A talk at User Group University (1987)

From Viewpoints Intelligent Archive
Revision as of 19:30, 5 December 2017 by Ohshima (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
Apple has finally decided that it's a product company what they mean by that
is that Apple one of the companies that
Apple sees itself as being like is sony Sony is a company that creates many of
the markets it sells in I think Apple came to that realization not so much
from the Apple two because the Apple two happened in a in a way where was hard
for them to even see that they had created in some sense a business for IBM
to move into but I think that the the
success of desktop publishing caught
their attention where they realized that oh yes Desktop iBM is just setting up a
2,000 person group in Colorado to do
desktop publishing that's about just about a third of the employees at the
Apple has I think now we realize oh yeah now we what we are is a product company
we're going to listen to customers but we also are going to try and come up
with new ideas creating new markets and see if the customers will go along with
us I like that very much of course because the the idea the the big problem
in all marketing theories where you go out and ask the customers what they want
is that customers almost always say oh I want something just like what I have
right now except with these little things fixed we have to listen to that
but you don't get new products from that a couple of months ago I was reading
Toynbee and the discussion was how do
you judge the worth of a civilization is
it by the quality of their art the size of their buildings the social systems
that they have and I think we 100 years
from now we will be judged yay by the complexity of the audio-visual
equipment we need to put on the presentation that's not much of a
picture is it now what I'm gonna try and
do is
instead of talking about future trends because I don't even believe in the word
trends and instead of just telling you
what I think the future is going to be I'd like what I'd like to do is to just
tell you the way I predict the future when people ask me what's going to happen so that you'll be able to do it
and then I won't have to give these talks anymore so four ways to predict
the future well the first way came about in the early days of Xerox PARC when we
had Xerox executives buzzing around like bees after honey saying what's the
future gonna be like what's gonna happen to us what's going to shape us and all
these things I finally got angry at them I said look the best way to predict the future is to invent it you want a future
come up with a vision and then build it and you've successfully produced the
future because you've simply made the future come out the way you want it
that's what research people are for is to invent the future another one very
sobering thing is it takes at least 10 years to go from a working good idea in
a laboratory to get it out in the outside world this is something I didn't
believe in for many years when I was younger because it was so depressing an
idea but in fact there are many instances of the 10-years story and if
it's a really basic technology like transistors it will it may take 20 years
does a basic technology has to displace the existing basic technology in order
to take hold I'm going to come back to that because one of the implications of
the 10-year idea is that it means that most things are going to happen to you
in the next five to eight years or in somebody's research lab right now so if
you want to know what's going to happen all you have to do is go around the country and visit people in their
research labs and see what they're doing and then you'll know the answer I
brought a satchel full of tapes of various things from people's research
labs and I'm going to show you at the end of the talk a lot of these tapes are older
than 10 years old and the reason for
that I'll explain in the last part of the talk another way of predicting the
future is to realize that what we're
about here is the invent the invention
of new kinds of information media there have already been bunches of information
media invented over the last several thousands of years so if you take
printing as an example there are four
phases to how you make money and how you make money pushes these media through
their four phases the first one is Hardware intensive when the new hardware
is invented you get high margins on the hardware the problem is that's the most
transitory phase because hardware is the
easiest thing for people to learn how to build and once you've learned how to
build it you are making it much more of a commodity item the margins go down you
have competition and so forth and that's really what happened to printing in the
early days of printing they made most of
their money actually on paper the margins on paper it wasn't until the
Industrial Revolution that made paper inexpensive that you started seeing
throwaway media throwaway media like the
novel now the novel have been around for several hundred years as an isolated art
form but it didn't take hold until paper became cheap and we had to go into the
second phase which is the software phase we all know that software is a lousy
business to be in alright because it's inherently a
cottage industry no publishing house has ever been able to sustain itself on
having a captive stable of authors they've always had to go out and most
people in publishing who have been smart going to distribution the distribution
is the first aspect of the third phase which I call service by service I don't
mean fixing things so much but in the
of going from these things which are basically selling items to selling
service you address much more closely this quaint idea that there's actually a
human being at the other end of the transaction and the business that we're
in is inescapably a service business even though most companies that are
selling computer products right now would think of themselves as either hardware manufacturers or software
designers but the urge sort of the the
medium the the forcing function of
networking especially forces us to become a service business and then the
one where you really make the money so if you can hang in there is the fourth
phase which is way of life a way of life i define as anything that
is only noticeable when it's not there so for instance if you walk into ten
years ago if you walked into somebody's office and you didn't see a telephone you would be really surprised nowadays
if you walk into a person's office and you don't see a telephone you would say
oh are you wearing yours now because your phone wired into your t-shirt we're
big on t-shirts down here now so let's
test this out here let's I like everybody in the room to hold up their
personal computer where is it okay some
watches a little calculator okay so I
rest my case I doubt if there's an Apple computer or anybody else's computer in this room and
that means that oh there's one over there well ignore that one
it's what physicists call experimental error I think the point is well made
that there's something about the computer right now that makes it not
it's it's basically not important enough for you to have one with you right now
something about it that causes you to leave it behind one of the things about
it is it's too big another thing about it is that it
doesn't do enough when you make it small so the real personal computer at least
the when I envisioned the phrase personal computer many years ago I
thought of something that you would never leave behind one of the tests at
Xerox PARC we came up with four is it a personal computer which would you be
willing to do anything so mundane with your personal computer as to put your
grocery list on it if you're willing to carry it into you at the supermarket and
then out again with two bags of groceries in other words a personal
computer is something first it is so portable you can carry something else
too because its own not the only thing you're gonna carry and secondly 0.5
herniation per block is not portability so this way of life Bayes is very
important it's the one that I have been
interested in for 20 years of what do you have to do to the damn thing to get
it to be a way of life for people and finally human trait amplification this
is sort of how you predict the big winners this is something I first
