Back to the Future of Software Development (2003)

okay anyway welcome

everyone to the April do in SD forums future of

software development lecture series co-hosted

by the compter History Museum Sanford

rockowitz chair of the series um

first all Laura Merling executive director of of

SD form is going to say a few words about the organization

or

um

hello everyone thanks for coming I think many of you have been to

few of the series I see some familiar faces out there

um at first I want to thank Sandy

for all of his work for these the entire

why I'm doing that is there's probably only a couple more

left for this particular series and we're looking for the next

series um so if you have thoughts or ideas I think

Sandy would love love the input and I know we would as well

do we continue this do you have other topics where else would you like us

to go um but um anyway for those of you

that don't know who a software development form

is uh we focus on emerging technology Trends and

education to you and the rest of the community to

try to say hey what's going on whether it's

wireless intelligence whether it's grid Computing whatever

topics and Technologies are so again if you have ideas

we like to hear them um but that said I and

announcement I have is announcement that I'm gonna do for Bill Gro

very quickly um next week on

Tuesday we have a home uh home networking

Expo uh kind of the connected home uh

so it's a series of two tracks one's on content one's on infrastructure

within the home and it's addressing different topics like I

TBO is coming to show their developer toolkit and talk about

connecting uh your PC to too and things

like that um we have a Broadband panel Microsoft's coming

to do a keynote on the future future of the home eight years from

and where it'll be so there's some interesting things coming that's next

Tuesday the 29th all right I'm done Sandy your turn I'm done than

k you everyone well we're

delighted that the Computer History Museum is uh co-hosting

this series and as many of you know the

museum is uh videotaping the talks in the series

for their archives and John tul the the executive director

of the museum is going to say a few

words keep very

but I want to really thank every everyone and it's been a real honor

development Forum we really have a lot of synergy

while this is about the future we're about the history

the authentic history we have a great great uh

program ourselves with a lecture which we We join with with

folks in having our next lecture will be in our new

phase one or Alpha phase building at

1401 North Shoreline Boulevard in our new Auditorium

on June 10th it'll be Jurassic software the

origins of consumer software Stuart alsup

Scott Cook uh uh and and a whole bunch of folks

that are going to be at that that particular lecture but our role

is really to preserve the history authentically and to

really make that happen we have Hardware artifacts to go back 25

years we've got software artifacts to go back 25

years we're putting together over the period of the next three or

five years are really first class worldclass

exhibition for people to really enjoy and to understand what

it was all about in the innovators like we're going to hear about tonight

it's particularly happy tonight because Allen is one of our Museum

fellows and it's a pleasure to be here and it's a pleasure

to work with the software development Forum thank

you well as

as many of you know the talks in this series are intended to

answer from a variety of perspectives

the larger issues of software development

and to this end we've enlisted uh

distinguished uh figures in software

development who can share with us their thoughts and

insights how on how software can should

and will be developed and the Technologies by which this

will happen um

simply put the talks in the series are intended to

answer the following question what is it that leading Ed

attention to

and why is it important now

as Laura said we're coming to the close of this lecture

series we expect we'll have one more talk either

the uh final Thursday of May or the 1

of June and we would very much like

your feedback um

on what's worked for you in the series what

You' thought has been valuable what might have been done

better um and looking forward to the coming

year what might we look at for

in terms of a of a series uh address

to a tech a senior technical audience

um what what who what

what people would you like to see what topics would

you like to see address and um

I know if some of you might have caught the slideshow that was

here earlier on as you were coming in it included my email

address which is uh rockowitz

r c wi TZ

at moft m n soft.com

and and I'd very much appreciate

hearing from you um your thoughts on what

we might do as a follow-up series next

year our speaker tonight

is is truly a seminal

pardon

it's a with Alan K it is a true challenge to

keep it short uh Xerox fellow

Atari Chief scientist Apple fellow Disney

fellow HP fellow president viewpoints

Research Institute um the list

goes on and the awards goes on

um alen has said that the best way to predict the

future is to invent it and my God has

he talked his he walked the

talk um his image of the

D book um sort of in

way back when computers filled rooms sort of clued

us into what computers might be what a personal computer

might be one of the creators of small talk the

first uh Dynamic o o

system he developed key aspects

of the window user interface that we

all know so it's a real honor uh to welcome Alan Kay

here this evening

[Applause]

