The Risks Digest

The RISKS Digest

Forum on Risks to the Public in Computers and Related Systems

ACM Committee on Computers and Public Policy, Peter G. Neumann, moderator

Volume 8 Issue 84

Wednesday 21 June 1989

Contents

o The risks of global editing
Martyn Thomas / Richard Tobin / Nick Radcliffe
o Re: I am not a number -- already in the US
Tom Comeau
o Re: I am not a number -- more in Canada
Vince Manis)
o Re: Computer electrocutes chess player ...
W. Scott Meeks
Brendan McKay
o Gigatext Translation Services Inc. scandal
Bhota San
o Info on RISKS (comp.risks)

The risks of global editing

Martyn Thomas <mct@praxis.UUCP>
Wed, 21 Jun 89 14:23:59 BST
From today's Guardian (UK national newspaper), Wednesday 21st June

     "Correction

     Because of the startling simplicity of computer programme logic, the
     Labour Party's successful candidate in the London South West seat of
     the European elections was printed on Monday as Ms A J Turnoutack,
     rather than Ms A J Pollack.  On Tuesday, her name was again spelt
     incorrectly.

     The computer programme which run [sic] to massage the results into
     printed form changed each occurrence of the work Poll, into Turnout."

(Spotted by Simon Seely)
Martyn Thomas, Praxis plc, 20 Manvers Street, Bath BA1 1PX UK.

   [Also reported by Richard Tobin <richard@aiai.edinburgh.ac.uk> (who added
   that the Guardian has been noted for its misprints since long before 
   computers), and by Nick Radcliffe <njr@etive.edinburgh.ac.uk>.
   I suppose Turnoutyanna would have seen the good even in that one.  But
   in a fault-tolerant illumination system there would have been an unfortunate
   ambiguity between Poll the Lights and Turnout the Lights.  PGN]


Re: I am not a number - already in the US

"Chairman, Von Neumann Catastrophe Relief Fund" <TCOMEAU@STSYSC.UUCP>
Wed, 21 Jun 89 10:29:44 EDT
In RISKS DIGEST 8.83, Walter Roberson discussed a possible Canadian move to ten
digit postal codes.  While most people in the US don't use more than the first
five digits, the US Postal Service (USPS) already uses a 9 digit ZIP+FOUR code.

Like Canada, the census people give out info based on ZIP code, and it is a
crime not to give accurate information.  So I trotted down to the local post
office, and sure enough, my address is a unique nine digit code.  There are
only four units in my building, and only nine people, so a person with census
info based on nine digits and just a little more (like, I'm the only married
male in my building) can find out anything the census bureau know about me.

We just got a note last week (delivered by the USPS to OCCUPANT) that the 1990
census is coming, with assurances that our responses to the census would be
kept strictly confidential.  In light of this RISKS discussion, I'll have to
take that claim up with both the Census Bureau and my elected representatives.

tc> Tom Comeau, Sr Systems Manager, Space Telescope Science Institute, 3700 San
Martin Drive, Baltimore, MD 21218                     (301) 338-4749 (AT&T)
Disclaimer:  My opinions are my own, not even my wife agrees with me.


I am not a number... (unique postal codes)

Vince Manis <manis@grads.cs.ubc.ca>
Wed, 21 Jun 89 08:19:12 PDT
Walter Roberson describes the proposal to create a 10-character postal
code, and mentions that it seems to generate a too-high degree of
discrimination. (The current postal code, of which V6T 1W5 is an
example, does an excellent job: the first character corresponds to the
province, and the first three to the Forward Sortation Area (FSA), or
sorting station; the full code corresponds to less than a single
carrier's route. Further, the mixture of letters and digits
substantially lowers the error rate, because transposition errors, the
most common among fast typists, result in invalid codes.)