pointed out to Xerox many years ago least like to try and do market surveys
on things and I have to tell you what Xerox story that Xerox in the 50s was a
very small company in Rochester and they had developed this obscure technology
invented by Chester Carlson into a machine called the 914 and they didn't
have enough money to build factories for it so they took it to IBM I think it was
in 1956 this is all described by the way four
people are interested in this kind of lore in John des hours book which is
called my years with Xerox the billions nobody wanted this is sort of a perfect
parable of life in the twentieth century so that's our story goes on to say that
IBM Xerox took this to IBM and they said here take this build factories build it
and sell it all we want is a small royalty an IBM scratched its collective
head and said well we don't know about that so they went out and got hired a
consulting firm Arthur D little is Mike
hammer his definition of a consultant is a person who knows a hundred ways to
make love but doesn't have a girlfriend
my definition is it's a person who steals from the smart cells to the
stupid sort of a sort of an inverse
Robin Hood but anyway after 18 months of
carefully doing market surveys and other things consulting firms do Arthur D
little came back to IBM and said nope it'll never sell there's too great a
disparity between it and the a be dick mimeograph machine and nobody wants
plain paper copying and there isn't enough copy volume to sustain copying
and all that stuff and so IBM turned it down and that really pissed Joe Wilson
off who is the head of Haloid corporation in those days and they bunch
of the executives got together they used their life insurance as security this is
the same way controlled data got started by the way out of Remington Rand raised
the money for their first factories and that was about a hundred and twenty
billion dollars ago it's a perfect example of when the technology is
actually new it's almost impossible to go out and do a market survey and find
that there is a need for it because as we'll see in a minute really great stuff
creates a need that's what all the so I
pointed out to Xerox that another way of predicting these things was simply to
look at traits that human beings have
that if the trait was taken away we would call a person no longer human like
communication humans are not natural Hermits we have to communicate we don't
do it just because it's useful we do it because we need to so whenever you make
a communication amplifier and no matter
sort of case law back then was even if it costs ten times as much as the
technology is trying to displace if it's really a better amplifier for
communication it's going to drive it out perfect example is the railroads in the
early 1900's thought they were improving their business by making better track
and faster locomotives they were looking out like this along the tracks what they didn't
see was their competition starting to buzz overhead because they conclusively
proved that there is no way that aircraft transportation could be as
cheap as trains and they were absolutely right and totally irrelevant because
people were quite happy to pay five six seven ten times as much for being able
to get there much faster and even getting their Freight there much faster
so here's another one I pointed out to
Warner's in Atari that fantasy amplifiers are big also and the reason
for that is that even when we think we're in reality we're dwelling in a
fantasy or in a hallucination that is partly a product of our culture and our
upbringing as well as what our sentence our senses bring to us and a third one
is control one of the biggest traumatic
shocks we have as a child growing up is when we start dimly sensing that the
universe isn't in our control I remember that the guy first realized that when I
was about three years old and I've never gotten over it since but in fact this
this idea of trying to be able to control things that amplifiers that
allow us to control various things and so forth it very important also and
simply you can tell very often whether a new product is going to make it if it
really does amplify one of these or other human traits you don't have to go
out and ask people they can't tell you that just as Sony did not market tests
when they when they went out and tried to do the Walkman nobody wanted to
walkman until Sony built 5000 of them
and gave them to people if you can give something to people and they can hold it
in their hand then they know whether they like it or not that's one of the
reasons the Japanese have been so successful is that they know the way to do market analysis is not to go out and
ask the questions before you do the product what you do is you do the product and then see what people think
about it because then they actually have an aid to their
imagination let me just follow up on
three ways to invent something this is
the most popular way take whatever's around shove it together and ask what
you've got that's called brainstorming something
these these productivity books think is important but in fact it's the worst way
to do anything because people are usually lousy at thinking on their feet
and they're terrible at thinking inside of a room people need sand under their
feet to do good thoughts or maybe a shower or something better way is the
problem-solving method find a need and fill up but the one that really wins is
to make something that creates a problem only if can solve that's exactly what
the Xerox machine was nobody had a copying problem until the Xerox 914 came
along nobody had a calculation problem
until the HP 35 came along the HP 35 and
calculators were predicted against for the simple reason that nobody did
calculations up till then there are desk calculators around and people didn't
realize that the that the simple access
to calculation gratuitously was an important part of it and of
course the personal computer itself is an example of that nobody had a personal
computing problem until pcs came along and now we all do possibly very possibly
okay I'll just give you a few more slogans and homilies here
another one is about literacy my
definition of litter literacy is that literacy is fluency not familiarity I
usually show this slides to the educators because educators think if
they make a child familiar with the computer then the child is computer
literate but in fact it's not true I believe literacy has three parts
there's an access part that means
somebody else makes up the material and you have the skills to get to whatever
that material is we call that reading in the print medium and the Macintosh is a
good example of how to solve the access problem to operating systems and
applications second area we don't do so
well at which is called creation fluency and the print medium that's writing and
writing is an example of something where we expect somebody who's literate to be
able to write a 2-page two pages of something advancing a point of view and
substantiating it in order to convince somebody and this two-page thing is very
important because most people have a very difficult time dealing in large
number of pages of anything even for people who can write so it's very
important for English through the use of metaphor to be able to express fairly
complex ideas in a couple of pages programming languages are terrible for
that almost all programming languages to do anything interesting in takes you far
beyond several pages and one of the things we found out at Xerox PARC years ago is that the if it's a
trade-off between an easy programming language like say basic or logo that
requires tens and twenties and hundreds of pages of program to do something and
a very abstract programming language such as small talk which can do many
interesting things in a couple of pages then it's better to teach the abstract
language to end users non-programmers
than it is to try and teach the so-called easy programming because in fact having the programs be
very small allows the person to use their visual field as an augmentation to
their short-term memory and that is the critical thing in end-user programming
and finally the third fluency I call
shanwa fluency and by that I mean that
for anything you can be literate in has a literature and it's a very nature of