you

yeah thanks um

you know it's it's great to be back at Park

but of course this part of Park never existed when we were

we were here was built after a whole bunch

of us left so this is actually a new part

of Park and since

this is uh

Forum has kind of the fabulous

Paradox of being about the future

software but sponsored by the Computer

History Museum uh

it opens a very wide range of possibilities for

giving a talk and so what I what I

thought would be interesting is to

start off uh with maybe

a little bit of a CET about the

uh last 20 years years or so of of non-development

in so many areas take a

look at some of the there's some interesting

40th birthdays this year there's a really

interesting 35th birthday this year there's

a 30th birthday this year that are all I think

have something to do with the future software

uh development and uh

many of these uh are what

I would call really promising roads

um not actually taken

roads that had

tremendous potential and actually paid

off uh tremendously well many years

ago and then the vicissitudes of various things including

the way commercialization got done these roads

were not taken over the last 20 years or so and I

think to the detriment of most software

so the the KET is a simple

one and actually this KET is pretty much

the same about the last 20 years and

plus years ago we made the same one because this

was kind of um IBM

and looking at business and thinking

of business as the ultimate Target for trying to

sell computers and the problem with

uh business

and um organizations of

this ilk is that they are

kind of an epitome of two

conservative uh properties that most human beings

have um not

just in advanced uh societies

but in traditional societies as well and one one of these

is called instrumental reasoning instrumental

reasoning is a form of reasoning

by which when a person is presented with an idea

or tool they judge this idea or tool

solely on the basis of whether it contributes to

some goal they already have

okay and I think if you think about that you can see that many

many people you know are basically instrumental

reasoners um

only as far as we can tell from working with children only about 5%

of children are not instrumental reasoners

basically built

um built in a person who's not an instrumental Reasoner

when presented with a new idea or a tool will actually

transform themselves in the presence of the idea and

the tool and so these are people

are called uh early adopters

that aren't just fists some early adopters

just want to be early so they don't have much taste

but there are other people who respond actually

to the to the new idea and these

people are to be found in each era and often they

are the ones who come up with

things that are thought of as far-reaching Visions

um that then

unaccountably take many decades to actually spread

generally even if they were a good idea in the first place

this happens over and over again um

and the other uh

Allied idea is that

um we humans are very

very uh prone to do case-based

reasoning it's the main

kind of reasoning built into our nervous system

so it's a compartmentalized

U way of dealing with past experience our

legal system is very case-based

engineering happened thousands of years before

science because it was case based

have to have a great theory if you're Gathering up things

that worked so you tend to make cookbooks

of things and you always basically start off that way

uh the problem is is that you start losing it's a

great first order Theory and a bad second order

Theory so when the commercialization of

PCS happened a little over 20

years ago and it was aimed at business was basically

a customer that didn't really even want the computer that

much and when they did want it they wanted it for

automating paper so as mclen said we were

moving quicker and quicker into the future but steering only

by looking in the rearview mirror and

this commercialization um

almost completely occupied the

space so we'll come back

little bit and of course this has happened in history

so uh when the printing press got

invented and for

the next 100 years or so

much of the printing that was done was to automate

done by manuscript

uh U with monks many

of them religious tracks and the if you've seen a Gutenberg

Bible you will have

been astonished at how much it looked like a handdrawn

book Gutenberg actually had more than 250

character characters in his font because he wanted

not just to do the upper and lower case

but every ligature every abbreviation

medieval scribes used he wanted to duplicate those

exactly so he carved not just

the 50 or 60 things that he needed he carved

253 and then they

brought in people after they had saved all his

printing they brought in people to hand illuminate

the books so they looked real

now and the books were big

like this and the reason is nobody knew what a

book looked like should look like

the only thing they had as a model with these old manuscripts 50

years later Alis who is a printer

in Venice is his uh last name was

maker was eldest

Manus and actually I his Italian name is Aldo

and I think of him as as uh Aldo

minu so it sounds more friendly than

Aldis minucius and he wanted

to do a portable library um

around 1500 and went out in the streets

of Venice measuring saddle bags to see how big

the books should be if they were going to be portable

and the answer he came up with was the size book we use today

as our main size book and now you'll never forget

that because it's such a wonderful

that Aldis actually went out and did user centered design

to make his portable

library but in spite of

people like Aldis and Rasmus understanding what the book

