COMMENT ⊗   VALID 00003 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	\input kermac.tex
C00005 00003	\arabicnumbering{1}
C00028 ENDMK
C⊗;
\input kermac.tex
\input papmac
% \input macmac
% \input codmac
% \input indmac.tex[tex,mmd]

%chpar12=1100				% Use magnify, not this !!!
\magnify{1100}				% Now this is ok.

\font V=timesroman at 18truebp
\def\Tit{\:V}
\def\<#1>{{\:m <#1>}\¬}
\def\lv{\:B}
\def\te{\:a}
\def\ni{\:b}
\def\ei{\:m}
\def\se{\:d}
\def\si{\:c}
\def\s8{\:p}
\def\it{\:?}

	% Change formats:
\def\sectionbeginb#1#2{\addq1{\varsec}
	\sectiondeclare{\tit\hangindent4.5em\sectionnum.\ \ #1\¬
		      }{\sectionnum\hquad #2
		      }{\sectionnum\hquad #2} }
\def\subsectionbeginb#1#2{\addq1{\varssec}
	\subsectiondeclare{\tenpoint\bf\hangindent4.5em\subsecnum\hquad\rm #1
			 }{\subsecnum\hquad #2}}

\setmargin{1truein}{1.25truein}{.9375truein}{.9375truein}
	% Modified to moveright the even pages only with hshft set to .45
% \def\shftthepage#1{
%	\ddvnmsel{\moveright\hshft truein\thevbox{#1}}{\thevbox{#1}} }


	% To produce special TEX printing and imitations
\def\TEX{\hbox{\:aT\hskip-.1667em\lower.424ex\hbox{E}\hskip-.125em X}}
\def\DEK{\hbox{\:aD\hskip-.1667em\lower.424ex\hbox{E}\hskip-.125em K}}
\def\MAX{\hbox{\:aM\hskip-.2000em\lower.424ex\hbox{A}\hskip-.200em X}}
\def\ALS{\hbox{\:a \lower.424ex\hbox{A}\hskip-.1667em\hbox
                                    {L}\hskip-.1667em\raise.424ex\hbox{S}}}

	% Initializations
\def\projname{Short WAITS}
\def\date{January 16, 1981}
	% Set parameters to taste:
\parskip .1in
\parindent 0pt
% \def\hshft{.45}
	% To change spacing on contents page
\def\hdltopskip{-0.05in}
\def\hdlbtmskip{0.10vu plus 2pt}
	% To change table of contents to eightpoint
\let\tofctypeface=\ninepoint

\secsubsecstyle

% \chcode'20←13
% \def\⊂#1⊃{\indexk{#1}}
% mtex.dmp[1,3].
\arabicnumbering{1}
\normalpage

\ctrline{\lv On the State of Artificial Intelligence in 1984}
\ctrline {\lv an Early Prediction}
\ctrline{\lv by}
\ctrline{\lv Arthur L. Samuel}
\te
\par
We have been asked to think back to 1956 and recall what we then thought
would be the state of computers, computing and in particular of
Artificial Intelligence by the year 1981.  Thinking back is always a
hazardous undertaking, our thoughts are so conditioned by our knowledge of
what has actually happened that it is extremely difficult to remember
exactly what we did think at some earlier time. \par

The task of reporting my early predictions happens to relatively easy for
me because I actually put my predictions down on paper in two different
papers that were published a few years after that famous Dartmouth
meeting.  These published predictions were for the year 1984 and not for
1981 but it is easier to look forward for three years than to look
backward by twenty five, so I will report on what I actually wrote and not
on what I now think that I would have written had my target date been
1981. Having been so foolish as to commit my predictions to writing, I
must, perforce, reveal them to you.

I have often observed that people usually over-estimate what can be done
in a short period of time and under-estimate what will be done over a
longer period of time. There is a simple explanation for this effect.
People usually think of progress in linear terms while progress is apt to
be exponential.  In any case, I do remember making a deliberate effort to
think exponentially when I was writing this early paper.  I also wanted to
make my predictions as advanced as possible because one of the reasons for
writing was to make people sit up and notice this new factor in society,
the computer.  Remember, I was working for IBM in those days.

I am now afraid that I overdid this exponential thinking, although, of course,
we still have three years to go. Certainly, the number of people working
in the field of Artifical Intelligence seems to have gone up along some
exponential curve but I will leave it to you, individually, to decide if
the rate of actual progress has, in fact, kept pace with the effort
expended.

