LAWS@SRI-AI.ARPA (01/20/85)
From: AIList Moderator Kenneth Laws <AIList-REQUEST@SRI-AI.ARPA>
AIList Digest Sunday, 20 Jan 1985 Volume 3 : Issue 5
Today's Topics:
AI Tools - LOOPS & LISP for PC & Golden Common Lisp,
Reports - Functions in LISP & UTexas Reports & Recent Articles,
Opinion - Reminiscence
----------------------------------------------------------------------
Date: Thu, 17 Jan 85 16:01 EST
From: Araman@HIS-BILLERICA-MULTICS.ARPA
Subject: PORTING OF LOOPS
I would like to know whether LOOPs has been ported on to the Symbolics
3600. I would also like to know whether it has been converted to
ZETALISP or has been ported using the INTERLISP COMPATABILITY PACKAGE.
Pointers, names & phone nos. would do. thanks in advance -sankar (617
671 3018)
------------------------------
Date: Thu 17 Jan 85 00:33:35-PST
From: Sam Hahn <SHahn@SUMEX-AIM.ARPA>
Subject: Lisp for PC
If you're using PC's and looking for a Lisp, I'd suggest
TLC-Lisp, from The Lisp Company. I myself have not used GCLisp,
but have been quite impressed with TLC-Lisp, which has a compiler,
an object-class system, packages, auto-load entities,
and costs less than half what GCLisp costs.
TLC is John Allen's (The Anatomy of Lisp) company, located in
Redwood Estates, CA. I have no connection with TLC except as
a customer.
-- sam hahn
------------------------------
Date: 18 Jan 85 0228 EST
From: Dave.Touretzky@CMU-CS-A.ARPA
Subject: review of Golden Common Lisp
This is a review of Golden Common Lisp; I was one of its beta test users. I
first got involved with GCLisp when I contracted to develop an intensive
week-long tutorial on Common Lisp to be marketed nationally by Carnegie
Group, Inc. The tutorial is designed to be held in a classroom full of
Compaq Plus computers (PC/XT clones) with 640K of memory and 10.5 MB hard
disk drives, running Golden Common Lisp.
I'm used to working on a Symbolics 3600, yet find I am quite comfortable
moving to the PC using this software. GCLisp is a very respectable subset
of the Common Lisp dialect. It includes DEFMACRO, generalized variables
(with SETF, etc.), one-dimensional arrays, named structures, closures,
stack groups, multiple values, lambda lists with &optional/&rest/&aux
keywords, a window package, FORMAT, the sharp sign and backquote macro
characters, i/o streams, and pathname objects. The online documentation
facilities are excellent. Example: you can start to type in a Lisp
expression at top level, like
(+ foo (nth
and, if you forget the order of arguments to the NTH function, you can
hit meta-L and the system will look at the expression you are typing,
find the documentation for the NTH function, and display its argument
list. The GMACS editor (a descendant of Emacs) is reasonably powerful
and contains its own online help facilities. Some of the other programming
tools provided are TRACE, STEP, DOC, APROPOS, DESCRIBE, and a pretty printer.
Now, here are the negative aspects of GCLisp. First, it is not a full
Common Lisp implementation. Missing features include: ratios, bignums,
complex numbers, many FORMAT and # options, some of the sequence functions,
most of the optional keyword arguments to sequence functions, hash tables,
character objects, user-defined packages, and keyword arguments in
user-defined functions. Of course, if Gold Hill had crammed a complete
implementation onto the PC there would be no room left for user code or
data structures. I think they did a good job of deciding which parts of the
language to omit.
Another shortcoming of GCLisp is that it is dynamically scoped (like
MacLisp) rather than lexically scoped (like true Common Lisp). This choice
was made for efficiency reasons; it is hard to do efficient lexical scoping
in an interpreter. Case in point: at AAAI-84 I saw a PC running GCLisp
outperform a Vax running interpreted Vax Common Lisp; both were executing a
recursive Fibonacci function, but the Vax was using lexical scoping. Also,
to be fair, it was an early version of Vax Common Lisp which wasn't yet
tweaked for speed, and the Vax was using 32 bit arithmetic while the PC
used 16 bits. But on the other hand, the PC wasn't just slightly faster
than the Vax; it was a good deal faster. The folks who created T use a kind
of pre-compilation hack to interpret lexically scoped code efficiently. As
far as I know, no Common Lisp implementation uses this trick.