was going to become the actual

printing Revolution didn't happen till about 150 or

maybe even 200 years later if you look at it from the

standpoint of the way the printing medium was

used to argue rhetoric

gradually changed to the rhetoric that we

view as modern rhetoric for arguing about politics and about

real world and Science and

all of this new way of expression and the

new way of using the printing press was done by children

because 150 years later nobody who was involved

in the invention of the original technology was still alive

and the people who understood back then

what the book was going to be had also

died right so the

people who stood the great idea

uh were uh did understand it

back then but the the large group of humanity

actually had to grow up into it several Generations over

before the children actually found this thing I think

this is what's going to happen with uh

Computing because we have a similar kind of thing sort

millions of dollars thing is kind

kind of mainframes owned only by

institutions and the desktop computer workstation

is kind of in it Gutenberg Bibles by the

way cost about $60,000

three years of a Clerk's wages

the nurenberg book fair proclaimed them

and they were all marvelously similar the

advertisement said they showed 24 of these books

that were as similar as they could possibly

be um so this is like a

workstation here and here's our

little notebooks designed to be portable

and this is not hard to figure out because of

Moore's law but the thing it's hard to figure out is

when are we going to get something that is

equivalent in the large dimensions

omething closer to Universal literacy

only one person in a 100 in Europe could read back

here um 80 and 100 could read

a little in here uh schools

had developed to teach reading to everyone

by here reading and writing

thinking about the ideas in a new way and so

forth so when you look back

40 years

you can see a lot of promising stuff

that happened then we'll talk about it a little bit because

it's it's worthwhile thinking about how

some of this great stuff got started and then when commercialization

happened uh it kind of died

away this may be overly

harsh but if

we take a look at 140th anniversary

this happens to be the 40th anniversary of the funding of project

Mac

and uh Doug engelbart's

proposal to arpa was done in

62 right before so it was just starting

to happen here and lick

lier and this is a picture of him actually

from that era when he was

the perceptual psychologist went to Washington

they had some money left over from the space program

as it went over to NASA and they decided to give it to him to do

with whatever he wanted and he decided to

try and deal with this idea that he' had

about man machine symbiosis and he wasn't the only one who had it

but he had a nice phrase he said not too

many years human brains and Computing machines will be coupled together

tightly and the resulting partnership will think as no

human brain has ever thought so if we look at

where we are today it just hasn't happened

yet it's happened in the Sciences

The Sciences particularly the physical sciences have been

absolutely transformed by the computer

scientists are thinking in ways that no scientist ever thought before

because there are computers but in general this is

not true it has not happened yet

so let's look at a couple of the arpa

goals um

sorry

one of them is this notion of end user computer

computer literacy

and

now last year was the actual I'm fudging a little

here last year was the actual 40th anniversary of sketchpad

but Ivan's thesis was signed in January

63 so I'm claiming

something here it's always worthwhile I always

find a way of playing this because it's worthwhile thinking

about this period when uh

we like to say there was only one personal computer

and that was when Ivan was on a machine

that was much larger than this room from 3 to 6 o00 every

morning

so this machine did not even draw a lines and notice he's

pointing to the segments here and he's

saying uh now make everybody mutually perpendicular

and sketchpad there just figured that out and

little flange so

continuous zooming and clipping on this first

this is the first actual window ever

done this display could only plot

points so I have an add uh

write the line drawing routines as well so there the

constraint was uh

parallelism now the constraint is collinearity

so he's using those solid lines as guidelines

for drawing the dashes and then he'll make the solid lines

invisible

so about half the capacity of this huge

Sage uh computer the tx2

was used for doing the

graphics

okay now he's got the hole in the

flange and the actual sheet

of virtual paper he's drawing on is about a third of a

mile on a side so it's a continuous zooming

interface this is why

it was called sketch pad the idea was to just draw

quickly

an idea and then get the machine to clean up the drawings

by telling it what the additional rules

were so use the uh

center of the Cross piece is there to for

the center of the radius of the circle and so when he tells the thing to

become mutually perpendicular that drags the

circle and makes them a nice little rivet

and so sketchpad is actually a

continuous nonlinear

constraint solver so a little bit like a

a graphical uh

uh spreadsheet many

the spreadsheet because it

nonlinear problems like stresses and strains

on Bridges and so forth

now he's got himself a little rivet

here and

here's one of the first ideas this ever appeared in software

is a notion of making an

instance so this is an instance

of that rivet and you can rotate it and position

it and scale it

individually