It is too bad that I was not asked to predict for the year 2000 because it
is likely that I would have made about the same predictions and many
of my still un-met predictions might well come true by this later date.

But to get to my predictions. And remember, I am quoting from my
published statements and not trying to relate what I now think that
I then thought, 25 years ago.

Firstly, as to the size, speed, memory capacity and cost, I predicted that
computers would get very much smaller in size, that they would be roughly
1000 times as fast as they then were and that their memories would be very
much larger---large enough to store in one data bank all of man's recorded
information.  I was a bit indefinite as to the reduction in cost but I did
state that computers would be very much cheaper.  I further expected us to
learn a great deal about computer organization but with a disclaimer that
the actual changes would be more apparent to the computer designer than to
the user.  These expectations have been largely realized, although I
failed to restrict my specification of man's recorded information to
specify the archival information that would otherwise be recorded on
paper, since the very availability of computer storage has led to the
accumulation of vast amounts of stored information and I certainly did not
mean that all the information stored in all the data banks in the world
could be stored in one data bank.

I did anticipate that two rather basic problems would have been solved
well before 1984, and alas they have not.  The first of these had to do
with machine learning, which I fully expected to have been perfected to
the extent that a learning mechanism would be incorporated into the
operating systems of all of our computers.  Machine learning was then, and
still remains, my main AI concern, and although I was none too confident
that I would be able to achieve the desired goal, I thought that others
would do it.

The second basic problem had to do with the imperative nature of our
then-existing programming languages. I fully expected a vast expansion of
interactive, question-asking capabilities so that our programming would
consist of making statements as to what we wanted done, not imperative
statements of actions to be taken, and we would expect the computer to
query us as to details that were unclear and finally to report on what it
intended to do. Programming as we knew it in 1956 would have ceased to
exist.  While some progress has been made, I fear that this still will not
happen by 1984.

Finally, I expected extremely cheap, convenient to use and portable
terminals to have been developed and that communicating with the computer
would be as easy and natural as communicating with an intelligent servant
of the human variety.  In spite of the rash of terminals that engulf us,
terminals are still not cheap, they still will not fit into one's vest
pocket, and they still require us to adhere to a host of artificial
conventions in communicating with the computer.

Most of my predictions were concerned with the effects that the
anticipated advances in computer technology would have on society.  It is
here that I erred in expecting much too much, much too soon.

I predicted the wide-spread use of personal computers that would be both
computers in their own right and intelligent terminals for communication
with large central computers and large central data banks, and that one
could communicate with these central data banks from one's portable
computer regardless of where one might be.  I think that we are well on
the way toward this objective, except for the matter of portability and it
may well be met by 1984.

Actually, the telephone industry let me down rather badly.  I was an early
advocate of packet switching for phone communication and I firmly believed
that the entire phone system would be converted to this well before 1984.
In fact, when IBM retired me in 1966, I tried, without success, to
interest several organizations in hiring me to help make this a reality.
The existence of such a phone system would greatly simplify the use of
data banks. I still believe that it will come some day but the inertia of
billions of dollars invested in a telephone plant, based on wired-through
connections, inhibits change within AT&T and many organizations are
fearful of challenging Ma Bell in her own monopolistic field.

I erred completely with respect to the development of large data banks to
replace libraries.  If I may quote: {\it ``One will be able to browse through
the fiction section of the central library, enjoy an evening's light
entertainment viewing any movie that has ever been produced (for a
suitable fee, of course, since Hollywood will still be commercial), or
inquire as to the previous day's production figures for tin in Bolivia ---
all for the asking via one's remote terminal.  Libraries for books will
have ceased to exist in the more advanced countries, except for a few that
will be preserved as museums, and most of the world's knowledge will be in
machine-readable form.'' }  Well, why hasn't it happened? 

I was over optimistic also with respect to language translation, stating:
{\it ``...automatic translation via the telephone will also have come into
use--- although perhaps not into general use, because of the cost and the
gradual drift toward a universal language.'' } This presumed both language
translation and speech recognition.  While my papers did not go into detail,
I can remember arguing that the proper role for machine language
translation was in its use with speech, because of the usual absence in
normal person-to-person communications of cultural and literary allusions
and of complex sentence structures and because of the ease with which
poorly translated portions could be questioned and rephrased.  Just as
people in the past have had to contend with poor quality transmission and
have adapted to it, so they would automatically restrict their utterances
to those speech forms that were properly translated.

