedman, D P & Felleisen, M: _The Little LISPer_: MIT Press,
Cambridge, Massachusetts, 1987. ISBN: 0-262-56038-0
(which is very much an introductory text, but great fun) and
Coxhead, P: _Starting LISP for AI_: Blackwell Scientific Publications,
Oxford, 1987. ISBN: 0-632-01544-6
(also a beginners text, with good clear style, but takes you up to
the beginnings of knowledge representation, inference, and language
parsing. Also, usefully, explains how to use it's example programs
with many non-standard LISP variants)
Have fun! Remember, computers are only toys. Computers (and languages)
which are not fun to work with are only unsuccessful toys.
--
------- ·····@rheged.dircon.co.uk (Simon Brooke)
-- mens vacua in medio vacuo --