literature that the literature is in different styles so it's not just enough
to be able to read and write but you should be able to access other people's
ideas by being able to absorb other styles of talking about things and again
this is something that will be more and more important in the future as
networking allows us access to more and more potentially useful tools that right
now we can't use because we can't sit
down and absorb a different style and here we're talking about more than just
to use their interface having a uniform user interface like on the macintosh is
important but it's not enough
and one final set of ideas
McLuhan pointed out mentioned that
Whitehead had called the greatest invention of the 19th century being
invention itself and Whitehead substantiated that by pointing out that
the number of crackpot inventions in the
Patent Office in England had gone up by more than a factor of 30 starting in the
early 1800s that everybody was inventing people are inventing automatic hat
tipping of devices and strangely shaped bicycle seats and every other kind of
thing amazing anything that could be thought of in some things we'd never
think of were patented a huge explosion and the important thing about that was
that the number of good things went up by a factor of three that factor of
three coupled with the Industrial Revolution changed the way we do things
McLuhan said if the greatest invention of the 19th century was invented there
was invention then the 20th century is the century in which change changed our
whole notion of what we meant by change over the last 10,000 years has changed
in the 20th century and so here's an example this is a list compiled by a Jim
cook who's a corporate thinker on the East Coast is kind of interesting so the
mission of the Industrial Age was to amplify the body and the mission of the Information Age is to amplify the mind
he says structure of the Industrial Age of the hierarchy the Information Age is
naturally a network the output products and services but the new output is ideas
and decision decisions advantage economy of scale now it's timeliness and
leverage old gold efficiency new goal effectiveness manages activity now we
manage communication control with centralized now it's going to be
decentralized medium is immeasurables new medium is
concepts so forth evaluation as in reports now we
believe sensitivity is to organizational
structures now it's the talent cohesion is through discipline nucleation is
through commitment risk return incremental now it's volatile the
leaders used to be dogmatic now they're charismatic and the heroes used to be
specialists but now they're going to be generalists I subscribe to most of these
things that this way of looking at
things says that it's more than just incremental changes that are going on in
this century it's a whole different way of thinking about doing things
well it's just that the basically
there's a guy by the name of Lewis Mumford used to write interesting books
about how civil how civilizations organize themselves to accomplish tasks
one of the the terms he made up was
called that civilizations he called the mega machine he was very interested in
what the mega machine was set up to do he thought of offices and businesses as
being communications devices when you set up a business it's it's not it's
ostensibly to sell goods and services perhaps what it really is as a communications device he's very
interested in what what the business was
as a system and what you can do with a hierarchical organization what you can't
do how flexible is a hierarchical organization say to corruption at the
top versus how flexible is a network
organized organization to corruption anywhere he was interested in the
difference between simple mechanical systems as applied to human sociology
versus more biological systems systems that are more tissue-like and so I'm
just putting those out there as a as a way of pointing out that the the very
structures in which organizations use to
communicate and move ahead is changing is partly changing simply because the
information the way we can move information around is different one of the ways of thinking about how humans
have gotten ahead over the last 10,000
years could be thought of is each stage of civilization has found a way of
making an abstraction into an object 10,000 years ago words were not objects
oral society words are not objects they can't be put down in a way so that you
can stand back from them people marveled
at st. Augustine because he was one of
the few people in his time that could read without reading aloud and moving
his lips when people read meta evil manuscripts
even up to the 1400s they read aloud and
they read slowly because it was they were essentially an oral culture the
writing system that they had was a way of capturing oral culture not changing
the culture McLuhan tells a really interesting story fascinating story
about a tribe in Africa that had just recently achieved literacy via some
missionaries that had helped them make
up a syllabary like Hindi but a syllabary of some 60 or so signs that
stood for the syllables in their language and several people in each
village had learned to read this syllabary and the story that McLuhan
tells is that when a letter came into the village it would go to the person
who could read it to the family who would received the letter and if the
letter wasn't was a personal letter the person who read the letter to the family
would put his fingers in his ears as he read the letter so he couldn't hear now
think about that think about that because that's what I'm
trying to say that McLuhan said I don't know who discovered the wall who
discovered water but it wasn't a fish in
other words it is the stuff that we're embedded in that we take for granted
that shapes us there's no reason why human beings have to be the way they are
in any particular culture there are many different ways of doing things and the
thing that we're at least aware of is the way our culture does things so people like Mumford and McLuhan are
tremendously useful in that they point out how how different are the ways that
people have solved their problems and how different it is for instance to read
fluently in a print medium in which none
of the emotion is expressed in the form of the text the illuminated manuscripts
were there as a kind of a stage setting to bring forth the emotion that was
essentially a poetic expression of things
prose is something that really became possible with the printing press and our
reaction the way we read prose is totally different the way we get emotion
from prose is totally different from the way we do when we were in oral culture
so in fact this is a good good place for this another one of these triads solving
the problem is all is often futile it definitely is in futurism changing the
problem is what the books tell you to do but in fact changing the context is
almost always the most powerful the reason for this is that logic is basically a weak method we're not any
smarter than we were thousands of years ago and if we were put back take any
person in this room and put them back in 60 AD and try and get them to multiply
two numbers using Roman numerals we would see just how smart we looked the
representation systems are the things that are powerful our ability to do
logical thinking has changed about the same so this was a slogan that we liked
at Xerox PARC so I made up call point of view is worth ad IQ points and that
slogan was made up when we were developing the user interface that led to the Macintosh what we're trying to do
is instead of trying to change the human
species so that they could run ms-dos
or for the ms-dos of the day thank
goodness it wasn't around when Park was thriving but they were equally bad
systems back then what we asked was what is the context that can make people look
smart on a computer what is the one that allows them to use
the logical operations that they've learned all their life and well that
will translate into actually controlling the computer and that's what led to the
familiar Macintosh user interface
so
so what caught business data processing off-guard in the past decade
I think desktop micros certainly
surprised them they were out buying IBM mainframes spreadsheets surprised them
Universal bitmap graphics such as on the