so he's going to Anchor it into the flange

there can see that

the success of sketch pad led to a desire for nicer

displays and

here's a very powerful idea one of the first times

ever in software is the idea of multiple instances

and then he notices whoops I've got that cross

piece there let me go back to the master

today we'd call a class let me make the cross

pieces transparent and now when I go out

drawing the instances have all felt that so this is a

true object-oriented Software System I believe the

first real one in all

important

details and

it's a prototype oriented system because when you make something

so he's got this rivet in the flange

can make that into a a

class or a master and now he's getting

instances of that thing he just

constructed so this is dynamic

objectoriented programming done by

continuous problem solving

could you possibly

in one year have invented computer

Graphics have invented

um um

object-oriented software and

done the first real-time Problem Solver

and he looked at me and said well I didn't know it was

hard so he had a tremendous advantage

in that nothing was known about computer Graphics

nobody had really done it in any interesting way

and uh one of the wonderful things about his thesis

which by the way is is available from

MIT you get online and go to the Barton

library or one of the library services in MIT

you can get the PDF file for it and

it is astounding I believe it is the closest thing

to an act of a Newton in

our field in the sense of Newton

stuff was impressive in any age uh and maybe

most impressive from where we were before Newton to

where we were after Newton and so

Ivan here in one PhD

thesis written in machine code on this

huge ungainly machine sort of gave

ntire image of what it was like to be able to sit down

and not just make pictures on a computer

but make things that were simulations so this is

in in many many ways everything uh certainly that

I think I've been doing and a lot of

us have been doing have been kind of footnotes or fleshing

out this incredible

uh vision and

realization um unfortunately

Ivan is too Restless to actually

uh he you know he could do

most unbelievable things in little fiveyear periods

and he had to move on to something else and

uh those of us who followed him um just simply

lack uh his incredible talents

so uh it's actually hard to go out and buy a system

today that will do all the things that sketchpad could

do back then kind of an interesting

commentary so another paradigmatic

system is sort of the almost the opposite done exactly

at the same time was the original video

done in the pdp1

here it is if you ever programmed it is

really one of the first machines that anybody would call

a honey because uh

it just right there it was sweet

you could really make it do things and it had a

uh one of the early Graphics displays from

deck on it and

uh one of the first things they did on it was the one of the first

Graphics text editors which they called uh

expensive uh typewriter

because it was basically a typewriter except it cost

$110,000 or so

and then uh pretty quickly Steve Russell had

been reading the doc Smith lensman series

to take a shot at

uh trying to do that and he came up with

uh something like this this is a a

kind of a recreation of it

but the basic idea is you got F equal ma

going and so when you

h uh

try and steer the thing you have the problem is you already

have quite a bit of

velocity built up

and so if you want to uh stop this

thing you have to point him away and try and kill off his

velocity so it's

quite interesting and challenging to use

even before they put a sun in there with a gravity itational

field that made it even more interesting

so now space war is one of those

things that was so

simple so doable

pretty much every computer that had a graphics display

on it somebody would sit down and uh

figure out a way to do space war and one of the

most important things about it it also use

this object idea as you want wanted

have not just one spaceship but a space ship for each

person playing it so it had many of the interesting elements

of what we think of as computer Graphics

today um and it's

fun

okay so

another part of uh the arper Dream

from way back was the idea

of being able to do group

collaborations and this happened in two

forms which later got combined one was this

desire to do the intergalactic Network

as lick called it back

then um

and uh from Doug engelbart's

proposals about boosting the combined

intelligence of groups

and so it just happens

to be the 35th anniversary of

was ever given actually a little bit later in the year

right bill so uh if you don't

everybody here must know Bill English but bill raise your hand please

because everybody's heard of Doug

angelar but Bill was the guy who made all the stuff

work so he was

the uh partner he was the co-inventor of

the mouse which was invented in

64 and this picture

here is

is a picture that uh could have been

taken yesterday at somebody sitting at their

desk but it goes all the way back to the 60s

so this is kind of the birth of personal

Computing the way I look at it

whoops and do we have

sound somebody giving me sound from this

thing

let me look just that low let me stop this

for a second this is one of

the advantages here of having a system where the authoring

is already always on is I can just uh

rewind this

guy let me just move him up to

the about

here start

him again so I can say all right I'd like to go to

produce but I'd like

go to produce they get big I'd like to say one

branch only and uh let me look

just that low and 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 there's a root I said I have

