Alan Kay at IRC Conference (1991)

From Viewpoints Intelligent Archive
Revision as of 20:31, 5 December 2017 by Ohshima (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
Thank You Lora I'd like to extend a special welcome to all of our colleagues
from Lille and also our special guests
from Indiana guests from Apple as we
look forward next with the IRC conference to me it's very exciting and
say interesting in which we are as we
think about what we have done with information technology the best is in
front of us and last week we had a meeting where worldwide si came together
the eighties and talked to accomplishing
the issues we'd address and we look forward to the and looked at the
challenges that face us and as mr. Wood
was summarizing mr. wood as chairman of the company was summarizing the issues
that face us in the 90s one of the key
things that he focused on was using information technology and innovative
ways to really give us competitive advantage and I think that's an exciting
challenge for a lot of us in the room as we think about the things that are going
on right now and while as you will see tomorrow when we have a segment called
show business we're colleagues in the company will share what they're doing
with information technology we have so much more ahead of us in terms of the
way we can work in cloud collaborate together and you use these phenomenal
tools to really change transform
business for us to be successful in the
90s as we look at our vision for enterprise mm transformation as you
heard Laura comment on some of the key phrases and our visions extending
ability promoting global collaboration natural
access people that's our dream and we
think that all of you are going to play a significant making with an Eli Lilly
and Company I'm very pleased to our
guest speaker just small dr. Alan Kay
Alan was a has been a leader in terms of
computer technology thinking about the future where it's going to lead and he's
already aided in many important things that we're using day by day earned his
electrical engineering his people here science the University and Xerox Palo
Alto Research Center where they worked on some very unique and exciting
concepts these those we see today in the
Macintosh that many of us use he then went to Atari where he was the vice
president and chief scientist for that company and has subsequently joined Apple where
he's an Apple fellow he's one of the key people who dream about the future and
what's going to happen and as we look forward to his comments on the third
computer revolution I'm sure he's going to provide us some exciting kinds of
inputs so please join with me in welcoming Alan Kay
[Applause] [Music]
well thanks you for inviting me here it's wonderful to discover there's
actually a non-stop playing between Los Angeles and Indianapolis which I took
advantage of when we think about the
future people think about it flows out
of the past in a way that seems real
there's a progression of events that we can understand it been since the future
became a concept back around the time of the Greeks when Heraclitus this is kinda
change that's something we often say today that was a new back then because
for most of humanity in fact for most of the people even in the world today
they live and die pretty much the same
world not a lot of change for half of
the Earth's population now something happened of 90 years
that made everything different McLuhan said about the twenty that it was in
which change changed and this read
concentrate definition page is something we haven't been able to cope with yeah
it's something that is much more like trying to predict a thunderstorm
something that obeys the laws of physics
does come out of the past and the present somehow but it's nonetheless
very surprising when it happened I remember what systems programmer of
Anton ash Viereck research in the 60s we
were doing these most fear weather
stations on the supercomputer I want
urologists like today and to migrate and
looked out the window and I said what
are you doing that for you know dollars the weather why can't they shit he said
Alan you the atmosphere is so that if
you go out and plow a field that would
be enough to set up and be the start of
a storm and even a single oak tree being
in one place and not another can set off the instability and I was just looking
outside to see if any clouds were forming from any of these in the patient
models we were using use ninth out data for the entire northern hemisphere
so all these pauses the thunderstorm
trees and plowed fields didn't show up in this simulation that's sort of where
we are that the the future we'd most like in this century is the like the
thunderstorm because I guess people are
against in fact many of the things that it's by storm I have a collection of
made over the years one of the minutes
from 95 board meeting Union the
Telegraph comes in which they sigh Civ
Lee turned divestment elephant and six
in my mine minutes is one of the board
members in person would ever through such a contrived and there's actually
for it because they had recently come up with a printing number telegrams they
white paper and pasted on it came by and
they they come up with a printing they
now could have a of every business action who much read that way because
they can keep track like the reasoning for electronic I've every turn
his action so people soon to something
better but the fact they did the great
prediction was by lord help who was the
minister head of the Ministry of War great but in nineteen he said the
heavier-than-air aircraft will never fly I was pretty good because that was for
sure these are the kind of predictions
the government people wait until the
thing has really happen and then predict against it so to get a handle on this
stuff and particularly with regard to information which has did four hundred
years of what has happened with book we
have two ways that don't have to do with
extreme in the 70s when we were trying
to do personal computing people kept on saying well we can do that on the mainframe this problem was that they
couldn't imagine computing outside of the computing that they were actually
doing this we see quite a bit when Edison invented the movie camera in 1895
there was about a 20 year lag between the invention of the camera and
somebody's decision to first move it the camera was thought and in fact the there
were ads back then that said everybody
has a third row front senator seat and
the back out come to Cincinnati the idea
is the camera is a stand-in for the perfect place to sit in the audience and what it was doing was Vance surveys
and nobody questioned that because
thought of as being a replacement from the theater he went to see it in things
that were like theaters until DW Griffith came along and realized well
you could move the camera you could do tracking shots you can do montages you
could do quick cutting you could do could do all film language to the 20
years that you could actually with the
mainframe and the advent of computing
word an idea is that when a new media
makes its initial content from the old media and this comforts people to a
certain extent and then this comes along
because the thing that's really for emerges some years later and then
everything changes that has happened many many times in history so what I'd
like to do is to talk about the revolution so the first revolution is
not the invention of the computer because that was a revolution in
information representation and processing so it's in this same level of
the invention of the printing press but to talk about the revolutions within
computer dome itself and just to give
you an idea of how this started in a
kind of innocuous way let's take a trip back into the past for a bit and see you
know lights down for this and
here's a it could be in those office in
Lily's today not everybody not only has
the mouse and not everybody Lily looks at a 19 inch high-resolution
black-on-white television screen but some people do and for those people fact
for all of the people interested realize it sure 19 those pictures 24 years old
there's does Venter of the mouse and I
think of as the font desktop computing
what we call personal computing and here's what his system was able to do in
part and what I'm going to show you now is a demonstration in front of 3,000
people and a 22 years ago
so I can say I'd like to go to the Provost but I'd like to produce they get
big I'd like to say one brand look just
that low when I see it oh I can say I'd like to see one line
only I can see it but there's another
thing I can do so here I'm afraid I'll
need a different pictures of you so here's my dream with a picture drawing
capability or so slight laugh if I start from New York and here's the route I
seem to have to go to to pick up all the materials and that's my plan for getting
home tonight my supposed to pick up
there I can just point to that you know I see overdue books in all well there
was a statement that name go back you wanna fight once more to pick up the
drugstore I see everything all right can do things
if I want to change the products so that
was 22 years ago and since it's been so
long we can confidently predict that something like this will be in most people's offices by the year 2000 so
here's an example inking about non-traditional uses of computers around
1962 and it was part of a larger effort
that said batch processing stinks there
must be a way to have a closer relationship with the computer and as
Jack Licklider said to have the computer
it kind of a complementary partner doesn't have to be like a human but it
wants to be a complement and amplifier to the kinds of things that we're trying
to do an angle Bart's project was one of the responses to this and he had to
invent a lot of basic technology the first existing hypertext system there's
another tape that I don't have with me that shows face-to-face conferencing
through this system all done and back in the 60s and my response to this was to
design this machine this is the first personal computer I did this is around
19 was a desktop machine for very special
desks namely those made out of steel so
weight about 350 pounds had a fan 747 so
it's full of those little black boxes that hold the silicon in the machines
which you have about 50 and your IBM PC or Mac they're about 700 in this machine
because the integrated circuit was just coming in the integrated circuit is the
forcing function for this kind of technology we've taken some of the ideas
like Engelbart had which were implemented on a time-sharing system and
discover quickly that when you act direct which in it's almost impossible
to do it on time sharing because you can't guarantee these cycles that each