Macintosh that is graphics without any modes where you don't have to say ahead
of time when you're designing the machine what can go on the screen
pointing window interfaces like the Macintosh desktop publishing local area
networking all of those things were a big surprise over the last ten years and
the business data processing people are still scrambling to try and figure out
what it all means they want to play it safe but they suspect now that sort of like a
Christian Scientist with appendicitis you know it's sort of one of those
ultimate conflicts that you don't quite know what to do I'm not going to talk
about I just want to go through a list that obvious changes the next five years
I'm not even going to talk about many of the changes in the next few years you can see by going over the Apple World
Expo multi megabyte Ram faster CPUs for
user pervasive networking three-pound computers application kits and
generators higher-level language programming very high level language
expert system prototyping fast deductive reasoning and some more attempt at
standards particularly at the network level those are all obvious things I
don't think they deserve too much comment here
so less obvious changes in the next five years now machines we did it park first
machine we did it park if you were to
bench market today was about as fast as
the Mac two which is wonderful wonderful
at the Mac to all of us that some of our
the old systems we used to do like small talk run beautifully on the Mac too and
we just love it and but what's been happening in the industry is something
is that's pretty good pretty annoying and it has to do a lot with the physical
form of RAM memory and that is the band the bandwidth per bit in memories has
been going down what do I mean by that well I mean that the you know you in
order to get bandwidth in memory you have to configure the memory you know in
a certain width if you make a 16-bit wide memory and you have to buy a one
megabit chip then 16 of those one megabyte megabit chips will give you 2
megabytes of memory 16 bits wide that's
not very fast these days so if you want
to go faster you can go to a 32-bit wide memory that says you have to buy 16 more
of those chips that get says that in order to have a single 32-bit wide
memory you have to have 4 megabytes of memory and it's still only 32 bits wide
what I mean because the memories get longer you have to configure enormous
memories in order to have a reasonable bandwidth per bit in memories and that's
not so good and what what that means is
that the MIPS per megabyte ratio has
been going down so the Macintosh for
instance has about is about a 1 MIT processor and it has about 1 megabyte of
storage and that we think is pretty bad because the machines we built at Parc
had three MIPS let's see let me work it
out it had about a myth per hundred
kilobytes so every hundred kilobytes of memory we built had a park we had 1
million instruction per second of processor and was we built a very big
machine we actually built in 1978 a machine that had 40 million instructions
per second for doing all that stock and no that is really important so today's
micros in my opinion are underpowered in processor speed for the amount of memory
they have on them that's going to change this is going to be this change of
having many many more millions of instructions per second for the size
memory you have on your machine is going to be a big surprise to people over the next few years would not surprise me at
all before 1990 to have 75 or even 100
nips on a desktop no
well show you why you need it say I told
you these weren't so obvious part of the
part of the way we're going to get that is we're going to have many more CPUs per user I'll give you an example there
are lots of different ways of trading off these two but in a good example of a
new kind of architecture is the system called the connection machine at
Thinking Machines Corporation ok this machine is is a machine that has 65,000
processors each processor is a one myth processor and each processor is attached
to four kilobytes of memory ok so this is a one MIT processor for four
kilobytes of memory and that's the machine so in fact there is no separate
Ram we have a one myth processor on four
kilobytes of memory you have 65,000 of them and that is now what can you do
with the machine like that well an example is inductive reasoning I'll give
you two examples that you find it hard
pressed to do on any other machine one of them is doing information retrieval
over large document spaces one of the hardest things to do is to get a set of
indexes that represent some theory you
have about a set of documents ok because
people have a hard time assigning keywords and full-text retrieval tends
to retrieve too much if you've ever used a good full-text system like Nexus so
the process on this machine is you start off like Nexus you come up with a couple
of keywords and it retrieves a bunch of documents and you look at these
documents and you see you score them you say oh this is just the kind of document
I want this one is way off and then what
it does is it takes all the documents you like and intersects every
non-trivial word in the document with
each other and when it comes back with is another set of keywords gotten out of
this massive intersection does this intersection in about one
second and then what it does it then it
retrieves from the entire database all the documents that matched those guys and you get to score again okay and
what's amazing about this and there are several papers you can get from Thinking
Machines Corporation if this intrigues you is that after you do this two or
three times the keywords stabilize and
you get a keyword set that will retrieve just the kinds of documents that you want and almost always exclude the
documents you don't want just a couple of passes and this whole process to
establish that it takes just a few minutes now this is just simply isn't
possible to do I'll give you one more example one of the problems in expert
systems is that expert systems know a
lot about a narrow domain but because
they're basically deductive there has to be a rule in the expert system to cover
every case so let's take a case where we're describing some symptoms to a
medical system and at some point the system needs to know the common-sense
fact that women that men don't get pregnant and somebody forgot to put that
in there and the typical expert system will just bolt that's the end of it
because the problem is is that the there we figure there are somewhere between
200,000 and 500,000 common-sense rules that we all learn over the first 20
years of our life that we used from getting from one expert island to another and it's very hard to codify
those and people are trying to codify them now because that's what's one of
the things that's holding back expert systems now in the connection machine
when it comes up to something like that it doesn't stop back it doesn't even
have a common-sense rule base what it actually looks at is the last six months
of reuters dispatches
they just got the last six months of Reuters fashion this is about 15
gigabytes of information and in a second
literally a second it is able to come up with a 99% probability that men don't
get pregnant because it has not found any mention of that in this general
corpus that has to do with human life and so it uses this inductive assumption
to carry on in the the the deductive
process of trying to figure out what disease has so there are many others I'm
gonna mention some of the the graphical needs for densely parallel processors in
a bit but I guess the point is that the
almost everything that has to do with the economics of building computers has
to do with what it costs you to package the components and most of the things
are going to surprise us in the next five years are going to be when certain
packaging thresholds change abruptly and surprisingly one of them as we see is
going to be 3d graphics 3d graphics that will in done in real-time has been
incredibly expensive for the last twenty
years or so I was in the original group at Utah that invented 3d graphics and
none of us had any idea it was going to take so long but it's completely
dictated by the economics of building systems that have 2,000 multipliers in
them and that's what it takes to do 3d graphics I'll get to that in a minute