There is, however, a RISK which is entirely separate from the assault on
privacy which a 10-character code would inevitably produce. It is necessary to
realise that Canada Post's service is legendarily atrocious (though not perhaps
as bad as that of the Italian Post Office, which once shredded 40 tonnes of
mail which had taken so long to sort that it wasn't worth delivering). The
introduction of postal codes in 1974 was supposed to speed service. In fact,
service has deteriorated. (It is quite common for a first-class letter to take
a week to be delivered to another address in the same sortation area.)

Labour relations are also atrocious at Canada Post. There are at present
several tens of thousands of in-process grievances; the leadership of the main
postal union is highly radicalised; and management has made it very clear that
it is not interested in improving the situation.

When the original code was introduced, the postal workers opposed it, fearing a
loss of jobs. They stamped `Boycott the code' on letters, and protested
strenuously. Management didn't do anything in particular to smooth the
transition.

I can see the introduction of a longer code as an excuse to create further
confrontation, thus excusing the awful service. So long as long delivery delays
can be blamed on `lazy workers', whereas the Corporation is using the latest
technology to expedite things, nobody really has to do anything specific to fix
the serious management/labour problems at Canada Post.

Ain't technology wonderful? 

Vincent Manis, The Invisible City of Kitezh, Department of Computer Science,
University of British Columbia, Vancouver, BC, Canada V6T 1W5 (604) 228-2394


Re: Computer electrocutes chess player who beat it!

W. Scott Meeks <wscott@flab.flab.fujitsu.junet>
Wed, 21 Jun 89 11:12:59 JST
It occurred to me that there may be a very plausible explanation for this
story--there was indeed a murder comitted, but not by the computer.  It seems
to me that the programmer, research scientist, whoever, who programmed this
super computer may have been embarrased (perhaps fatally in political terms) by
the fact that his system was being beaten my a mere human.  To get revenge, he
reprograms the computer, possibly modifying the chess board as well to make the
fatal short circuit, so that it electrocutes the chess master the next time the
computer loses.  This may even turn out as an added bonus for the researcher --
the computer may not be able to beat a chess master, but it is intelligent
enough to stand trial.  Once again, the computer gets all the blame.

W. Scott Meeks, Bellcore, Morristown, NJ 07960-1910


Re: Computer electrocutes chess player who beat it! (RISKS 8.75)

Brendan McKay <bdm@anucsd.oz.au>
21 Jun 89 10:18:35 EST (Wed)
[Dave Horsfall doubts that 12 volts can cause substantial electric shock.]

On the contrary.  About a year ago I drove my car into a creek and got
stuck.   I tried to disconnect the battery but found it quite impossible to
hold onto the positive terminal without insulation; the shock was sufficient
to make my arm flail around uncontrollably.  I was standing in about one meter
of water, and the car was turned off.  (Of course, I should have disconnected
the negative terminal instead, but nobody's perfect.)

PS.  Haven't people been electrocuted in swimming pools by remarkably small
voltages?  I defer to someone who knows the details.

Brendan McKay.  bdm@anucsd.oz.au


Gigatext Translation Services Inc. scandal

Bhota San <VOGTWARR@UREGINA1.BITNET>
Wed, 21 June 1989 1:35:43 CST
For some of you, this will be the third time you will have been exposed to this
article. I apologize for this. The first time was intentional, but only the
subscribers on Bitnet got it. The second time was because of an anomaly in the
file transfer mechanism, unintentional. This time it is so that the non-Bitnet
subscribers can read it and so that I can finish the story of the whole ludi-
crous affair. Also, one of the respondents sent me the complete references for
the books I consulted to form an opinion of this misbegotten venture. If you
wish you can skip the repeated section of this posting and go straight to his
upgraded bibliography and the story's denouement by typing FIND BIBLIOGRAPHY or
some other search command to move your editor to that point in my account. But
that is several lines ahead. The repeated introductory section begins here:

