COMMENT ⊗   VALID 00009 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "memo.pub[let,jmc]" source
C00007 00003	.skip to column 1
C00009 00004	.bb Computers and Communications - B. O. Evans, IBM Corporation
C00015 00005	The Return of the Sunday Painter or The Computer in the Visual Arts
C00017 00006	Scientific Uses of Computers - by Sidney Fernbach
C00021 00007	Computers and Modelling - by Martin Shubik
C00024 00008	The consequences of computers for centralization and decentralization -
C00026 00009	Towards convivial computing - by Terry Winograd
C00029 ENDMK
C⊗;
.require "memo.pub[let,jmc]" source;

.bb Computers and Automation by Michael L. Dertouzos

	I agree with the picture presented in this paper except for
some secondary issues which, however, seem to me important in getting
an accurate picture:

.item←0;
	#. The shoe-making scenario shows too many distinct machines
to be economically feasible.  If the made-to-order shoes are to be
available while the customer waits, then there must be a great
excess of production capacity, because demand concentrates on certain
days and at certain hours.  Therefore, the machinery serving a customer
must be inexpensive.  This can be achieved, but it will require that
the same manipulator do the cutting as does the sewing or other
fastening.  It may even have to be the same machine that gets the
materials from the boxes in which they are stored.  Thus it will
have to be closer to the general-purpose robot than is shown in
the illustratons and text of the article.

It may be appropriate for the shoe making to take place in a
shop that can make other items, e.g. clothes as well.  Incidentally,
home computer terminals and keeping personal files in public
computer systems will come before automatic manufacture so that
your foot dimensions will be kept in your private files, and you
needn't carry a special shoe card.

	#. The heterarchical systems predicted seem too ⊗sociomorphic (if
one may use that word for a structure modeled on human society).
The part about each traffic control computer being connected precisely
to its four neighbors reminds one of the Illiac 4 disaster.

	#. The intuition that distributed computers can do more
than a single computer of equivalent total power seems disprovable.
On the other hand, it can probably be shown that some tasks cannot
be divided among a number of processors without arbitrarily large
losses of efficiency.

	#. The results of Julesz using random dot stereograms shows
that edges, etc. can be perceived after the images from the two
eyes are fused.

	#. There is a qualitative difference between servo-mechanisms
and problem solving programs.  The latter can decide to go in the
opposite direction from the goal if analysis shows that this is
required.

	#. The postulated arithmetic atrophy seems plausible, but
the article reads as if it has been actually observed.  Has it?

	#. Unless an innovation actually displaces labor, it shouldn't
be adopted and usually isn't.  With so many people working harder
than they wish and getting less pay than they wish, the saving
of human labor is still the most desirable goal for society as
a whole.  Of course, an individual whose skills are made obsolete
by an innovation may be adversely affected and may bargain for
a share of the benefits of the innovation to compensate.
.skip to column 1
.bb Noyce

	There is one dimension of possible improvement of integrated
circuit technology not mentioned in this paper.  This is the possibiliy
of a technology that will reduce the time necessary to make a small
number of integrated circuits, the cost of a small run, and the
capital required to make them.  The extreme would be a computer controlled
device using (say) electron beam technology that would permit the
production of a single integrated circuit in a few minutes according
to a computer prepared design.  Let us suppose that such a device were
within the financial resources of a laboratory.

	How far can one forsee going in this direction?
.bb Computers and Communications - B. O. Evans, IBM Corporation


	There are two considerations not present in this report
	It seems to me that two factors not considered in this paper
will have a strong effect on the future of computers and communications.
The first of these is within the direct scope of the paper - costs
of communication between computers and between users and computers.

	This is dependence of cost on holding time.  There are two
extreme cases:  The first is a user at a terminal: he may be a programmer,
an airline clerk, a student using computer aided instruction, or
a reader of a book from a computerized Library of Concgres.  His
average input rate is slow, about 1/3 character per second has
been measured, but he will need higher rate output, preferably in
bursts, averaging three to ten characters per second.  He will have
a holding time measured in tens of minutes.  The second use is for
sending messages.  Most messages are very short.  If they are sent
over the American dial-up network, the costs are dominated by the
minimum - three minutes or one minute depending on tariff considerations
apparently unrelated to the cost of providing the service.  The Wester
European dial-up network - at least for pay phones - treats this
customer better.  The minimum time of a call is as low as 3.7 seconds
and the price is linear with time.

	A critical factor in the growth of message traffic is the
relation between the price of a minimum telephone connection and
the price of a stamp.  Already this relation favors electronic
communication in most localities, and if the cost of a stamp
rises to 35 cents by 1985 as the Postal Service has predicted
(in some statements asking for subsidy), then electronic communication
will be favored for all classes of service.

	The second consideration is within the scope of the title
of this article though not within the domain the author has chosen
to cover.  Direct computer-to-computer communication between
different businesses has the largest potential for increasing the
use of both computers and electronic communication.  Many potential
applications are frustrated by the fact that putting the information
into computer readable form dominates the cost of processing it.
If the information is generated using a computer, then it must
never leave an electronic form.  This requires a standard "Business
Communication Language" that can express inquiries, orders, invoices,
etc.  The topic has received extensive treatment in the narrow
application of electronic funds transfer, but no start has been
made on standardizing other forms of company-to-company business
communication.  In an advanced system, a company would be able
to declare itself a supplier of a certain item by getting
the information entered in a standard way in an industry wide
computerized catalog.  Then the purchasing programs of buyers
of the item would notice it in the catalog and would automatically
check out its Dun and Bradstreet rating call its computer on
the telephone with inquiries and eventually orders.  The standardization
required here is on a much higher level than the current bit stream
standardization efforts.  I hope to treat this problem more
fully elsewehere.