another example is what you might call
ultra high level language end-user programming and tailoring this is this
again is sort of a prediction and I
believe it's going to come true not because it's easy but because it almost
has to become true if we're going to have a hundred million computers in the
United States instead of something like 20 million that there's no way
applications packages are going to be able to anticipate all of the needs
without getting much more patchwork quilt II than they are and there has to
be some way for users to tailor the end results of the
system and even build some from scratch it's not going to be in any programming
language that we know now common sense
database I've already talked about a bit this means the difference between
specific ways experts do things and those 200,000 facts that everybody knows
but are harder to write down there's a wonderful research project in Texas at
MCC right now done by Doug Glennon that
is using an ingenious set of techniques for coming up with a common sense
database interesting thing about how big is this database he's done enough
already he's been doing it for four years now he's done enough already to
establish that the database is somewhere between 8 megabytes and 100 megabytes in
size for all of the common-sense facts that people in our civilization know if
you think about it sounds large but in fact it's only 1/5 of a compact disk
only 1/5 of a compact this to have this and you can think of this common sense
database in many ways as being the quick-draw of expert systems in the
future it's that set of stuff you want to have an e ROM that every expert
system builder can rely on being there in a general inference engine in fact
this next point speaks to that fact the
next several points do that it's going to be the integration of expert systems
that will make expert systems as useful as the tools that you now enjoy on the
Macintosh expert systems right now are
completely isolated and they've partly isolated just because they're done in
different languages but a lot of it is that they're isolated that there's no way for the inferencing mechanism to
move from one place to the other another non-obvious technology this is a
technology that has been chugging along pretty nicely for 10 years but nobody
has noticed it in the micro world is what's called explanation generation
and a explanation generator is a program
that can look at a data structure knowing what the data structures are
generally attempting to represent and generate perfect English prose at any
level of detail about what that data structure is this is the exact opposite
of a system that understands English input which is a very hard problem but
systems that will be able to generate beautiful English output from
representational structures will burst
on the scene in the next five years there'll be an overnight sensation because it's only they've been worked on
for the last 20 years or so just a technic technology that there's no hint
of in the micro world right now I mentioned inductive reasoning another
less obvious change is that I believe that user interfaces are going to change
completely over the next five and ten years and the reason for that change is
going to be networking one of things you
probably noticed when you put a hard disk on the Macintosh is that the nice
browsing techniques of the finder don't work so well anymore you have to make
things real small and some people go back to lists and stuff and all of a
sudden being able to browse a hundred items changes completely when you have
to browse upwards of 2,000 items and this is something we found out at Xerox
PARC in the mid 70s we started thinking about what that meant especially when
you get hooked up to networks that are also hooked up to networks and you have
hundreds of thousands of resources and
we thought that was rather like what happens when you wander into the Library of Congress in Washington and you
discover to your dismay that the card catalog room is much larger than your
hometown library and what do you do at the Library of
Congress well the first thing you do is find a person who is an expert in where
things are that person will act as your agent during the time you're in the
Library of Congress the person does not know the answer to your question but the
person is an expert at the resources and pathways for finding the answer so this
notion of agent is an old the idea in computer science it goes back actually
to the early 50s McCarthy I mean late 50s McCarthy thought it up but there
hasn't been enough pressure from the technology to develop agent based
interfaces until now and finally one of
the things that we just absolutely need a factor of 100 more cycles for is what
I called remember we talked about the one of the messages of the Mac is
Universal bitmap graphics that means Universal just means you don't have to
know ahead of time of what's going to be shown on the screen and the big change
in the next five and 10 years is going to be extending that to 3d real-time the
animated graphics and simulation now let me give you some examples I'm sort of
going to be popping back and forth here
they're a clicker up here yes there is
it's always nice when every Apple celebrates their invention of the
personal computer I always like to show the actual first personal computer
because this is the machine called the
link and it was invented at Lincoln labs by Wesley Clark in 1962 near about 3000
of these built in the in the 60s and a lot of their architecture were adopted
by deck when deck invented the mini is
really kind of a shame because I sort of believe when deck look at this machine
they thought oh what a great idea but people don't understand displays so
we'll take that off and make it an option
and these tapes that look like deck tapes are actually invented for this
machine they actually held the pages of the virtual memory this first personal
computer had so it had a swapping virtual memory run off these tapes and
Dex said oh but people don't understand virtual memory we've got it we'll put
files on these tapes so they took this technology and made them back into
something like computer tapes so this was a great idea that was way ahead of
its time now this is the invention of computer
graphics Ivan Sutherlands sketch pad and sketch pad was not just a graphics
system but in fact a it was a system
that could do general-purpose simulations the graphics had meaning and in fact the system was called sketch pad
because it would actually complete the drawings for you the drawings you just
put it in roughly and then you told the system what the drawings should be and
sketchpad would complete the drawings all this in the very first graphic
system and I'll show you an example of that
some over 1962 you see the familiar rubberband technique Ivan and someone
actually invented it you something as easy for links to the edges and says I
want these to be all mutually perpendicular and sketchpad figures out how to do that so you're seeing a real
time expert system in action notice gets bad as the first system to ever have a
window here's another example
draws a couple of lines he points to them he says make them parallel and
mutually perpendicular to the edge and sketchpad straightens up the drawing a
sketch pad is also the first non procedural programming system and it's
far far more capable than visicalc
or Lotus 1-2-3 because it can solve nonlinear problems in many unknowns so
now he's making the guidelines invisible
you'll see this display jump around a lot because the tx2 computer that this
is running on is many times larger than this room and it's putting up every dot
on the screen by brute force that goes
the last computer to ever have its own roof so here's another example he wants
to draw a rivet he uses that cross place there as the center for the arc and now
by simply pointing to the edges and telling sketch pad to straighten it up
that will drag the cross pieces which will Center the rivet and sketchpad
completes the drawing
reason I'm showing you this stuff is because the Macintosh to here he's
showing it no matter how you distort it the same general rules or follow
Macintosh to is the first micro computer fast enough to be able to do this kind
of computer graphics and I predict that we'll see sketchpad what are called