user needs when the user needs them to
have these famous lags that you get from from time sharing at a reduction in the
kinds of expectations you have from the system and really don't have something
like interaction you have something as a little more like negotiation you do
something and then you wait to see if I'm going to do support and so forth so
this is personal computer with first cuter years earlier but was the first
one that that had the name personal computer and it was very much tool like
to physical like levers and wheels there
are also things you would and math all
of them I think of gesture the M word is
manipulation because even if it was being abstract is actually a concrete
way of dealing with the abstract take all the abstractions make them into
concrete little symbols and all of a sudden they're out here okay after the
movement so all of these ways were all their extensions of the Box get their
way now while this was going going on
some people were working on a very different way of dealing with the computer and pull with it this is done
it RAND Corporation economists the
economists are which if you lay them end to end in the world to conclusion they
said people we have to do something
computer none of us know how to type so
here's what the rand people did the
connector paddled away so that we may draw a box in its place next one well
now it's recognizing his handwriting the printing in the box is commentary on
case the box is slightly too large so we
may change its size modern-day windows literally Android
that's an element to the box and drive flow from it to scan we then erase the
floor area and move the box to a new
position to draw a new box well I had a
chance to system for half an hour in 1968 and it felt completely different
from the Engelbart system Engelbart
system felt a little bit like you're doing a dangerous experiment and radioactive chemistry where there's some
feeder ladder and hoping nothing would go wrong and in in this system it felt
like you're sinking your hands right through the glass of the display and touching the information structures
directly this system fell intimate and
so I I thought of what this system was doing was kind of an more of an intimate
relationship this was a whereas the
Engelbart system as a customs fool is something that you watch and you
manipulate it and this kind of system
was something it was watching me and it was more like I was managing I was sort of showing it
more or less what I wanted it and it was figuring out the rest and completing it
so I have a very different way of thinking about it and so I made in 1968
I made this cardboard model of what an intimate computer would look like the
dimensions were arrived at by a variety of means one is that we quickly realized
that portable you have to be able to carry two
point-five herniation spur block is not portability and this this model we
loaded up with BB shot to see how heavy you could make it before you didn't want
to carry it around anymore and we asked crazy questions like what kind of a
computer would it have to be so that you would do something so mundane as to put
down your grocery list on it carry it in park and be willing to carry it out
again with two bags of groceries now let's let's many people hate that looks
like this right now can you hold it up
well then you don't have one like this so most people have paper and pens and
stuff on them but this is a different kind of thing than a desktop computer
and so now we're confronted by three
different ways of thinking about things the computer done was going from batch
processing to time sharing changing the route of the user
way the next big change was going from
time sharing to personal desktop computers again a huge change
particularly is your interface the new
user interface of using windows and icons and stuff is Wellum what we're
looking here is a third revolution these are not smaller desktop machines but of
course laptops right now our smaller desktop run the store because they're
just imitating what came before the way ms-dos imitated the interaction that you
did on the mainframe so this early thing
we don't just really realize that when you change the relationship of the user
with a machine a big way but the person winds up doing and the kinds of software
that gets gets built and used is very
very different now if integrated circuits where the where the forcing function for the for
the desktop machine what's the forcing function for this machine well it's networking
networking yeah these computers pervasively network
there will never be a time when the computer talking to a network watch they
won't have an on/off switch so testing
whether it's intimate or not non intimate technology has on/off switches
and the reason they can afford to have on/off switches is they aren't all that
important and use them they use them every day but they aren't your life as
paper does not have an on/off switch thank goodness and your base to make
some why the digital watches had LCD
displays that used a lot of battery power you had to put press them in order
to get them to light up remember those build a cajones they didn't last because
that's you don't want to go to that much trouble to find out what the time is you
want to find out the time as being part of your own environment not something
that's outside of your own environment so the metaphor here is another way of
extending humans but not the tool way what we call agents and an agent is
something that can take on our goals and carry them out on our behalf so simple
agents can manipulate tools for us that's nice we can take a nap while
somebody else is wielding the screwdriver and then agents can do
something the tools don't do and that is that they can deal with other agents so
once an agent has cloned our goals they can spread that goal around and in fact
we're part of just such a gold cloning set up right here in this room because
as Lewis Mumford pointed out many years ago the most of the machines made
people for most of human history have had other humans as their moving parts
like all the mecha and they were things
as large as civilization small hunting
parties businesses Boy Scout troops
whole panoply of things where a bunch of
people get together and they're a goal
work on it and so the M word is management here so tool is something you
look at and manipulate an agent is something that looks at you and you manage it and our boy that just go from
something like a control panel by the
human scale I'm sharing to the desktop computer so we found a tool metaphor for
people as we go to the end of it computer where we're going to be
connected with a pervasive network that
stretches all over the globe and more than a trillion objects that the whole
or of dealing with the computer and what
it's attached to is going to change and why why do you networks force agents
well when you have a trillion objects continually online you're not going to
spend a lot of time using sequel to try and find things all current information
receiving retrieving tools over retrieve and when you're dipping into trillions
of objects instead of a particular file where you happen to know the information
is you can't afford to try and do the retrievals yourself just as in the
Library of Congress where the card catalog is much larger than any library
in Indianapolis as an example Library of
Congress has some 85 million documents in it the card catalog is larger than
most hometown libraries when you go in there you don't use the card catalog you
actually are introduced to a person and that person attempts to find out what
you're interested in that person is an expert at finding those kinds of things
amongst all the different kinds of document collections and so we think
something much more like that it's going to happen so the here's another way of
thinking about these three revolutions institutional and we can make an analogy
to books and I will will in just a second but let's take a look at three of them like where is it well the mainframe
is owned by the institution here's a computer that you can own if you want
sometimes it's still owned by institutions because they're right in that border line we haven't decided
that's quite as useful as a car yet so only some people own their own machines
and then wherever you go
the intimate computer will go small and I am nearly connected you don't have to
be in any particular place in order to use the thing that was a nice parallel
here or the history of books the invention of writing is very much like
the invention of the computer it's a new way of representing information and the
first rebel revolution in the invention of writing you can think of as the
manuscript books that were copied by monks in the Middle Ages it's hard for
us to get a sentence back one of the
largest libraries in Europe at the end
of the Middle Ages was the Vatican Library as you might guess the Vatican
Library in the year 1400 had 392 books
now did a library look like it did today was
books on the shelves know there's still a an ancient library in Florence you go
into one of these things each book had its own desk they weren't
in the wall they're actually on a desk open and chained to the desk because
these books were almost priceless they took about 15 years to copy so you went
from desk to desk to read a book and if you look in one of these antique
libraries like the one in Florence nothing dry and that this looks like a
time-sharing bullpen all right it's it's
like a room with 50 terminals in it too expensive for any one person to own now
when Gutenberg came along he made books
but the books were the same size as the middle-aged manuscripts why because
nobody knew what size a book should be that's the size that books were what did
the what did the fonts look like including bars books well I look like
script finest at the monks but there are thirty-six point type 30 and 36 point
type half an inch high big big books and
their thought was is that they would still be owned by institutions but now
cities could own them so huge advance
with Gutenberg but it didn't make it how
many people here have ever heard of Aldous how many people here use a piece of
software called Aldus PageMaker okay
Alice's last name was not PageMaker would be great if it was his last name
as Manutius and predator that born two
years before the printing press was invented so here is a guy tire life
growing up in printed books and when he
was in his 30s he decided a couple of
important things one is that there should be lots of books Bibles weren't
enough and so Aldous and his sons over
the century and operated printed some
