Copy Link
Add to Bookmark
Report
AIList Digest Volume 2 Issue 077
AIList Digest Friday, 22 Jun 1984 Volume 2 : Issue 77
Today's Topics:
AI Tools - Q'NIAL,
Cognition - Mathematical Methods & Commonsense Reasoning,
Books - Softwar, A New Weapon to Deal with the Soviets
----------------------------------------------------------------------
Date: 19 Jun 84 14:59:27-PDT (Tue)
From: hplabs!hao!seismo!cmcl2!lanl-a!unm-cvax!janney @ Ucb-Vax.arpa
Subject: Re: Q'NIAL
Article-I.D.: unm-cvax.962
The April 1984 issue of Computer Design has an article on Nial
(Nested Interactive Array Language).
------------------------------
Date: 18 Jun 84 15:10:07-PDT (Mon)
From: ihnp4!houxm!mhuxl!ulysses!unc!mcnc!ncsu!jcz @ Ucb-Vax.arpa
Subject: Re: Mathematical Methods
Article-I.D.: ncsu.2622
It is not surprising that mathematicians cannot
remember what they do when they first
construct proofs, especially 'difficult' proofs.
Difficult proofs probably take quite a bit of processing power,
with none left over for observing and recording what was done.
In order to get a record of what exactly occurs ( a 'protocol' )
when a proof is being constructed, we would have to interrupt the
subject and get him to tell us what he is doing - interferring with the precise
things we want to measure!
There is much the same problem with studying how programmers
write programs. We can approach a recording by saving every scrap of paper
and recording every keystroke, but that is not such a great clue
to mental processes.
It would be nice if some mathematician would save EVERY single scrap of
paper ( timestamped, please! ) involved in a proof, from start to finish.
Maybe we would find some insight in that. . .
John Carl Zeigler
North Carolina State University
------------------------------
Date: Wed, 20 Jun 84 20:39:01 edt
From: Roger L. Hale <rlh@mit-eddie>
Subject: Re: Commonsense Reasoning?
From: Roger L. Hale <rlh@mit-eddie>
Subject: Re: Commonsense Reasoning?
I get 4 quite a different way:
If 3 (2) were half of 5 (4), what would a third of 10 (9) be? 4 (3).
This way twice 5 (4) is 9 (8), [rather than twice 6 (5) is 12 (10)
the way you describe.]
The transformation I have in mind is "say 3 and mean 2", which is simply
difference-of-1. The numbers I *mean* are in the stated relations
(half, a third, twice) but they are renamed by a distorting filter,
a homomorphism. "If arithmetic were shifted right one, what would half of 5,
a third of 10, twice 5 be? (Answer: 3, 4 and 9.)" Partly it is a different
choice of who to believe, the numbers or the relations; but I find this
form most compelling due to the components being so fundamental.
The extended proportion [for your method in parallel form] would be
3 : 5/2 :: 4 : 10/3 :: 12 : 2*5,
the 12 (v. 9) serving to show that our two methods differ concretely.
I think that the critical point for AI is that we make sense of
a nonsense problem by postulating an unmentioned linear
transformation since only a linear transformation permits a unique
solution. [...] -- KIL
In the first place, any critically constrained transformation has discrete
(locally unique) solutions, barring singularities; and it is false that
they are only unique for linear transformations: it takes a fairly special
domain, like the complex analytic, to make it true. In the second place,
what confidence should one gain in a theory on fixing a free parameter
against one datum? Surely one should aim to constrain the theory as well
as the parameter, and you have used up all your constraints. Where would
we be if twice 5 were neither 12 nor 9? ?8-[ Back to square one.
Yours in inquiry,
Roger Hale
rlh%mit-eddie@mit-mc
------------------------------
Date: 19 Jun 84 14:00:58-PDT (Tue)
From: hplabs!tektronix!orca!shark!brianp @ Ucb-Vax.arpa
Subject: Re: Commonsense Reasoning?
Article-I.D.: shark.836
About "if 3 is half of 5, what is a third of 10?"
It is interesting to note the assumptions that might be made here.
One could assume that all numbers retain their good-old standard meaning,
except 3, when compared to 5. Then the chain of relationships
(3:5/2, 6:5, 12:10, 4:10/3) can be made. What I first thought was
"so what's a '10'? " I.e, let's toss out all the definitions of the
numbers along with 3. 'Half' could be redefined, but that says
nothing about what to do with 'third'. One could redefine 'is',
in effect, making it mean the ':' relation of the previous article.