I still believe that we will finally be able to have quite satisfactory
machine translation, but I am a bit more cautious now in predicting when.
I believe that it is only a matter of economics and as long as human
translators are so relatively under-paid there is really very little
incentive to justify spending the money that it will take.

A lot of money has been spent on speech recognition, as many of you know,
but much of this effort was directed toward reaching a specific goal in
five years and it was only at the end of this period that a few
organizations buckled down to task of really solving the problem. In
justification for my prediction, I now maintain that if we had had an
adequately financed program beginning in the early 1950's, when, in fact,
I did start a small speech recognition program in IBM, if this or a
similar program had been supported continuously until the present time, we
would now have in operation the kind of speech recognition system that I
predicted.

I also envisioned a much greater advance in teaching machines than has
actually occurred, both in terms of many slave teaching machines and of
many large master-computers which {\it ``will alter teaching methods on
the basis of experience.''}  Something has gone wrong here.  Teaching is
still a labor-intensive field of endeavor and it shouldn't be.  My present
guess is that the personal computer manufacturers will enter this
field in a big way, sooner or later, and that, when they do, the teaching
profession will be in for some disturbing developments.

I was right on target in my predictions as to The Banishment of Paper-Work
(this was the title of one of my papers), meaning by this the use of
computers in business.  In a sense, this was the least far-out
prediction to make because computers were already making inroads in this
direction.  However, some of the most critical mail that I received
(mainly from England) took exception to this aspect, pointing out the need
for an audit trail, the necessity for having human bookkeepers who could
testify in court as to the accuracy of their records, etc..  From the
vantage point of 1981 it is hard to understand how some people could have
been so short sighted.  Today the operation of such diverse activities as
maintaining the Social Security System, running a bank or scheduling
reservations on an air line could not function without the computer.

I further stated that {\it ``Process control with the attending automation,
while delegated to simpler computers, will have reached a very high degree
of development so that the number of factory workers will drop
precipitously.  These displaced people will go into the service
industries, and a few into the design and maintenance of the machines
which replace them.  Science and the arts will be flourishing.  The work
week will have been shortened to four days, but we will still have an
unemployment problem.'' } I did add that {\it ``Making this dire prediction will,
of course, help to nullify it.'' } Obviously, it did not, and I completely
failed to say anything about inflation.

I had one final word to say regarding intellectual achievements of
computers. I wrote: {\it ``The world draughts} [the paper was published in
England]{\it , chess and go champions will, of cource, have met defeat at the
hands of the computer.  ... Computers will have largely taken over the
task of composing and arranging music, at least for popular entertainment,
and many people will vie with each other in regard to the quality of the
mood music which their own personal computer, or personal program can
produce.  As far as literature is concerned, the computer will still be a
neophyte although who-done-its will be turned out by the million.''}
Well, well.

My concluding sentences were: {\it ``Computers will not have contributed as
much as some people have predicted with respect to making basic
contributions to mathematics and science and, somehow or other, all
attempts to invest them with truly creative abilities will have failed.

There will still be a place in the world for people!''}

Now what do I now think of my earlier point of view. 

Well, in the first place, we were all overly optimistic in those days, but
I can remember railing against certain people who were then making quite
unfounded statements as to what was then being accomplished or of what
would happen in the next few years.  In fact I also said in print, and I
quote, {\it ``As always, with any revolution, there is a lunatic
fringe---people who believe in magic, or who are carried away with their
enthusiasm for a new cause and who make wild claimes which tend to
discredit the entire undertaking. The field of artificial intelligence
has, perhaps, had more than its share of these people.''}

We also have had to contend with people who attempted to discount the entire
field of activity.  For example, in a 1961 book, Mortimer Taube of Columbia
University charged the workers in the field with {\it ``writing science
fiction to titillate the public and to make an easy dollar or a synthetic
reputation.''} Life hasn't been easy.

So, all in all, I would like to defend my 1956 point of view and claim
that my predictions were not too far out of line in the light of what we
actually knew at the time.

I went wrong, however, because I did not make due allowances for the fact
that developments in terminals, in data transmission facilities and in
programming languages would not keep pace with those in main frame
development, and progress in A.I. is critically dependent upon these
facilities. I also did now allow for the stulifying effect of Government
research support policies which were to demand immediate results rather
than long range progress.

\par
\vfill
\end