40,000 different books in their lifetime and they were one of the major sources
for ideas of the Greeks and Romans because they gathered up all the Greek
and Roman writings they could find and publish them as as books but Aldous did
something they'll always endear himself to me decided the book should be the
size they are today and wonder if anybody here knows why Elvis decided
books should be this size oh okay
you heard me give the talk before yeah that's not fair
he decided they should be this size because that was the size that saddle bags were in Venice in the 1480s okay so
Aldous had come to 30 years after the printing press had come to and a
tremendous realization and that is that books could now be lost
think about what that means because books could now be lost they can now be
taken with things could happen to them and they had to be a different size and
that size change forced Elvis to become one of the first typographers because
these script fonts did not shrink down very well he started developing fonts
based on Roman fonts and developed
lowercase characters and most of the fonts that we read from today Aldous had
a hand in designing so he had to design the display now for doing these newly
intimate books so important thing for us
today is the thing that if when you look
back in history that what happened between Gutenberg and Aldous which is
where we are today in computers counted not a whit nothing that happened between
Gutenberg and Aldous was significant because all of the things that we think
of as flowing from the book actually happened after Aldous had made it into an intimate thing that could now be lost
so we have to ask what our computer is
going to be like when we can afford to lose our computers something that most
of us would not countenance today now if we jump into a pool with the Japanese
watch on that cost 1295 we say yeah
that's too bad but I'll go get another one if we jump into a computer with our Mac
or our IBM PC we have a very different feeling about it because all of a sudden
we've ruined something that costs everything we have that the multiple
thousands of dollars is not the right scale for something that is like a
dynamic book and so when we talk about
things one of the best ways of predicting the future is to try and figure out what paradigm er in so in the
not not actually too many years in the future there was a paradigm shift in astronomy
where went from most people believing
that the son went around the earth the Ptolemaic
way of thinking about things suddenly got turned around by Copernicus and
Kepler and Newton into a heliocentric
way of looking at things and the world was never the never the same again never
the same on a book called the structure of scientific revolutions Thomas Kuhn
pointed out that the reason most scientific revolutions take about 25 or
30 years to happen is that you have to wait for the old physicists to die off
now if this is if this is happening to scientists who are supposed to be always
open to new versions of the truth but if scientists actually get imprinted like
ducks on old versions and old worldviews
and have to die off before new ideas can come along then it's no wonder that we
have the hardest time recognizing these paradigm shifts even wanting to see that
they're happening and adapting to them now the problem we've had in computing
science is that the technology is moving along in such a way that we've been able
to recapitulate 400 years in the
development of most technologies in just 40 and so we've gone from batch
processing to time sharing one of the
funnier announcements that has been made in the last 30 years was when IBM
announced that it had invented time sharing that happened around 1970 or so
which is just about ten years after it had been invented but yeah but now it
was real now it was time sharing with capital T and capital s actually IBM had
been against time sharing in many ways for years until it it actually come up
with a version of its of its own that's a perfect example of a company that has
a vested interest in one particular technology lagging behind and then all
of a sudden saying oh here it is and of
course the personal computer companies like Apple have to be very careful
because if we think of ourselves as a desktop then we're in exactly the same place as
IBM and Dec are for this new revolution that's coming along which is quite
different and in fact it's the kind of hardware revolution that the Japanese
are much better at and we are having to do with a combination of both highly
dense silicon artificial intelligence and digital wireless communications
Communications is an interesting thing the one of the simpler ways of thinking
about it is my friend Nicholas Negroponte of the Media Lab at MIT likes
to say he says everything that is going through wires right now is going to go
through the air and everything that's going through the air right now is going to go through wires so it's a complete
inversion in other words television is going to be forced into fiber for
various reasons including hive eye and
telephone which now goes through wires is going to be forced into the air
because we can't afford to wait to get
to a telephone for all the things that we need to get through it not just being able to talk to people when we want as
we do through cellular telephone but we need to be able to exchange data in
various ways and so we must be able to do it Wireless so the sleeper technology
in the network revolution in the next 10 years is wireless networking of various
kinds we are not in the u.s. ahead in
this because we're still conservatively trying to we think improving what we're
doing is making better wires and so
forth very similar to the railroads in the 20s who thought improving what they
were doing was to make faster locomotives and lay more track so while
they were looking like this their competition had taken to the third
dimension and was flying over them no railroad in America ever invested in an
airline company it's interesting a survey in the 20s would have shown them
that there would be more track and more railroads but they would never have
asked the question of whether they'd still be in business in 50 years okay
and that's what these paradigm shifts due to you now 10 20 years from now
there still be mainframes 10 20 years from now they'll still be desktop computers 10 20 years from now they'll
still be COBOL right that's not the
problem the problem is is what are the tools that are going to give you the
leverage that you need and those tools are not going to be the mainframe
they're not going to be the desktop computer and they are certainly not going to be COBOL so let's take a look
at contrasting these three we call them paradigms after scientific revolutions
how do we measure performance well in the mainframe it's response time there's
a phrase called path length which is the
millions of instructions that you execute before you get a response back
at the terminal I wants to ask three thousand Macintosh developers if they'd
ever heard the phrase path length and none of them had because the issue on
desktop computers is MIPS sort of like
in the 50s you know those great old dodges with 450 horsepower that you could lay a strip of rubber a block long
it's in particular most of you here are not old enough to remember that
wonderful that was when gas was still 25 cents a gallon you could but that's not
the issue because the issue actually is going to be access it's can I find the
resources I need I'm going to be connected up to damn near everything in ten years but can I find what I need I
don't even know what its name is a lot
of things I need I don't even know that they exist they're actually analogous to
the things I'm working on not directly related how am I going to find them once
I find them and I get them shipped to me what's the probability I'm going to be
able to use them I can barely use another application that I get from my personal computer at egghead software
what if I get what if somebody in in hindustan has done the perfect
application for me and it's delivered to me and it's sir
this is exactly what I want and I can't sit down and use it pretty frustrating
those are gonna be the issues of the next 10 years kinds of data yeah I
thought these green and white listing papers had gone forever and I saw one in
my doctor's office the other day and that caused me to look into his other
state-of-the-art medical practices
eventually changed my doctor they are
still around fonts and graphics on
personal computers on the Macintosh style of doing things and all other media coming in in this next revolution
and it doesn't mean just all other media on your machine but because the next
revolution is a networked revolution that means that we have to be able to retrieve all other media and it doesn't
mean just shipping film clips and stuff
over the networks that means that we have to be able to retrieve into the interior because most retrievals are
done by asking things about the content so it implies that some of the work for
instance being done by the Getty foundation I'm being able to retrieve pictures by talking about their contents
or even making a little sketch of a painting that you've seen 20 years ago
and having the computer find all the things that are like the sketch are
implied here how do we integrate
information very painfully on the
mainframe usually done through files
sometimes with a relational database I was shocked when I did this slide that I
did not know what the actual penetration of relational databases was I presume
that was like 90% so I actually called up some friends of mine at the Gartner
group and discovered that for the fortune 1000 companies in the u.s. less
than half of the company's information is in relational databases and I presume
that's more or less true of Lily - that is remarkable considering
relational database was not a big improvement over flat files and as more
than 20 years old so that was a very very slow adoption rate on Macintosh
style computers you can cut and paste
from any application to any other application and to the extent that
things are object-oriented the result
will be compatible in the future instead
of cutting and pasting pictures the things that we see on the screen what
will be actually moving is the applicant views of the application itself whether
the application will actually still be active but to be used as a component of
any other application now some of the operating systems are just starting to
happen now allow you to do or at least
start to do things like this how do you give commands remembering type so one of
the ways of thinking about is what is the job of the user interface well job of the user interface and this way of