Anybody have hypotheses on which assumptions or definitions one would
tend to drop first, when solving a puzzle of this sort?
Brian Peterson
...!ucbvax!tektronix!shark!brianp
------------------------------
Date: 19 Jun 84 18:34:05-PDT (Tue)
From: hplabs!tektronix!orca!tekecs!davep @ Ucb-Vax.arpa
Subject: Re: Commonsense Reasoning?
Article-I.D.: tekecs.3861
> From: brianp@shark.UUCP (Brian Peterson)
>
> It is interesting to note the assumptions that might be made here.
> One could assume that all numbers retain their good-old standard meaning,
> except 3, when compared to 5. Then the chain of relationships
> (3:5/2, 6:5, 12:10, 4:10/3) can be made.
If one redefines "3, when compared to 5", shouldn't the 3 be redefined in all
instances of the "chain of relationships"? If so, one could conclude that
one-"third" of 10 is 24/5 via 3:5/2, 6:5, 12:10, 12/(5/2):10/3, 24/5:10/3.
David Patterson
Tektronix, Inc.
Wilsonville Industrial Park
P.O. Box 1000
Wilsonville, Oregon 97070
(503) 685-2568
{ucb or dec}vax!tektronix!tekecs!davep uucp address
davep@tektronix csnet address
davep.tektronix@rand-relay arpa address
------------------------------
Date: Wed 20 Jun 84 18:37:45-PDT
From: Jean-Luc Bonnetain <BONNETAIN@SU-SCORE.ARPA>
Subject: softwar, a new weapon to deal with the Soviets ?
[Forwarded from the Stnaford bboard by Laws@SRI-AI.]
This is my translation of an article published in a French news magazine,
"Le Point"; i have done my best to translate it, but i am sure there are
some inadequacies. I just hope they don't occur in important places.
I am just wondering if any one has heard about that, and if this is real,
pure computer fiction or so well known that it's not worth flaming about.
"Between the atomic bomb and conventional weapons, there was nothing in the
American warfare equipment against the USSR. Now the time has come for
"soft bombs", to launch a destructive war without any bloodshed. This is the
topic of "Softwar", a forthcoming book written by a French computer scientist
working in New York. The idea: as simple as it is machiavelic. In the programs
that Soviet people get from Western countries are placed what amounts to "time
bombs": devices that can be triggered from afar to hamper the functioning of
Russian computers and paralyze the economy. With "Softwar", nuclear blackmail
becomes obsolete. Le Point asked the author, Thierry Breton, how his relations
with highly skilled American engineers has convinced him of the existence of
the new type of weapon.
LePoint:
is "Softwar" just an computer thriller, or do "soft bombs" really exist ?
ThierryBreton:
I never used any, but they have been used for a few years already in our
trade. Some countries from Africa or South America, who are customers of
big American software companies, have booby-trapped programs running in their
administrations. The aim of the providers of the software is to be protected
against customers who won't pay. These soft bombs are set in vital areas,
like payroll routines, which are then paralyzed. The customer has to call the
company, and won't get any help until debts are cleared. In this case people
talk about technical problems in the computer, but obviously never say that
the program contained a bomb.
Since now, these techniques had never been used for aggressive purposes. But
there is absolutely no technical difficulty in doing that, so we are led to
believe that this new weapon could be used through non strategic networks
giving access to databases. For example, the Stockex network, which gives
information on stock exchange values, or the WMO network, about worldwide
meteorological information.
LePoint:
Has softwar begun yet ?
ThierryBreton:
For me, there is no doubt about that. The Soviets use 80% of the American
databases. It is this dependency on communication between computer which is
new, and which allows to enter a territory. Until now, the "bombs" had to be
triggered on the spot by someone inside the place. The bombs were there, but
could not be triggered remotely. Today, thanks to data transfer, they can be
reached from thousands of kilometers. In the book, I imagine that one bomb is
controlled, through Stockex, by the rate of exchange for a particular company
determined in the software, and the Pentagon, as long as it does not want to
detonate the "bomb", avoid the critical value by buying or selling actions.
LePoint:
You give some names of American organisms working for the Pentagon whose work
is to set bombs in the programs, and to activate them. Is this real ?
ThierryBreton:
The names quoted have been slightly modified from the real ones. I took my data
from a group founded in 1982 by the American Army, called NSI (National
Software Institute). This institute works on all programs which have military
applications. In 1983, the Army has spent 500 million dollars to debug its
programs. Written in different languages, they have now been unified by the ADA
language. This is the official objective of NSI. But for these military
computer scientists, there is not much difference between finding unvoluntary
errors and adding voluntary ones...