There is a GCLisp compiler in the works. The first release is due in
30-60 days, although that might be just a beta test version. A random
comment one of the Gold Hill people made to me suggests that when the
compiler is ready they might decide to switch to lexical scoping.
A third criticism of GCLisp is that you can't buy just the Lisp; it is
bundled together with a piece of tutorial software known as the San Marco
Lisp Explorer, plus copies of two books (Steele's Common Lisp manual and
the 2nd edition of Winston and Horn.) If you already know Lisp, and you
already have copies of these two books, you have to buy the extra stuff
anyway. I don't know exactly how much the bundling adds to the $495 price
of GCLisp ($395 for academic institutions), but I hope Gold Hill will
reconsider this decision. For large institutions, a site license is
available that permits unlimited copies. We are presently arranging to get
one for CMU; we plan to move two Lisp courses (one for novices and one for
experienced programmers), and the Lisp segment of a third course, onto PC's
using GCLisp. This will provide better computing resources and a higher
quality environment than what is available running MacLisp on our
well-loaded academic DEC-20's.
GCLisp needs at least 512K to run, and you really ought to have 640K (the
max addressable on a PC) in order to use the system to its fullest without
running out of memory too quickly. You don't have to have a hard disk,
although it helps. On the new PC/AT, GCLisp can address up to 1 megabyte of
memory. (You can put up to 3 meg on an AT, but Gold Hill uses some of the
bits in each 32 bit address for type codes and other stuff, so only 20 bits
are available for addressing.)
Gold Hill's address is: Gold Hill Computers, Inc., 163 Harvard Street,
Cambridge, MA, 02139. Telephone (617) 492-2071.
In summary: this is a superb product. It puts state-of-the-art Lisp
programming technology into the hands of anyone who can afford a PC.
-- Dave Touretzky
------------------------------
Date: Wed 16 Jan 85 17:31:31-PST
From: Emma Pease <Emma@SU-CSLI.ARPA>
Subject: Report - Implementation of Functions in LISP
[Forwarded from the CSLI Newsletter by Laws@SRI-AI.]
NEW CSLI REPORT
A new CSLI Report by Michael P. Georgeff and Stephen F. Bodnar, ``A Simple
and Efficient Implementation of Higher-order Functions in LISP'' (Report
No. CSLI--85--19), has just been published. To obtain a copy of this
report write to Dikran Karagueuzian, CSLI, Ventura Hall, Stanford 94305 or
send net mail to Dikran at SU-CSLI.
------------------------------
Date: Fri, 18 Jan 85 06:28:48 cst
From: Laurence Leff <leff%smu.csnet@csnet-relay.arpa>
Subject: Recent UTexas Reports
Technical Reports
U. T. Austin
TR-84-29 Translating Horn Clauses from English Yeong-Ho Yu
TR-84-30 Automated Proof of a Trace Transformation for a Bitonic Sort
Chua-Huang Huang and Christian Lengauer
TR-84-33 From Menus to Intentions in Man-Machine Dialogue Robert F. Simmons
TR-84-35 Modelling Concepts for VLSI CAD Objects
D. S. Batory and Won Kim
------------------------------
Date: Thu 17 Jan 85 17:30:40-PST
From: Ken Laws <Laws@SRI-AI.ARPA>
Reply-to: AIList-Request@SRI-AI.ARPA
Subject: Recent Articles
Communications of the ACM, January 1985.
IJCAI Award for Research Excellence Announced, p. 123:
$1000 plus expenses to be awarded every second year to honor
sustained excellence in AI research.
AFIPS Publishes Volume on Artificial Intelligence, p. 123:
Describes "Artificial Intelligence", volume VI in the AFIPS
Information Technology series, a collection of NCC and joint
USA/Japan Computer Conference papers edited by Oscar Firschein
[then at Lockheed, now at SRI]. AFIPS Press Department, 1899
Preston White Drive, Reston, VA 222091, $23 (plus $3.50 if not
prepaid). [The collection includes three 60's papers on solving
geometic analogy problems, decomposing line drawings in 3-D
polyhedra, and representation in GPS; five on knowledge engineering
and expert systems; two on planning and problem solving; two on
AI languages; six on applications; and six on image understanding.]