computing is sort of like a control panel of a nuclear reactor you know it's
to give you access to function there's all this function in there and we'll map
it into control keys and different
screens and so forth it's not a bad way of doing it it's used for other kinds of
things in this user interface it's
seeing point and the job of the user interface is not to provide access to
function anymore but to make the user aware of what the possibilities are in
other words the user interface in the Macintosh style is a learning
environment it's job is regardless of what you're trying to do its job is to
make you aware of the whole machine and
the result is about the average number of applications
used by ms-dos users as an example is about 2 and the average number of
applications used by a Macintosh users about seven and a half and the reason is
is that you expect in this kind of user interface that the new application that
you may have borrowed from somebody or may have gotten off the local area net
that you expect that the new application is going to be very much like the
applications that you've already used we we look for a transference of something
like 70 to 80 percent between applications of familiarity so a person
can do 70% of the operations from one
application to another without having to open the manual and most people never
open the manuals which actually is not great because is not a perfect interface
there are usually unfortunately things in the manuals that are worthwhile knowing the interface doesn't make up
for and in the end of it computer because we're dealing with agents it's
going to be much more answer for morphic it's going to be gesturing like you saw
that's going to be speech in a variety
of ways could be written English it could be spoken English but it's going
to be much more casual it's going to be something much more like advise field of
interaction feels like editing a line of text feels like layout on the Macintosh and
it feels like conducting on the intimate computer because what you're dealing
with our active elements rather than
passive ones so I called the user interface that with the overlapping
windows and so forth when I wrote my thesis was called a reactive engine it
was designed to react to the kinds of things that you're doing and the new
user interface will be proactive it will be actually doing things ahead of you
trying to anticipate the kinds of things that you'll need
how do you print big impact printing and
enormous big laser printers now personal
laser printers and no printing in the
independent say now people have been predicting this for years and as people
like to say in the newspapers there's more paper now than there ever has been so how can this prediction possibly be
true and here's the reason this is really going to happen two things are
going to happen over the next ten years that are going to make printing a very
low yield operation one of them is hyper media not just having different kinds of
media but having the media interconnected in many different ways
that means that printing out any piece of hyper media is going to destroy a lot
of what's interesting about it hyper media is designed to be dynamically
dealt with it doesn't layout linearly in a very good way because it makes use of
this idea that the computer always has more dimensions than your data so if you
have n dimensional data space you're the computer can give n plus one dimensions
and bring all of the things to one place
extra if I had a if I had a pencil I'd illustrate that but I don't and the
other thing is that the we're going to be rapidly going from data structures
which are passive representations very much like the words in a book we're
going to be going from data structures to active simulations people don't want
just to see what happened in the corporation three months ago they really
would like to be dealing with a model of the company like people do in a crude
way with spreadsheets a model that they can ask what if questions of to try
different kinds of simulation and other kinds of gaming to try and figure out what they should do next
so both the multi dimensional and the simulation aspects are absolutely
antithetical to printing I believe that both of them are going to dominate to the place where we will not be as
interested in printing and the little machines with the page size displays
that way less than two we'll come along on just the nick of time so that we don't care whether we
print or not but we actually want is an active copy of that information on the
machine itself that we can look at whenever we want so I think all of the
technologies are going to hit together just at the right time and printing for
most things will go away just as there
are very few things besides and university will probably still print
University diplomas yeah I can write because we still use vellum and I guess
most universities today don't even go to develop or parchment anymore because
it's so expensive they probably use a Mac to do it cheap what kind of software
one purpose custom application typical
kind of thing that an Arthur Andersen does for a customer like Eli Lilly or
could be a big billing system it could be a chemical analysis system it could
be an airlines reservation system
personal computers we have we can run those kinds of things but what we
generally run on them are what are called generic tools these are tools that are like frameworks desktop
publishing systems spreadsheets they're frameworks that the end users do stuff
that is dangerously like programming where ease Mis people know no end
because they're all of these end users out there subverting them some end user
needs something that puts it into the Mis Department and gets back a friendly
note saying yes we'll be glad to do this for you you will do it in COBOL on the
mainframe and you'll get it in about a year and a half so the user goes away
and a few months later they think well maybe I should try that in Excel they
sit down and start hacking up something in Excel and by the time the COBOL
program line comes along a year and a half later the
personal computer user has forgotten he even asked for it because he's built
some incredibly hairy Cluj all by himself that does 90% of what he wants
so their various ways of going you know one of the things you could do is stamp out this kind of subversive activity by
instituting thought police in the in the companies so the only M is COBOL
programmers allowed to write code or the other thing you could do is to say well
users do have a tendency of trying to get on with it and maybe what we should do is give them
better tools so this whole notion of end-user programming was something that
was born at Xerox PARC was the idea of
the end users are going to do stuff let's try and figure out safer
frameworks for them to do this we don't expect them to be architects but they
sure are going to build dog houses of various kinds for themselves so let's see what kind of a dog house we can let
them build before they have to go and study architecture to do it and this is
even more so in the next ten years
because we're going to go to active self customizable components I'm going to try
and explain that a little bit in a couple of minutes that will make it much
much easier for the end users to put together their own their own tools so
the languages change COBOL is a data
structure and procedure language like Fortran and Algol and Pascal that's what
we call in the biz weekly modular one of
the ways to think in other words as modular modules are things called data structures and things called procedures
and this is not a happy division of the computer stuff because when you move a
data structure from point A to point B and there's maybe a thousand miles in
between them you have to write a lot of
code to use that data structure when it comes to the other end if you don't send
the code along with it and there's a there's a rule in this kind of program
I mean that you never send all of the code that it's needed and when you
change the representation of it you never change all the procedures that new
the old representation and so the result
here is some lightning bolts so you wind
up with very large fragile spaghetti-like code with this way of
doing things nonetheless this is the dominant way of doing things there are
whole companies that make their living doing nothing but writing programs like
this that seem to work another ah Allen
let's see if I did I do a whole bunch of
them like this No so ignore that nor the one on the right take a look at this
take a look at the Center for a second in object-oriented programming what you
do is there are two things one is you don't care how the data is represented
you just want to use it so you can stick all of the state-specific stuff inside a
thing called a little cloud called an object and the way you make it
independent of its representation is that you put all the procedures that know what the actual representation is
inside there to you make a little package when you send that package a
thousand miles you only have to write a very little bit of code in order to use the object safely by sending messages to
it the representation is quite independent of how you use it in fact the representation can change on the fly
while people are in the in the throws of
using it so you have this interesting thing that in some object-oriented
languages these systems programmers can be improving the efficiency of objects
that people are actually using in various structures and the inside and
the outsides are so sealed off that there's never any time when if one is in
danger of having a having breakage now
that's a object-oriented stuff is about 25 years old so it's just starting to
come into vogue now as I saw in one of the conference
things for this conference that it is now the 1990s buzzword which is really
kind of funny because I made up the term object-oriented in 1967 I did not invent
object-oriented programming I invented some of the languages they're used in it
but it was an idea that goes all the way back to the mid 60s and but this is the
paradigm shift that programmers are just
now starting to take you say you think you're conservative take a look at a
programmer some time just because they're using fancy technology doesn't
mean that they are they were about as conservative as a person can possibly
get once they've learned something they imprint on it like a duck to Konrad
Lorenz and never give it up no matter how spidery what it is that they're
doing now in the next 10 years we really
have a problem that is that the pervasive Network actually isn't strong enough to deal with objects being sent