LePoint:
What is the Trojan horse used to send those soft bombs to the USSR ?
ThierryBreton:
The USSR has a lag of about 10 to 15 years in computer science, which is the
equivalent of 2 or 3 new generations of computers. This lag in hardware
causes an even more important lag, in artificial intelligence, which is the
type of software running on the machines Soviet people have to buy from Western
countries. They are very eager to get those programs, and some estimate that
60% of the software running there comes from the USA. The most important source
is India, which has very good computer scientists. Overnight, IBM has been
kicked out, to be replaced by Soviet Elorg computers ES10-20 and ES 10-60,
which are copied from IBM. The Indians buy software from Western countries,
port it to Elorgs, and then this software goes to the USSR.
LePoint:
Can a trap be invisible, like a buried mole ?
ThierryBreton:
Today, people know how to make bombs completely invisible. The first generation
was fixed bombs, lines of code never activated unless a special signal was
sent. Then the Polaris-type traps: like for the rockets, the programs contain
baits to fool the enemy, multiple traps, only one of which is active. Then the
stochastic bomb, the most dangerous one, which moves in the program each time
it is loaded. These bombs are all the more discreet that they can be stopped
from a distance, failures then disappearing in an unexplicable way.
LePoint:
Have there been cases in USSR of problems that could be explained by a soft
bomb ?
ThierryBreton:
Some unexplained cases, yes. In November 1982, the unit for international
phone calls has been down for 48 hours. Officially, the Soviets said it was
a failure of the main computer. We still have to know what caused it. Every
day in the Soviet papers one can read that such and such factory had to stop
its production because of a shortage of some items. When the Gosplan computers
break down, there are direct consequences on the production and functioning of
factories.
LePoint:
By talking about softwar, aren't you helping the Soviets ?
ThierryBreton:
No. For 30 years, we have seen obvious attempts from the Soviets to destabilize
Western countries by infiltrating trade unions, pacifist movements. The Eastern
block can remotely cause strikes. But since now, there was now way to retaliate
by doing precise desorganizing actions. In the context of the ideological war,
softwar gives another way to strike back.
The book also shows that the Soviets have no choice. They know that by buying
or getting by other means this software, they are taking a big risk. But if
they stop getting this software, the time it will take them to develop it by
themselves will increase the gap. This is a fact. So soft bombs, like atomic
bombs, can be a means of deterrence. For political people who are just
dicovering this new strategy, the book is that of a new generation showing to
the old one that what was a tool has become a weapon."
[This reminds me of an anecdote I heard Captain (now Cmdr) Grace Hopper tell.
It seems some company began to pass off a Navy-developed COBOL compiler
verifier as their own, removing the print statement that gave credit to
the Navy. When the Navy came out with an improved version, the company
had the gall to ask for a copy. Her development group complied, but
embedded concealed checks in the code so that it would fail to work if
the credit printout were ever altered. -- KIL]
------------------------------
Date: Wed 20 Jun 84 20:07:35-PDT
From: Richard Treitel <TREITEL@SUMEX-AIM.ARPA>
Subject: softwar @=
[Forwarded from the Stanford bboard by Laws@SRI-AI.]
The article Jean-Luc (or whoever) translates sounds like a typical piece of
National Enquirer-style "reporting", namely it describes something that is
*just* feasible theoretically but against which countermeasures exist, and
which has wider ramifications than are mentioned. I'm sure the Russians are
too paranoid to allow network access to important computers in such a way as to
trigger these "bombs".
But: it is widely rumoured that IBM puts time-delayed self-destruct operations
into some of its programs so as to force you to buy the new release when it
comes out (and heaven help you if it's late?). And in John Brunner's book
"The Shockwave Rider", one of America's defence systems is a program that would
bring down the entire national network, thus making it impossible for an
invader to control the country.
I love science fiction discussions, but I love them even more when they're not
on BBoard.
- Richard
[Another SF analogy: there is a story about the consequences of developing
some type of "ray" or nondirectional energy field capable of igniting
all unstable compounds within a large radius, notably ammunition, propellants,
and fuels. This didn't stop the outbreak of global war, but did reduce it
to the stone age.
All that has nothing to do with AI, of course, except that computers may
yet be the only intelligent beings on the planet. -- KIL]
------------------------------
End of AIList Digest
********************