constraint oriented graphics systems popping up in the next several years
have to have enormous computing power to
do this in real time now he's showing
that anything that you can make a drawing of you can get instances of the
instances can be dynamically again you see the contrast here between this and
Mac draw which doesn't allow you to dynamically drag complex things it makes
up a simplification of the complex thing
for dragging purposes and here are some more instances of the rivet the reason I
call them instances and not copies you'll see in a minute because he
doesn't like the cross piece on the rivet so he goes back to the master
drawing and makes the cross pieces transparent and we now see that the
rivet has the rivets have felt that change dynamically so sketchpad is also
the first true object-oriented programming system and this again we
will see on the Mac 2 in the next few years I asked Ivan once I said Ivan how could
you possibly have done the world's first graphics system the first non procedural
programming language and the first object-oriented programming system all
in one year all by yourself and he said well I didn't know it was hard
so this year is the 25th anniversary of
this system this is the 25th anniversary of computer graphics this year okay
yes here's another something that we're
going to see in the next few years I'm moving up in time this is Douglas
Engelbart in this is about 1968 or so
and Doug Lingle bar was the guy who invented hypertext systems and he also
invented the mouse as you see black on
white graphics view specs for windows
and he had this interesting idea that people shouldn't have to hunched over
their desk so the way you ran this system was this was a lap board that
went over your office chair so you could tilt put your feet up on your desk like
you're used to or scooted around the room it's an idea whose time is yet not yet
come in the microcomputer industry here's an example of that system what
I'm showing you here was taken live from the demonstration of the system done in
1968 so this is 19 years old and
something that I predict will be able to write you as an intellectual worker were
supplied with a computer display backed
up by a computer that was alive for you all day and was instantly responsible responsive
Anthony responsive every action you have how much value could you drive from that
well this basically characterizes what we've been pursuing for many years in
what we call the Augmented human intellect Research Center at Stanford Research Institute
so look what else we can do in here I've got this file is structured if I want to
see what's in there I can walk down
Cinci return but there's another thing I
can do does remove dice that I have here
so here I'm afraid I'll need different pictures of you so here's what I do with
a picture drawing capability here's this light map if I start from work 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 but if
I want to I can say the library what am I supposed to pick up there I can just
point to that you know I see overdue books and all while there was a
statement there with that name wanna go back what if I once my supposed to pick
up the drugstore mmm-hmm I see you've everything all right oh I've already
seen that J that's too much anyway so we
have this feature the structuring our material hierarchically being able to
move around it very well when we get a hierarchy
such as I can show you here
I can do things if I want to just say I'd like to interchange the Protestant
camp materials bingo know if I care to look
interchanging them very quickly cans are going to get interchanged with products
they do it and all gets me numbered so
I've ways of studying over making different views moving around going to
specified points okay well there's lots
more but you might have been surprised by the speed of that system all right
I've got a I'm sorry I have to show you a couple more things I have at the lab
of the love this this is one of the great inspirations of the stuff you did at the same time I've seen that I have a
repertoire of different entities my character knock off a character replace
the character make that P so I have
entities of all sorts that I can say I want to do operations on and this basic
structure that I can move over and study and get about very quickly so that is
the essence No
that's the essence of the tool we have a lot of details your character if I hit W
it'll say delete word the arrow moves back and forth to give me feedback my
tracking spot changes that gives me feedback now tells me since it's an
arrow that it's armed I can do something we get a lot of feedback let me restore
if you like this to show you this is more normally the way we work with
feedback up here okay now the thing to notice here is this is actually a
multiple multiple window system and each of the views had a view specs
he's going to be talking about those the view specs are much more than just that
mucis specs to control the level of detail he sees how things are presented
whether they're presented in an outline form or so forth it's something that has
never been done on a micro up till now I'm shown not working that's an echo
register that normally gives you the last six characters that you use the
last six characters and let shifts continuously so you can look up to any time and see what have I just struck and
that's for a good feedback here are characters that show me the different
viewing parameters few specs single
strokes I change those parameters now
notice here that the theory of interaction here is different ankle Bart
understood very early that the thing you don't want to do is go moving constantly
back and forth between the pointing device and the keyboard and so there's
five finger keyboard he has for the left hand allows him when he moves his hands
out to the outside he can do anything with a five finger keyboard that you can do with the regular keyboard just a
little bit slower and so when he's zooming through the system he'll just keep his hands constantly in
navigation mode it's only when he's typing paragraphs that he uses the
regular keyboard in the center it's a more efficient way of doing things than
we do things today and each of those
means something to me and they're being larger particular times tells me I can
have very quick abbreviation for changing the view so I can say I'd like
to go to produce but I'd like to go to
produce they get big I'd like to say one branch only and let me look just that
little can I see it oh I can say I'd
like to see one line only I can see so
these ways I move around the way I get feedback up here the way I use both
hands to coordinate to tell the computer what command and what's short literals I
want all perfectly designed to go together to make the repertoire I'd like
to link to bill and I don't know what his terminal is right now
so I have to ask for somebody connect me to him audibly audio so bro will you
come in through this intercom oh I I need to know what terminal you're on
though 13 okay I'd like to have him see
my text and so the special thing if I label 13 we'll switch switch over so I
just display he sees my text so I'll execute it and sure enough it does but
what's that running around well he's looking at my text he'd like to have
something to say about it so we put on a marker a tracking spot that he controls
so he's sitting at Menlo Park looking at this text and he can point to it but
we've carefully reserved from either right to control and operate on this somebody so what's going on here is
teleconferencing because this system is organized angle Bart is in San Francisco
and the rest of his group is in Menlo
Park but they're sharing the same screen and every time a person joins the
conferencing system they get a new cursor so that they can point to things
too now the system becomes a shared blackboard it's more powerful in yours
but we can have an argument yeah
so alright so in case you haven't been
listening bill we've been back to lots of examples and setting up in
collaboration here so that we can go on into information retrieval and we set up
now audio coupling and we're both looking at the same display and that'd be very handy to work we could talk to
each other in point and maybe later I could hand you the chalk on this