randomly all over the earth and being done by various people and have any
possibility of using them you want something more biological for instance
if you take a a muscle cell from your shoulder and put it down in your thigh
muscle you don't have to reprogram anything down there in order for that
shoulder cells to survive that shoulder cell like all the cells in your body
knows what it wants from the environment and it's really proactive about getting
it and if there is stuff out in the environment it can use to survive it
well that's why we can build a trillion cell organism as how many cells we have
is about a trillion is because each cell spends about 90% of all of its effort
looking out for itself and protecting itself from the ravages of the
environment around it so most of us believe that there's going to be a new
way of programming which is called agent oriented programming or something is called modular control programming in
which what sand is something that is very shiny in
other words when you send something from A to B no code has to be written at the
other end in order for this object to be used it knows enough to adapt itself to
the environment that it finds there all by itself and no code has to be written
and this way of doing programming has been around now for about ten years just
give me an example of so these are not new ideas in the sense of research but
these are ideas that are about ready to come out of the research labs and start
making it an impact how big is the big
program COBOL million lines that's a big
one Arthur Andersen does them all the time an enormous Li large program with about
the same functionality is only a hundred thousand lines they're only a few of
these around cuz most programs are much smaller than that and what is quite
remarkable in the last 10 years of doing these agent oriented programs is that
programs with about the same functionality that have been deployed out in the real world have come to about
10,000 lines in other words there are these huge factors of 10
between these paradigms this is where it starts getting interesting right because
Americans are not interested in new ideas just because they're interesting that is not our culture never has been
our culture but here is actually a practical reason for getting interested
in that there are factors of 10 differences between these paradigms why
bust your chops working on one of these 1 million line fragile programs when you
can do it in a tenth the time with a tenth the resources wind up with
something that is much easier to validate much easier to maintain
now these used to be hand waving back in
the 70s when we're trying to promote object-oriented programming it's only
been in the last 5 years or so that very large companies have started to deploy these side-by-side
with some of the older ways of doing things than real comparisons have been done arthur anderson is a good company
to talk to about this because they have done some very large programs in
object-oriented style and large ones an agent oriented style and gotten
tremendous results so these are now field tested results
so another way of looking at it you can think of each one of these is a
kind of an architecture and in fact you think of each technology is the kind of
an architecture so if you know how to put posts down at the ground and put
lentils over them you can build structures of a certain size and we
still use that technology today for building small things Greeks used it to
build the largest things that they ever built with your temples if you want to
build something much much larger and moreover that lets light in you need a
different architecture the Greeks never built a structure with an arch except
for some grave structures and they never could actually see that you could get
the graves up above ground and they'd work there too so they missed completely
out on this architecture and it's quite remarkable that the amount of stone in a
Gothic cathedral like Notre Dom is very similar to the amount of stone in a
structure like the Parthenon seems incredible but the Gothic cathedral is
mostly air because the stresses have been balanced out and there's plenty of room for putting windows in when you go
to a completely new way of doing things that buckminster fuller called tensegrities
which means using things you can pull on as well as push you wind up with things
like the geodesic dome which is quite easy to build a geodesic dome that will
fit over all of the Gothic cathedrals that we've ever built on this earth so
each of these architectures is a qualitative difference in approach and
for companies it's the qualitative difference that gives you the leverage
now of course this is risky I remember
when I first did the overlapping window interface at Xerox I showed this to a
Xerox executive and this was 1971 or
something like that and I bound up the demo with a flourish and I said to him and what's what's more what's even
better about this is that this stuff only has a 20% chance of success I'm
taking risk just like you asked us to and he looked at me right square in the
eye and he said boy that's great but just make sure it works
so he wasn't interested in the romance of it and like most executives he wanted
us to be in the 20% 100% of the time right that's not the way you do it the
best lifetime batting average of any
baseball players Ty Cobb's it was 367 that meant about 2 out of every 3 times
they went up to the plate he made out and so research requires actually having
some outs as well as having some hits now what's the key to all this stuff key
is not to get embroiled with just
function if you ask somebody for a screwdriver and they give you this you get a very angry at them you'd say why
this is the functional part of a screwdriver and you'd say no I need to
use their interface because it's the function plus a user interface that
gives me the tool now when I made this slide it forced me to look at a
screwdriver in a new way and when I looked at it I thought boy this is a
terrible way of designing a screwdriver just think of it this the mechanical
advantage on the screwdriver is the ratio of the diameter of the handle to the diameter of the shaft it's not big
the best way to grab the screwdriver is like this but then it will slide off the
screw and if I grab it on the top the way you grab screwdrivers I don't get
them very much purchase on it so look at this and I thought god that is a terrible design for a screwdriver that
is the ms-dos of screwdriver design
yeah the screwdriver this problem my hobby is organ building and I have
one of the great old books of organ building it goes back to before the French Revolution and it has pictures in
there of the tools they use and included in there are screwdrivers that look just like the screwdrivers we have today so
this is a design that goes back more than 200 years and it's terrible
it doesn't fit to the human being so what if we wanted to fit it to the human
being well we wanted diameter as large as we possibly could get for maximum
mechanical advantage we need to grab it from the top and we need maximum
friction so what should it look like it looks like a ball does anybody ever seen
you know screwdriver look like a ball yeah Pyrus so years ago somebody looks
at the screwdriver the same way I did and said boy that's a terrible design maybe it's the COBOL of screwdrivers the
design design but if you think about if you want to fit to the human you
actually have to look at the human so if we want to do use their interface with
primates we're a primate then you have
to go with whatever is going on in the primates mind and that is not easy
because most of our models of the way
the human mind works are not are made up
out of common sense theories that are unfair I give you an example of that so
here's an upside down face and see if you can look at it look at this picture
for a minute and then see if you can see anything wrong with his face
yeah what is it the mouse upside down
okay see anything else
okay ears no how about eyes
no fair turning your heads to the side
okay so let's pretend we're in school school is a place that treats knowledge as a
fluid that can be dispensed from the
full teacher vessel to the empty student vessel drop by drop and it's usually
done by encoding the knowledge into facts in English sentences so let me
give you a fact about this picture what we've done is we've taken a picture of a
young girl extracted her eyes and mouth turned them upside down put them back
into the picture and then turned the entire picture upside down okay is that fair okay so having said that
you should be quite prepared for what it looks like right side up
I gotta turn this back because I discovered nobody will listen to me when it's on the other way okay so what
happened there I gave you exactly the same information twice once and in the
terms of a sentence it went in through the ears and went over to the left side
of your brain got evaluated as a proposition you have an okay reaction to
it and then I showed it to you as an image and went to the back of your brain over the right hand side of your brain and
got an emotional reaction out of you this is why McDonald's does not run
print ads saying if you eat a McDonald's hamburger you'll become a better-looking person because in English that's absurd
but if they show a picture of really good-looking people eating McDonald's
hamburgers then you'll make a different inference because it goes to a different
part of your mind in other words we are multiple minded we have different ways
of dealing with the world and they're not in sync with each other in fact we
have a special part in our mentality that thinks things are more true if they
rhyme so if I were to say to you a Big Mac and fries will slender your thighs
sound a little more convincing
if you think about it you can imagine why truth and rhyme go together if you
think back over the several hundred thousand years of how we've passed our culture down to each other that the
truths of the culture were passed down in written rhythmic schemes for memory
and the association with truth is something that advertisers still used today when they make up slogans it's
something that we're entirely unconscious of and because our mind
isn't a seamless fabric we can get
tricked very very easily so just very
quickly PJ said we have three stages of
intellectual development going from a body-centered stage where an object is
to grab it visually centered the kid thinks there's more water in the tall
glass and symbolically centered where the kid says oh there can't be more
water Bruner who's an American psychologist took a kid who said there's