IEEE Computer Graphics and Applications, January 1985:
Consensus Bibliography Technique Aids Access to Literature, p. 94:
Ware Myers describes a series of annotated bibliographies being
compiled by John L. Burch's Ergosyst Associates. (No address is
given, but they have a division called the Report Store, 910
Massachusetts St., Lawrence, KS 66044, that sells copies of the cited
documents.) The company identifies top people in a field and
uses the sources they cite as the core of their bibliography.
Each citation they include is accompanied by a capsule review of
about 1 or 2 pages, and the volume as a whole has a short introduction
to the field (about 17 pages). They have produced two so far,
including "Artificial Intelligence: Bibliographic Summaries of the
Select Literature" by Henry M. Rylko, 210 entries of about 2.4
pages each.
------------------------------
Date: Wednesday, 16 Jan 85 22:03:25 EST
From: shrager (jeff shrager) @ cmu-psy-a
Subject: <no comment required!>
[Comment: rather long and pointless reminiscence. -- KIL]
Forwarded from the CMU opinion bboard.
Date: Wed 16 Jan 18:10
From: Purvis.Jackson@CMU-CS-CAD
Subject: trees (3416)
Perhaps what makes me most insecure about artificial intelligence is
the feeling of my general lack of natural intelligence that becomes
most pronounced when I'm in the company of those well versed in the
ideas and language of AI. It seems as though I can be made to feel
somewhat small, rather insignificant, when gauging my own intellect
by that of certain AI folks I have known. For example, I was at one
time a fairly close friend of a certain postdoc who exhibited an uncanny
ability to explain endless phenomena without recourse to the well-worn
phrase, "I don't know," which seems to characterize an exceedingly large
percentage of my responses to questions having to do with the howness of
my reasoning. Whether sitting at my elbow in the Squirrel Hill Cafe
slugging down beer and Schnapps or fixtured in a corner at parties, he
could usually be seen pressing close to his captive audience, eyebrows
contorting with emphasis as he clarified point after point in a ceaseless
stream of what sounded to me and others, or so I assumed by their wide-eyed
expressions of amazement, to be the voice of an Olympian mind looking down
through the clouds into the abyss of lackluster intellect where we squirmed
and lived our trifling days.
He had a temperament that could freeze a simple-minded fool such as me with
no more than a gesture that seemed to imply "How drab" when I offered a
tidbit of speech that deviated from the course of his conversation. I do
not mean to imply that he was an unfriendly person, for he wasn't; quite
the opposite really. He once invited me to his apartment for dinner and
to discuss an issue about metaphor that troubled us mutually. I arrived at
about 6:30 in the evening, and he immediately began to explain all that he
knew about metaphor, dissecting point after point, with something akin to
anger gently boiling below the surface of his voice, escaping
once like the spray from a geyser, which he quickly brought under
control again. The apartment itself was intensely stark, devoid of all but
the elemental furnishings that seemed to meet his functional requirements.
Somehow, midnight turned into the bottom of a bottle of Jamisons, and
dinner never came to pass. We spent the last half hour sitting silently,
staring at one another as though there was an ethereal presence between us
that we neither understood nor worried about. We were vastly different
yet so much the same, and the connections were at once the discrepancies:
I came from humble soil, he from a veritable garden; I worked at existence,
he at the fruit of knowledge; I had been the stoic, he the epicurian. As
theme met phorous the final time, I passed 50 dollars to him, and he left
me with memories.
Perhaps what makes me most insecure about artificial intelligence is my
general lack of natural intelligence, especially when I'm warm in winter and
away from the hawk with food on my table unlike what it was when I began and
grew. And perhaps it's most so this tendency in me to not capture the
fleeting obvious until it's too late to say it to who it is that needs to
hear it said. I sit inside walls of wires that bring in voices and pictures
that move and keep the time accounted for on the walls, but that leaves me
less than intelligent, when gauging myself against my memories and wondering
how they work and how long they will.
------------------------------
End of AIList Digest
********************