blackboard like saying here you control it but let's status mode now and add another feature that hardware-wise is
available to the kind of display we have I'd like to see you while I'm working on
it so before I can do that I have to set
up my display in a certain way
set it up so I see it over like that that leaves a corner up there and I say
now computer do the automatic switching that will bring in a camera picture from
the camera mounted on his console such as the camera mounted on - I built
that's great now we're connected audio you can see my work you can point out it
and I can see your face and we can talk so let's do some collaborate I had this
immensely intuitive feeling that humans are going to be able to drive a great deal of capability from this and along
without it coming with very real images of my mind is sitting at a display
console interacting with the computers seeing all sorts of strange symbology
coming up that we could invent and develop to facilitate our thinking since
we would no longer be limited to working with paper and laborious means like that
and other people could be sitting at similar consoles tied to the machine and
you could be collaborating and brand new exciting ways and that you could be just
doing all sorts of things to control a computer so I've enough engineering background even though I didn't at that
time know how the innards of a computer work but I certainly knew that if they can read a card they can sense keys and
any other action you want to do and if they could drive a printer or a card punch I knew plenty electronics to know
that they could put whatever you want or not display okay well that was one of
the great pioneers and a lot of what he predicted is going to come to pass
because what he predicted was what we need and the difference between these
guys in the 60s and much of what's been going on in the micro industry is that
these guys in the 60s seem to have these complete visions so Ivan Sutherland
didn't just invent computer graphics he realized that the graphics wanted to simulate something with meaning and
Engelbart didn't just invent the mouse and a few things he realized that the
whole idea was that for people not just one person but for multiple people to be
able to work together in a shared way now this is the first personal computer
I designed it's called the Flex machine in 1967 a habit
it was influenced a lot by Engelbart stuff turned out to be very hard to
in user interface like Engelbart's was actually a fairly hard one for people to
learn and a little bit later we saw a seymour papert's work with logo in 1968
and also this system done it RAND
Corporation this is about 1968 also
again on a multi-million dollar machine larger than this room for a bunch of
Randy Kannamma that complained that they couldn't type this list of processes
takes us back to the topmost level so they got really smart edit the flow
diagram first we erase a flow arrow then
move the connector toddled away so that we may draw a box in its place
recognising his handwriting here the printing in the box is being used as
commentary only in this case the box is
slightly too large so we may change its size we're back window control came from
literally then dry flow from the connector to the box attached a decision
element to the box and joy flow from it to scan we then erased the floor arrows
attached to the process post new area and move the box to a new position this
allows us to draw a new box
then chop off its corner and label it sub scan now notice it this is the end
here he goes up he sees it he has two dates but it's modeless draw flow so he
can just change it goes to sub scan correct the label add a decision on sub
scan so that control may flow to the connector a zero then complete the
diagram from scan to post well that was
that this system was the one that actually changed my way of thinking about computers because this is a system
life felt to me intimate I had a chance to use it for a half an hour in 1968 and I was not
notice that the command there's you didn't even see a pulldown menu there
every command was analogical if you want a box you drew a box itta made the box of that size if you
wanted a character on the screen you drew a character you want to get rid of something you scrubbed it out the text
editor in it took normal proofreading symbols so this is one of the great
systems of all time that nobody has ever heard of these days called Grail
graphical input language so that led to
this notion that the pencil was the best
metaphor that thing we'd never withhold from children and that the information
vehicle that Engelbart liked wasn't as powerful as this notion of giving
everybody a pencil and also in 1968 the
first plasma panel flat screen display was done and so we realized that one of
the destinies of all of this stuff had to be something that looked like this we call this the Dynabook and remember that
it can't be just like a lap computer of today because lap computers of today
don't allow you to carry other things easily it has to be really slim it has
to be really light has to be really good
it has to be really fast to test those ideas we do this is the forerunner of
the Macintosh it's a photograph about 1973 you can see the screen is about the
size fact is exactly the same number of pixels as the radius screen has 808 by
606 and this machine was about this the
the computing power of the Macintosh -
it's about three times or four times the computing power of the Macintosh Plus
Windows you're all familiar with invented on that machine small talk was
how we tried to work with creative literacy and get people to do their own
applications programs especially children here's a quick example so
here's a pro this is a wonderful crowd to explain this - because this is a an
applications program like Mac draw
so it's an object-oriented graphic system you just saw some selection
handles there here's a color change you
see the menu down at the bottom of the screen here's a size change now the only
thing interesting about this program is it wasn't bought in a store it was done
in 1975 and it was designed and implemented by a 12 year old girl after
just three months of programming in small talk 72 this is an example of a
real live tool done by a novice programmer a child and one of the nicest
things about the Macintosh - is it is the first micro computer fast enough to
be able to run a language like small talk so this is the kind of stuff we'll
be able to do with schoolchildren now for real and in fact Apple has a an
extensive project here in Los Angeles that uses powerful Macintosh's in a
grade one through six school so end-user
programming is something that's going to happen in the next few years
that's the vivarium project I'll say a
few words about it if I can hold your
interest for another few minutes so one of the questions you might ask is how
young a child can actually use these
techniques and here's a
here's a tape that I got about five months ago from the woman who's my my
accountant this little girl is the daughter of my
accountant and her parents both work at
home each of her parents has a Macintosh and when I found out the little girl was
interested in computers I gave her an Apple to which she rejected he wanted a
Macintosh just like her folks and in fact we actually did I did go against my
belief in market sir we tried a market survey on this 22 month old and
discovered that she vastly preferred Macintosh's with hard disks fact this is
this is what we used to call in the in the bad old days to years ago Mac pause
actually it's a little surprising to child psychologists that she can watch
the screen and use the mouse perform this displacement at such an early age
given that it's not too surprising that she can hit visible menu items in Mac
Paint here so when I saw this I wasn't
you know I was intrigued but I wasn't too surprised that she could do it but
what happened next really surprised me she wants a fresh sheet so she goes up
to the clothes box on the window saves
your file with the pop up wizard the pulldown
to get a new one and she's off and rolling again
that's spray painting and here's
something that Pia Jet Ian's say she shouldn't be able to do yet which is a reversible operation of erasing and she
she thinks of the hand as animation and
the fact we're starting to do some animation things with her now she was in