more water in the tall glass and immediately put a cardboard in front of it and the child changed his mind said
what though but there can't be and Bruner removed the cardboard and the
kids said no look there's more water in the tall glass and Brenner put the cardboard back and the kid changed his
mind again so if you have any six-year-olds you'd like to torment
so what Brynner was doing was showing that it's not stages as Piaget thought
but actually parallel mental activity
the fact I was just talking to Brenner on the phone a couple of days ago talking about how well this model has
stood up over the last thirty years or so that in a very simplistic terms we
have at least three quite separate ways
of dealing with the world that you can think of as separate mentalities a body-centered way a visually centered
way in a symbolically centered way they can be appealed to differently they have
different logics they have different ways of remembering the world and you
can use them for designing ways of communicating with computers so if we
have a body-centered way of knowing the world in fact this is what Orient's us
in the world we need when we're communicating with a computer to have a way of sticking our arm into it the
mouse is one of the simplest ways of doing that it gives us a kinesthetic relationship we need something to relate
to so icons are a wonderful way of being
stand-ins both were symbolic things and for real objects in the world and here
are two but fact let me ask the audience here how many people flip channels on
television we have cable television here I noticed so whenever you have a cable
you flip it's okay we're not we're not gonna take a record of this so a bunch
of people do how many people have ever flipped in the middle of a movie that
they've hadn't seen for like 20 years
okay how long did it take you to recognize that movie that you hadn't
seen for 20 years a couple of seconds think about what that means 20 years ago
you saw this movie probably with somebody that you were in love with or
something so you you're really paying more attention to them than the movie
you didn't know you were gonna be tested on that movie 20 years later you pop it
into the middle of it without any preparation and in 60 frames you have
recognized that you saw that moving before it now let's test it out how many people have had that happen to
let me just just raise their hand so we can see okay how many people knew what
was gonna happen next yeah it's just
about right about half the people know
what's gonna happen think of what that means over the context of your entire life your visual mentality is set up to
memorize every scene you have ever seen
you can imagine why if you think back over the last couple hundred thousand
years and so appealing to the iconic is
something that is almost a no-brainer even though it's taken many years to do
it the other thing that's interesting is the iconic mentality can keep track of a hundred different things at once and if
we had a hundred images of animals you could find the elephant four times
faster if it were a picture than if it were at work that search is done by a
different part of your brain than the search for a work then you need a
symbolic environment which we called small talk an object-oriented
environment to tie them all together and you wind up with something like this
which is the first interface that looked like a like a Macintosh so this style of
interface I'm not here selling Macintosh computers this style of interface which
is now spreading to the IBM PC is one that was actually designed from a model
of the way the human mind works and it's fitted 'no stu people is what has made
it go out into the world and be
successful we're gonna have to do the same thing when we go and think about
agents now here's the screwdriver diagram for agents it's not function but
functionality is not a handle but the
agent has to have language and context what is context mean well an agent for
UNIX UNIX is one of these operating systems that graduate students love that has 500 commands that you have to
memorize Berkeley did a nice agent for
Unix and a user you could type to it in English anything that remotely had to do
with dealing with computers the
interface could translate and carry out a UNIX command to do it so one day about
five years ago somebody typed to it I just ran out of file space I need some
more those words 20 seconds later the agent came back and said okay now you
have plenty of file space
later on that afternoon the port user
discovered that it had deleted about a third of his files now think about that
that's a high score for an artificial intelligence right the user wanted more
or something the agent had to figure out it meant less of something they had to
find out how to do that went and did it now this story would not have a happy
ending if this were on an IBM PC or a Macintosh right because those files
would be gone forever in this case though there was an undelete in the file
this file system for this particular system did not throw away files when you
said delete it just marked them for being archived in a particular way and
so that user when he discovered his files were guy just said undulate and
they came back this is what I mean by context artificial intelligences can be
very very smart strategically but they don't live in the same world that we
live in and we have to put them there and so when you combine the three
together you get something that will function as an agent one of the things
that agents have to do is to track to
track our goals in some way they have to stay with us they have to watch what
we're doing and try and understand what it is we're trying to do and I looked for a way of illustrating this and I
will try and skip to it here
you just bear with me this is kind of a
grant this is an agent that was done at MIT that I think illustrates what all
agents have to do pretty well
so what we have here is a flute player and the computer is going to play the
harpsichord the computer can hear what the flute player is playing
[Applause] [Music] [Applause]
real people make errors so we want the
system to be able to track what we're doing even though we're not perfect even
at the things we said we were going to try to do
[Applause] now you can imagine the poor computer
saying what is this guy trying to do but
the point is in the next 10 years that's what we want it to say what is this
person trying to do and how can I help so in the next 10 years the apple cart
is going to be overturned again we're going to go from the second revolution
which we're now midway through most large companies are just struggling to
acclimate this new technology and this
third revolution which is going to be a much larger one
than the desktop revolution is going to be in full swing I pointed out the Apple
that we could be incredibly rich by 1995
because we have the best second revolution technology and we could be
out of business in 1997 because there's going to be a complete switchover just
like Digital Equipment Corporation was in great shape just a few years ago and now
it's struggling like mad why well it just didn't move with the paradigms so
these kinds of things are really scary because again it's like a thunderstorm now we can anticipate a little bit we
don't know when the tornado is going to hit but we can certainly build storm cellars we can decide we can be
proactive about the kinds of things that are going to happen to us we can start
getting interested in networks we can start asking questions about what are
the kinds of goals that we're trying to do in this company and how can we build agents that will help spread those goals
around and help amplify people but still it's pretty scary sometimes when
companies get scared they hire consultants it's a good idea but
consultants are like a person who knows 100ms to make love it doesn't have a
they can advise but it's the people who
are actually within the environment that can do the best if they can find a way
of getting outside the reason that consultants work is because the consultants come in from the outside
they have a different perspective but they're weak in being able to tell you
exactly what you should do and how you should do it if you can get outside the
current perspective that you're in take another point of view and realize that
point of view is worth 80 IQ points right we're not any smarter than we were
a couple hundred thousand years ago imagine a person with a 250 IQ in 10,000
BC or Leonardo in the early 1500s with
it maybe a 250 IQ he still couldn't transcend the century that he was in our
smarts come from the change in representation systems that our culture
adopts so we have to do it ourselves
point of view is worth ad IQ points and it's scary to get to go to that other
point of view but as in all scary domains the best weapon we have is the
one between our ears providing us loaded
thank you
now we have some time for questions and
there are actually some microphones around I think I'm not sure I see them
but they're microphones in the aisles and the balconies of have any questions
about either my major or minor arguments
I'd be glad to entertain them do you
believe everything I said yeah
well actually not so well not so well
one of things I've become very aware of is how much of my mind is in paradigm -
I bought the paradigm 3 ideas the Apple one I came because we could see paradigm
3 starting to happen in the late seventies as soon as we got very well
networked up at Parc and we realized that the windows and icons interface was
not going to survive pervasive networking we realized that there was
another way and people had thought about this way and this notion of agents goes
all the way back to the late 50s was one of the original dreams of John McCarthy
who is one of the founders of artificial intelligence and and so forth so it's
not an old AI it's not an idea that was startlingly new but one of the things
that I realized in the last couple of years is that I am so much more comfortable I think it's partly because
as a musician I'd much rather play in the band than be a conductor as I just
know I like to grab on to things and so
my natural bent is to be a tool type user and so even though I can see
philosophically what's demanded by this paradigm 3 and in a way I'm kind of
trapped by the stuff that I helped invent you know I I've not had
particularly good ideas in paradigm 3 except pointing out what it's likely to
be like and stuff but I mean fundamental technical ideas of solving some of them
the problems are being done by other people and the that's a it's a very
interesting experience to be a