  /
PRECIS                                                 Tues, 13 June 1989

The provincial government of Sask., Canada is now embroiled in a scandal over
its involvement with an AI language translation company owned by Guy Montpetit
of Montreal, a shady dealer currently defending himself against a $39-million
lawsuit for not paying back loans. Deputy Premier Eric Berntson had discussed
with Montpetit the possibility of its investing $125 million in a computer chip
foundry, the details of which would have been overseen by Montpetit.
         But the government has already invested $4 million in Montpetit's Gig-
aText, a company established to translate by computer Saskatchewan's statutes
into French. A couple of years ago, a local Francophone created a sensation
when he took to the Canadian Supreme Court his case that his constitutional
rights were being infringed when he had to go to court on a traffic ticket that
was made out to him in English, not French. The Court decided in his favour.
After trying to quash the decision by a decree of its own, the government swore
it would save millions of dollars by getting someone to create AI software to
translate its laws, rather than by hiring bilingual humans to translate them
"the old way". The government accepted Montpetit's offer to do the job and
payed him the $4 million to rent a sizable office complex, hire administrative
personnel and a few programmers supposedly bilingual and conversant in Lisp and
Prolog. It is alleged that the government also bought him a luxury home, payed
his plane fare to "conventions", and let him use a Mercedes to get to his Re-
gina appointments. It took a long time for expert criticism of the presumptions
of this venture to reach the public, including the NDP opposition which is now
raking the incumbent PC's over the coals for it. But Berntson confidently pre-
dicts the computerized translator will be unveiled in full working condition
on June 17.
         That's four days from now. It will be interesting to see how it turns
out. From everything I have read, Artificial Intelligence language translation
is still a long, long way from being practicable, and indeed may be impossible.
The world will stand up and take notice of what has been accomplished in Regina
Sask. on June 17, but I wonder if it will be for the reasons Berntson intended.
         I based my assessment of this case on readings from a book called "Tu-
ring's Man" by J. David Bolter. (See references below.) On a different subject
is a book called "The Fifth Generation Fallacy", by J. Marshall Unger. The au-
thor, a professor of East Asian languages at the University of Hawaii, Honolulu
states that the REAL reasons the Japanese are in such a rush to invent this new
hard- & software has little to do with fifth generation machines that imitate
human neural networks & thought processes to speak to the user in his own ton-
gue [sic]. It is their desperation to find a means to speedily process, store &
distinguish the thirty odd thousand characters of their Kanji script. These
characters are difficult to represent digitally or on a screen: It takes a 24 X
24 grid (576 bits) to store a good image of them. To keypunch, input them will
be yet harder: Business memos are still written out in hand. And it will seri-
ously test the software running at extremely high speed to tell them apart, a-
side from the grammatical difficulties involved. In short, the professor be-
lieves the Kanji character system puts an immovable limit on what any computer
technology can do for their quest for data-handling speed and efficiency. The
question is then: Do the dubious advantages of increased productivity obtained
by high-tech mean more to the Japanese than love of their culture and the ele-
gance of their 1000-year-old Kanji script?


BIBLIOGRAPHY:

Bolter, J. David, 1951-   <UBOLT@TUCC.BITNET>
  Turing's man : western culture in the computer age / by J. David Bolter. -
  Chapel Hill : University of North Carolina Press, c1984.
  xii, 264 p. : ill. ; 24 cm.

Unger, J. Marshall.
  The fifth generation fallacy : why Japan is betting its future on artificial
  intelligence / J. Marshall Unger. - New York : Oxford University Press, 1987.
  x, 230 p. : ill. ; 22 cm.