The Return of the Sunday Painter or The Computer in the Visual Arts

by Nicholas Negroponte

	The author doesn't make much of the computer as an aid to
the artist who wants to use it to get and refine an effect that
he has in mind before he starts.  It isn't clear whether this is
because he discounts the possibility, or because he is pursuing
other interests in this paper.  In my opinion, any artist up to
the twentieth century would have used the computer to execute his
artistic conceptions more successfully.  The specific characteristics
of computers would influence the art only to the extent that
painting and etching differ.  The artistic ideas executed in the
two media are often similar.
The use of computers in art will become
important only when art - for its own reasons - again concentrates
on realizing specific conceptions of what is beautiful.

	Even if the paper is about art, it still would be better
if the author would say more clearly what he is asserting.
Scientific Uses of Computers - by Sidney Fernbach

	The view that "It is the scientist,  however, who still seems
to set  the pace for development in  most aspects of computing" seems
justifiable only  if  one  leaves out  much  of what  is  called  new
developments in  computing.  Also Fernbach's  Livermore oriented view
of  science seems narrow, e.g.  it leaves out  computer science.  For
example, many hold that the most significant  recent developments are
related to  data base management and  communication oriented systems.
However, the  rest of  this  discussion will  be oriented  to  number
crunching.

	The   article  correctly   points   out  that   much   larger
computations will require parallel processing, the largest project in
which was  Illiac 4  costing more  than $32  million and  essentially
pre-empting the national effort in parallel processing for ten years.
The  future of parallel  processing may depend  on understanding that
Illiac 4 was a blunder and being very sure not  to make the following
mistakes again:

	1. The basic idea of a multi-processor  machine with a single
control  was wrong.   As was predicted,  the machine  has been almost
impossible to program in such a  way as to bring its full  arithmetic
capacity to  bear on a single  problem.  It is  necessary that future
parallel machines have a control for each processor.

	2. Arranging the machines in a fixed  square array turned out
too Laplace-omorphic - to coin a term.
  
	3. Parallel processing needs to be associated with very large
directly accessible  high speed memories.   Any future project should
be capable of addressing and  accessing all the memory technology  is
likely to make available in the planned life cycle of the machine and
its compatible successors.

	4. Projects  should  not be  based on  academic logrolling  -
where  engineers bargain to accept  each other's innovations.   It is
risky to innovate in more than one area at a time.
Computers and Modelling - by Martin Shubik


1. It is not at all clear that this is a raw-data rich society.
The basic conomic flow information - who bought what from whom
and when was it delivered - is still unavailable.  The technical
coefficients of the Leontieff matrix are most inaccurately known
and so re the lead-times of the various processes.

2. We are not on the brink of developing intelligent machines.
Skepticism is easy enough to program, but wisdom (as opposed to
intelligence) seems to be one of the refuges of the intellectual
scoundrel.


	I am left with an uneasy feeling that the article has
missed the main issues in the use of computers in modelling
aspects of society.  These are:

	1. Is the only use of computers in social modelling
simulation?

	2. How can incompletely known data or causal relations
be taken into account?

	3. 

I don't think the duel models have any of the right properties.
I also think the travelling salesman problem is for mathematicians.

Is fairness in accounting really a matter of balancing the
interests of the parties?  Is there not an issue of truth
involved?  Suppose one party is a swindler with a great stake
in staying out of jail.  Should the accountant include his
interests in presenting a "fair" picture?

The consequences of computers for centralization and decentralization -
Herbert Simon

The first part of the discussion assumes that the people at the top
can control the design of the organization.  This may not be true
under the same conditions under which they cannot control the
operation of a decentralized organization.  In other words, the
organization is assumed to have goals rather than to be a community.

How broad-band an information link is, is not always the decisive
consideration.  We can do a lot more with telephone bandwith than
we have done provided information can be automatically collected,
filtered and stored.  Computers often beat human systems merely
in staying on duty, being reliable, and not developing goals of their
own such as remaining necessary.

The example of whether the computerized doctor's adviser makes the
decisions or not is excellent.

A much more elaborate electronic democracy is possible including
delegation, continuous election, etc. and logical filtering of
proposals.  Getting the necessary information down is the key to
local influence on decisions.

Computers and democracy

	I agree that the expected effects of computers on centralization
and decentralization will be mixed.
Towards convivial computing - by Terry Winograd

	The paper proposes that computers be made more "convivial"
by making them use natural language and eventually speech.
The examples given don't leave the reader with much confidence that
a computer can be programmed to evaluate a letter from a doctor
that a person's back problem is cured and draw the conclusion
that the person is more creditworthy than his employment record
would indicate.  Such considerations are best left to bank loan
officers (not clerks) who in fact evaluate loan applications.

	On the other hand, store clerks must make credit decisions
with much less discretion and are not normally allowed to take such
factors into account with or without the aid of computers.
Let's consider what are the concrete possibilities for making
computer aided credit decisions more flexible.

	The most clearly implementable method is to allow the customer
to access the credit determination system directly - say through
a typewriter terminal in the store if the customer doesn't have
one at home.  Most likely the algorithm for determining the credit
limit is distressingly simple-minded.  The extreme case is that
a previous human decision is simply stored in the computer, but
it is possible that a batch updated credit limit based on previous
payment record is stored.  It seems unlikely that an ad hoc computation
of payment record is made at the time credit is requested, because
few such systems keep the necessary data on-line.

	A good system would be able to make a large number of
"obvious" decisions and document them well enough so that most
disappointed credit seekers would know that policy changes would
be needed to give them credit, and the "maybe" cases would be
referred in an on-line way to a human with sufficient authority.