soothsayer and inventor of technologies and to realize that you're in a
different world and the one that's going to going to happen fortunately I like to
play music so I'm I have my retirement well planned but you see what I'm saying
I mean this this stuff the the toughest thing I think is particularly for
managers who tend to be embedded in one particular way of thinking or a
collection of ways of thinking to be tolerant to be able to see that there
are other ways of thinking and to be tolerant the people who are doing doing those that was the biggest problem at Xerox
was that Xerox turned out to be very
intolerant of things that weren't like copiers after all and so when we gave
them the Macintosh style way of doing things in 1975 they literally could not
see what it was they just could not see it and we didn't give it to them as an
idea we gave it to we had 500 Macintosh like machines connected to Ethernet
networks with six or seven pages second laser printers in 1975 they still
couldn't see what it was so that to me
is an example of what the paradigm shift problem is all about and I think for
quote unquote older people are people who are very efficient and thinking one
way you just have to realize that that isn't the only way to think and so I
think that that way of doing things I think that the people who are coming
into a new set of ideas will have enough freedom to develop them but it's tough
that's why so many new products have had to go to different companies than the
companies that paid for their invention because innovation is distressingly easy
to schedule you just rub money and smart people together but incredibly difficult
to recognize because Brian Eicher
something that's innovative is something that you haven't seen before when you
haven't seen before the tendency is to reject it or not understand it
well it had a pen stuck in the side and
it was based very much on the pen based
system that I showed from RAND Corporation however anybody who has ever
used one of those systems and very few people have even the people who are
developing pen based systems today have ever used a good one once you've used a
really good one you realize that you still need something in order to do
volumes of text now for years in the
late 60s I tried to get people interested in steno type now where you
can go 180 words a minute it's like a court reporter does and I just can't get
people interested in learning that thing but there needs to be something the
fastest you can go on a on a pen writing system is maybe 15 to 20 words minute
that's not fast enough for doing all of the things that you're going to need to
do although the second version I mean
the the classier versions of the Dinah book that I did showed the keyboard
displayed on the display when you need it the display was touch sensitive okay
that was that was more of the vision and something like that will probably be
more true than than having a discreet keys yes
I think they're in paradigm zero aren't
they I mean not sure that well I think
the the major that's something I've been very interested in actually I've done
some work with the Indianapolis Children's Museum and and so forth and I
think the major thing on it is that the
values that the kid brings into any kind of learning situation are the values the
kid gets at home and I think the the
value space is more important than any kind of technology that can be be rigged
up I've worked with children and computers for 22 years and I've never been in a successful school situation
that didn't have very strong parental support that is the hardest thing to do
now saying that the technology is not a band-aid or imagine going event that
it's actually an amplifier or something it's like if you wanted your kid to
become a musician one of the things you could do is buy a piano or get a piano put it in a classroom but if there's not
a value system about music there's musicians will tell you the music isn't
in the piano right that where we'd have to let it vote so the musical impulses
inside of people and the technology is there to amplify it I think that model
works very well for computers in schools so the the really hard thing about
computers in education is that you have to fix the problems of Education without
worrying about the technology already got a great technology called Bo okay
basic organization of knowledge think about that solid state two and a half
megabytes cost about three bucks a megabyte a couple hundred million titles
and stuff and we've got technology up the kazoo and we always have had what we
need is actually some understanding of the way people's minds work so we can set up environments that are not painful
there's no reason why learning should be painful and we have to have some
understanding of what it means to
amplify those human impulses with technology then I think it's fairly easy and there
there are various examples of isolated
schools in the country that have no special budget but just have for
instance a principal who actually understands how to educate and the the
stuff happens the principal gets the parents involved realizing it's a
sociological process now not just see
what's once you get away from this transfer of knowledge fluid right this
is the factory model of Education that's so wrong if it were that easy we'd have
no educational problems because we could set it up as something like an
industrial process and that's and that's basically the way certain companies have
approached user interface you know where the idea is that this is an industrial process and we'll train people in order
to use the user interface of this computer the other way of looking at it is people have all of these human
dimensions and they have an interest in environments and exploring environments
and so forth and you can get them to learn this stuff painlessly while
they're using it and I think that approach works a lot better but as in
all things that we are not lockstep
fascist type control from above you have
to set up an environment that's actually well fitted to the human in order for it to work so it's one of those paradoxes
that education is a simple problem with
a very complicated solution because most of us don't really want to go through
what you have to go through to do it any more than most parents really want their
kids you know the parents sat down and thought about it for more than two
seconds they would not stick their kids in front of the television
even before the something like the Gulf crisis I I grew up during World War Two
in Australia where we actually had air raids and an occasional japanese suicide
sub and I had no trauma from the war because my parents told me about the war
on a need-to-know basis which was very very tiny for a 3-4 three-year-old
four-year-old that I was back then they listen to the radio after
I went to bed I realized now and stuff and so I had no trout you know there's
no way a three or four year old can handle something like that so why the heck let three and four-year-olds watch
unrestricted television and once you realize that that is a crime against
humanity to do that and think of all the other things that they were watching before the Gulf crisis television has no
secrets and childhood is all about that there's a difference between children
and adults so you know turning off the
television is the first value judgment that any parent could make to help their
child learn most parents won't take the effort to do that and take the
consequence of what it means to turn the television off that's why I say most
Americans actually don't care about education they want it to be a process
done by somebody else but they don't want to take the attack the responsibility for it it's a shame but
it sure is the way it is I saw another hand someone yeah
well I think that the there will always
be when you have networks you always have servers and I think to me what a
server is is a piece of computing
hardware that is designed to maximize certain kinds of processes that are not easy to distribute I can't imagine
anything like a 3090 continuing on but
of course they will I mean the customers have to stop buying these things before
they go away and the customers aren't sophisticated enough you have to figure
it out but I think that just as there will always be a COBOL they'll always be
things like the 30 90 but I think that the what we'll start seeing more and
more of our special-purpose servers and one of the places where you can beat the
heck out of a mainframe is in doing information retrieval even today it's
possible to buy massively parallel computing hardware for like a million
dollars like a connection machine from thinking machines that has 64,000
processors and can not only do conventional information retrieval tasks
much faster but can do information retrieval tasks that are not conceivable
on any kind of stream processing system
and I think we want to think very very much about what the server environment
actually means when we put networks in and to realize that every time we have
to you know the question is what are we send back and forth every time we send
things like characters back and forth we lose every time we can send something
like an object back and forth we win the problem with things like characters and
simple raw data is that we have to have very complicated protocols for dealing
with them at either end and what we
really want is to finesse solutions so just an example of a nice finesse
solution is where you have a laser printer one of the things you can do is
try and make a protocol for it right and all of a sudden you start thinking
about all the different kinds of page layouts this is the way people do
standards let's see all the different things people are going to use you trying to numerate them and then you
make that to standard that can't work there are too many degrees of freedom on
the page and so what the the Adobe people Adobe used to be a Xerox PARC lab
and what we did at Xerox PARC was to say
no we won't do that we'll just will let people send programs we'll just then the
program will interpret the program so we'll know how to execute about 80
operators and those 80 operators will span the degrees of freedom of the page
completely that's what PostScript is so the only
protocol is that you just send a program and at the other end there's an
information buffer that PostScript tries to build the image from that program and
if you think about the simplicity of the PostScript manual is about 20 pages long
the difference between that and trying to come up with a formatted way of doing
things is incredible this is true for things like cd-roms and and so forth