DENOUEMENT                                                Tues, 20 June 1989

         Well, a week has passed and now I can report what happened on June 17,
the day Sask. deputy premier Eric Berntson promised to show us the computerized
French-English translator, working as planned. For it is certain he had some-
thing planned, and it came as a surprise to everyone.
         Mr. Berntson did not have a scrap of translated laws to show us that
day. But he did succeed in showing us the NDP opposition was guilty of having
wasted more of the tax-payers' money when it was in power over something like
this than the incumbent PC's have over GigaText. As he said, in 1981 "the New
Democratic Party...invested $5 million -- the equivalent of $8 million today --
in an outfit called Nabu...from central Canada [and] not one job, not one ounce
of technology moved to Saskatchewan. The whole investment was written off and
these technological wizards over there blew $5 million!"
         Nabu developed informational or educational software programs for home
entertainment. Embarrassed by this telling repartee, the NDP critics fell into
sullen silence.  (The Leader-Post, June 17 p A4)
         Yet the NDP's venture, an investment in what we would probably call an
"expert system", from a purely technical perspective had a greater chance of
succeeding than did the PC's venture into the field of AI language translation,
a goal that continues to elude the most accomplished researchers in this area.
(See "Turing's Man" in the above bibliography.) Herbert Clute (ABC Translation)
agrees with ex-director of legislation translation in neighbouring Manitoba,
Greg Yost's opinion that machine translation of its statutes was "too primitive
to be of any use." He said weather reports are the only material that can be
translated flawlessly: "If the vocabulary is straightforward and limited, ma-
chine translation would be a great saving in time, but we all know laws aren't
straightforward."
         Manitoba gave the bulk of its translation contract to "Traduction Uni-
verselle", a Montreal-based company in contact with plenty of people who were
competent to translate the language in specific areas of legislation. In four
years the bulk of the translation, 4000 pages, was translated by the Dec. 1988
deadline for public laws and regulations. (The Leader-Post, June 19, 1989 p A4)
         Saskatchewan, in contrast, since July 1988 has funnelled $5.25 million
into the GigaText Translation system, so far without success. PC politicians
have had to face accusations of wasting the money on perks for Montpetit and to
take off his hands obsolete computers from his own company. The 20 GigaMos com-
puters that GigaText bought for its Regina research centre cost the government
$152 000 each. Their cost climbed from $1,5 million to $2,9 million as they
passed through the hands of firms owned either by Montpetit or his close asso-
ciates before being purchased by GigaText. Montpetit's U.S. company "Systems"
manufactured them. (The Leader-Post, June 16, 1989 p A4)
         Incidentally, Berntson acknowledged meeting am associate of Montpetit
named Dr. Alex Voshchenkov. He considers Voshchenkov to be one of the world's
leading high-tech scientists and declared, "if I can find a way that is reason-
able to get Dr. Voshchenkov and his technology into Saskatchewan I would like
to do it." I am unfamiliar with Dr. Voshchenkov and his achievements, but with
that June 17 deadline passed, the provincial government will need all the help
it can get! (The Leader-Post, June 17, 1989 p A4)
         The RCMP has been investigating Montpetit to determine how GigaText
spent SEDCO's first $4 million. There are allegations Montpetit spent it on o-
verpriced computers, personal debts, payments to those who helped arrange the
government financing, and jet travel, like a weekend flight taken by Montpetit
and a female companion to San Francisco at a cost to GigaText of $15 000.
         In exchange for its $4 million, the government got 25% ownership in
the company. In exchange for his unproven technology, Montpetit and his partner
were given 75% control and sole responsability for signing cheques. This con-
tinued, even after he was named in a $39 million lawsuit last October, until he
was eased out in March. A new $1,25 million SEDCO loan was provided to keep the
company alive, but the June 17 deadline was set for it to prove it could do its
job. That day has come and gone. You'd think that would be the end of it, but
the project has been given a stay of execution. Berntson said that because June
17 was a Saturday, the deadline is being extended to some time this week. If at
the last moment, it miraculously succeeds, I will faithfully report the details
to the list. But if you don't hear from me again, you can assume it will have
been the big flop everyone expects. After all, there is an old Latin maxim that
fits very well here:    "DE MORTUIS NIHIL NISI BONUM."

Please report problems with the web pages to the maintainer

Top