this she was 22 months old when this is
taking course kids have this compare she's not twenty-two months now unfortunately when something like this
happens you like to freeze them in stasis you know so great but after we
saw that her father had taken the picture what we got as a Mac with a
video out and we ran a line to a monitor in a bedroom and set up another camera
and then we got many hours of her interacting with the Machine and
determined that she was able to handle about 75% of the commands and the
pull-downs consistently and not just a Mac paint
but in several systems and in her
obvious interest in animation is something that we're planning on doing something about another thing that's
interesting is the intriguing possibility that if the Macintosh
interface is the first interface that reaches she was actually able to use the
mouse more jointly than she can use a crayon at this age and it raises the
possibility of building an ant question-answering system for a child
aged 3 to 6 they have lots of questions
then to allow them to ask and questions and get answers without having to use an
adult as an intermediary and that's something that I'm very interested in
but I fear it may change the sociology of the home
maybe for the better so very quickly
here's this this project that we're doing in Los Angeles very similar to the
one we did at Parc except at Parc we're asking the question what would it be
like for novice programmers especially children to do their own applications
programs and here in the vivarium project we're asking the question what
would it be like for novice programmers especially children to do their own
artificial intelligence programming so we want to know something fundamentally
new about what it means to program up mentalities and the way we're asking
that question is to set up a forcing function so we have this wonderful
School in Los Angeles called the open magnet school all 226 children and this
is actually an inner-city school though it doesn't look like it because this is part of their asphalt playground that
they tore up to make a garden where the 2nd and 3rd graders learn about biology
by actually growing things this is sort of the saran-wrap free system of
education food doesn't come in packaged
in saran wrap here's a typical thing that we do in the
vivarium is the children might study an ocean ecology like a tide pool lots of
different animals and plants interacting they might get interested in fish and
besides this drawing fish we would like to give them some magic computer stuff
like mac draw stuff that they can mold
use distortions like Darcy Thompson showed at the turn of the century that
related species are have related body plans conformal mappings can be made
from one end to the other animate the
fish and most importantly give the fish behavior actually program up a mentality
of the fish based on their observations of what fish do and for that we have to
invent lots of things we want to have multiple points of view like here's one
from the fish's point of view who want to see the world from the frogs point of
view from the herons point of view and from all the points of view simultaneously
because it's multiple points of view member points of view are worth ad IQ
points multiple points of view are how you actually learn and we want the
graphics to look something like this this is a scene from Bambi and we're
privileged to have with us the person who did Bambi Frank Thomas the old
Disney animator who is one of the advisers on this project and is in the
process of teaching every child in the school how to animate here's another
example a scene from Fantasia in which every frame was airbrushed by hand okay
only computers this this kind of
animation will never again be done by hand only computers can bring this level
of aesthetics and here's an example and
this to be able to do this in real time takes about ten times the power of the
of the Macintosh - so about 10 -
somewhere between 10 and 30 myths and you notice it's sort of hard to see
what's going on now to turn the Sun on like this the very same scene takes on
the order of a hundred nips now this is done by Bob Abel and it's done very
slowly this is done on a backs about the
about the same power to do this on a macintosh ii would take about twenty or
thirty minutes a frame but it's a very nice metaphor to make something alive
out of just raw material so the problem
with waiting twenty and thirty minutes a frame is they probably notice that children are very real-time devices
and so in order to to actually get
something that's closer to their rates of speed we use this Evans & Sutherland
flight simulator this is real-time computer graphics 60 frames a second and
there's a human pilot flying this plane the system does not know where he is
going to go wherever he goes it has to show him this planetoid size database
you get an idea of what the state-of-the-art this is the state-of-the-art in real-time computer
graphics you can get an idea of look at
the sky look at the leaves in the trees so the modeling is at a very high level
of detail and this system is able to show these simulations from multiple
points of view simultaneously
this is a Harrier Jump Jet
so this is the this is the equivalent so
we're seeing this from about a hundred feet off the ground and simultaneously
here's a ground level view
now think of it this is the equivalent of those million-dollar computers in the
60s that sketchpad and Grail we're done on and here's a third simultaneous view
from the cockpit of the second plane so
of course we weren't that interested in
marine jump jets as we've been talking
about fish so we took our Disney animators and a couple of Apple people
and we went off to Evans & Sutherland in Salt Lake City to see if we could do a
thousand miles of Australian Barrier Reef and some swimming things and Disney
animators have this way of turning everything into a character and one of
the things that's not been done successfully in real time graphics is sinuous motion so we got very interested
in how fish moved their tails and skull
their way through the water so we did lots of drawings of that and of course
sharks have a sort of a evil scanning eye which we put in so the result of one
long weekend with lots of cold pizza and
this several million dollar machine is this
so what you saw there was would be is
about somewhere between 100 and 150 million instructions per second to do
that in real time or about 50 McIntosh
two equivalents of computing now 50 is
an interesting number because that first machine I showed you at Xerox the
forerunner of the Macintosh was exactly 50 times as fast as a console on a PDP
10 time sharing system we forget what it
was like so many years ago it was incredibly slow to do computation
interactive computation back in the late 60s and early 70s and the real
breakthrough at Xerox PARC was that factor of 50 that made everything
different the breakthrough that we have today in the Macintosh 2 was a factor of
50 over the way people did things 20 years ago and another factor of 50 is
yet to come in the next few years the
most important thing about the 3d graphic stuff is that it has not been
possible to make a real-time system like this and anything smaller than a couple
of full racks of equipment because of the way packaging of chips has gone this
is one of these things that's going to go through an immense step function of several orders of magnitude change in
cost performance just in the next few years because the level of integration
of chips is quivering on the edge for being able to package a system like that
on a one or two cards that would fit inside a Macintosh to
don't need 50 slots so a lot of them a
lot of it I guess I should close this off with my favorite way of thinking
about this is that the Greeks held that the visual arts were the imitation of
life but the computer arts in which we're all interested are the imitation
of creation itself and Purvi see the
Italian philosopher said to know the
world one must construct it and it's
this through this constructive process the children actually learn computers
are the best construction material that we've ever come up with outside of our
own brains I see them as an amplifier to the things that we bring when we sit
down in front of the machine an amplifier not a prosthetic thank you
very much