most of the standards that people think about on CD ROMs and video discs are
really stupid because they worry about things like pixels you know who knows
what display we're gonna have 10 years from now the CD ROMs are still gonna be around so why should there be a cd-rom
standard that knows anything about the pixelation of current displays so all of
those things I think say what we want to do is go to representation independent
ways of conveying information and that's what objects are all about getting away
from having to know anything about how things are represented or going to be
represented but going to things that are more like saying I want you to do this
and I want you to do that and the object figures out how to do it you can change
the representation yes
laughs LCD screen
yeah but we had a chance you know virtual reality was actually invented by
ivan sutherland who was the guy who invented computer graphics in the early
60s and when I was at Utah Ivan
Sutherland brought the first head-mounted display system in fact the
first real-time 3d graphics was done for I hadn't been a display that you want
you warned your head and the computer knew where you were looking and could change it so back in the late 60s we had
quite a bit of experience with that and I think there's quite a bit to be said
for one of the the alternative there
were actually three designs for that Dynabook one was the notebook one was a
head man a display and one was a wristwatch the wristwatch was for 2010
when you have as many different network connections as you have plug electric
plugs in the wall so that all you need to carry around with you is something it lets the local computer in the room know
it's you so as you move from one place to another in the physical world you can
imagine your computer agents hopping through the network following you a lot
right that's sort of the I usually don't bother subjecting audiences to that one
because it seems but think about how ubiquitous power outlets aren't even
phone outlets are now so just a short step from that to have all the networks
in the world the transponding all the time with you
I think the virtual I mean the virtual
reality stuff it like multimedia can be
a red herring in the sense that both of
them are much more concerned with display modalities than they are with
content both of them for instance in order to make virtual reality work and
this is something that we had determined this is my opinion but this is something
we had determined very strongly in the late 60s is you absolutely need to have
Tac Toe feedback from the environment putting
yourself in a 3d environment not being able to touch and feel it is almost
impossible if you ever tried one of these things it is almost impossible to
grab things and know that you're grabbing them in all this you need tactile feedback until they do that it's
just the curiosity now the same thing multimedia is a little bit like saying
well now that we have books let's put pictures back into them and we'll put the pictures back in in lots of
different ways we'll put sound in that doesn't count unless there's actually a
rhetoric that allows you to organize this stuff in the meaningful way and
that's what really needs to be looked for you want to in particular you want
to have it in such a way that multimedia doesn't wind up looking like television
because television appeals to one part of your mentality and not another
so it'd be distressing if with all the technology we have on this earth if we
wound up reducing the entire world to the mental state that people had in the
Middle Ages but that is entirely what could happen it's happening right now
television is basically electronic stained glass windows and it has all the
pluses and minuses that go along with that and if that is a pre-scientific
scientist ill ization worldview it's quite good we don't have time to really
go into it here but I think it's one of the biggest issues there's something I when I get to invited to a multimedia
conference I'm usually on the other side not saying make multimedia going away
but I'm saying trying to understand it for God's sakes just like try and do a
user interface you might as well trying to understand how the human mind works a
little bit for yes
yeah well there you know there's this old joke about rocks are to keep out
amateurs so basically most I mean you
can boil down most security mechanisms to how much work do you want the person
to do before they can steal your stuff and the answer is you can you can make
it you can make fairly inexpensive locks
that are an incredible amount of work for somebody to steal your stuff and
that's pretty good I have been for the last five years I've been trying to
convince publishers to publish on an
open network providing we can prove that we can make geometric copying impossible
it's the same thing in the recording industry they don't care about numeric
you know arithmetic progression copying
but what they hate about dat is that you can do it geometrically because you got
a pert you got no noise so you get a perfect copy and the copy can make
perfect copies and the copies can make perfect copies and you don't get any degradation and so you can geometrically
fairly easily overwhelm something without having the degradation in what
they would like is to have something where degradation is introduced so that you can't sell a copy as them as the
real thing right and same kinds of things you have to do when you're doing
network security there's so many different kinds of things that have to be done and all I can say is technically
the problems have been fairly well understood for a couple of decades
none of the processors that desktop
computers use have any security mechanisms whatsoever built into them
right so if we wanted to actually get secure and we wanted to extend the
address space of the network into our machines then we'd have to get rid of all of the machines in
in place of a separate architecture I think that will happen fairly slowly but
it's likely that when the first intimate computers start happening which is about
1995 that they will be built with
complete hardware level protection in them right from the start just from what
they are and how they have to interact continuously with the network yeah one
more question having a hard time saying oh yeah given
that institutions and businesses have to stay in business continuously do you
believe that it's possible for a company to leap one of your evolutionary steps
in other words to go from institutional to intimate without going through the
personal computer stage yeah I don't know I don't I don't have a good I don't
have a good sense of that and we're probably not gonna be able to find out
because the we had one chance for that and that was if Xerox had decided to do
the quote-unquote Macintosh in 1975 that
technology was actually invented before any desktop technology was and then we
would have had a chance to to see what
it was but I think that the because this third one that's coming along is so
Network driven and we're so lacking in the kinds of servers that it needs and
it's going to take five or eight years to get the network's going and where
artificial intelligence is and everything else I think that it's actually going to be a real struggle so
my I believe that by say 1993 to 9 1993
we'll start seeing devices that you put in here that are a combination digital
cellular phone and things like electronic mail and other kinds of be
like these notebook computers but tied into communications and by 1995 we'll
see full function things that have a fair amount of AI and we'll be
retrieving things that the system thinks that will be interesting for you to know
even today you can buy 10 years ago one
of the first agents I did was called News peak and it was the thing that stayed up all night and found you a
custom newspaper based on the kinds of
projects you're working on and so and this year 1990 is the first year
when you can there are two companies that are operated offering a product
like that one is called News edge and one of them is called first news
once for the Mac and once for the IBM PC and these business system where you can
gradually develop a profile of the kinds
of things you're interested in and it will log in to a half a dozen to a dozen different news things continuously for
you and present you with about a you
know the hundred most interesting things that's discovered for you so this is a
very simple agent and all of the agents I think in the next three or four years
will be retrieval type agents because that's an area that just isn't being
handled by the sequel direct method way of doing things and then we'll have to
see a person would be very interesting
to invite here is a guy by the name of dr. Doug Lennon who is doing a system
called psych it's the psych in encyclopedia and what this system is is
a an attempt to model human common sense so he believes that human common sense
can be dealt with with between 200,000 and a million rules this is the projects
been going on for about eight years you know it's like modeling the human genome and his belief is that we can't do real
agents until we have a model of human goals that the agent can relate
everything that you're doing to some understanding of what humans are trying
to do in this world and that you have to have something like common sense
reasoning in order to pull this off so
that project will not mature for another five or six years if indeed it ever does
because it was a very hard problem so I think that's something that can be can
be tracked and what I would do is the company right now is to start doing
retrieval type agents for instance here's one you could do that would be
incredibly useful right now most companies have electronic mail but they
just imitate surface mail right this is not moving the motion picture camera so
what would it be like to have a mail system that not only retrieve
stuff with your word with your name after the word - but what if it
retrieved mail that had your name in it
what if it retrieved mail that was about the kinds of projects that you're
working on what if it retrieved mail that was about projects that were
analogous to the kind of projects they were see an electronic mail system gives you a database that is essentially a
group mind it's the mind of Lily if the mail system is that all good and yet
because we don't think of it as a retrieval system we're not using it and
that's what agents are all about is to take enormous pieces of data and to just
fare it away at it and trying to trying to find relationships that we can't get
at directly so thanks a lot for inviting
me and enjoy the rest of the conquerer [Applause] [Music]