here so here I'm afraid I'll need

ifferent picture The View so here's what

I do with a picture drawing capability here it's a slight 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 line Library what am I supposed to pick up there I

can just point to that and oh I see overdue

and all well there was a statement there with that name

on it go back what if I what am I supposed to pick up

the drugstore h i see very interesting

all right Market

can do things if I want to just say I'd like to

Interchange produce and canned materials

Pingo and they're all numbered right if

I care to look interchanging them very quickly

can are going to get interchanged with

produce they do it and all it's re

numbered so one of the nice

for showing this is

you get a nice this is from the uh 68

demo in San Francisco that I was very happy to

be there was one of the run like

a military campaign uh

nothing was left to chance right bill

was the most incredible thing you've ever seen and this

that thing on the puny little thing

there was bigger than this whole screen it was just this

huge thing done by a

uh you know one of the first light valves

that was pretty much using an atomic bomb as a light

source huge big

situation display but one of the reasons I show

that is the computer was actually down in Meno

Park while angelart was giving this uh

demo up in San Francisco and we'll talk about

that in a second but if you look at the response on this system

notice the response was subsecond on

all important interactions and I have to tell you the

was done on was about a half a MIP

192k bytes

and uh was time shared so

not the only user on this machine think about that for

just a second and think about what we don't

get off these several hundred

MIP machines we do not get

subs second response so how could these guys

have possibly gotten subsecond response

anybody got a

theoryof that's a necessary

but not a sufficient

condition

who's got a more a more close order

Theory okay I'll tell you why the

is is because they wanted to get subsecond response

they were not going to be settled for anything less than subse

response they worked their asses

get subsequent response because it was part of their image

of what it meant to fly that was

one of the metaphors to fly through n-dimensional

thought vectors and concept space

so this was a conception a grand

conception about what it meant to

have a corpus of knowledge that was useful to other

people and also an understanding that in any retrieval operation

you're doing you're spending most of your time

rejecting so the only way

you can make something like this work whether you're

browsing or any kind of searching is you have to reject quickly

so they had this thing set up it was just

incredible we could go on and on about it but it's

worthwhile when you're thinking about the future of software

you don't have to go much past this

in 1968 to make your first

list of 15 criteria for what you should do and what

you shouldn't do because they just really really

went after it um

and the next part of it is perhaps

even more interesting and was part of the original

conception as well so check this

out now computer do the automatic

switching that'll bring in a camera picture from the camera monitor

on his console such as the camera monitor on mine

is hi Bill that's great now we're

connected audio you can see my words you can point

at it and I can see your face and we can talk so bill is down

collaborating go off to a

a directive file and see what the directive

is to get Roman numeral page

numbers I'm into the file now here's the

first level of the hierarchy let's open up

page formatting this is angle we want

page numbering so we'll open that up we find

yes here it is Roman numerals we find

out the directive y your bug's right on it already

D here directive we want so we worked down quite

a way into a hierarchy as you can

see okay so

another part of this conception was that through every part of

the system not as a feature added on from the outside but to be able to do

immersive collaboration of various

forms and this was a a demo of

it and they even had their meetings

when they had a meeting in the same room they had the

consoles down

in this these sort of circular table Arrangements

had so that they were actually sharing the context and they

were using this as a sharable Blackboard back then

so um one of the nice things is you

can get videos of

these demos there's a lot of stuff written about it

and um if you're interested in any kind

of uh improving group

process uh

thinking about what criteria are for starting you

you have to start here it's

because uh most most of us aren't as smart as these

guys were and the software that we are given

today proves it because it was

who never looked at this stuff never understood

it so another thing that arpa was thinking

about was kind of a

form we saw one form

of what the computer environment was

going to be like here's a here's another one from Rand

was done about the same time and

it interest it it superficially doesn't

look that similar

but it actually uh is remarkable in the

points of similarity it has with the engelbart system

from a completely different point of view

which is uh from the from the standpoint of the

end users that they dealt with at Rand Corporation it's

called Grail and uh it's one of

amazing coincidences uh that both

the tablet and the mouse were invented in the same year

so it's part I didn't know whether you knew that bill but it was

so Tom Ellis did this these tablets

were actually kind of like a sign

that you were a real graphics person back

cost $118,000 in

1968 which is like like

$100,000 today to get when they're made by

hand so they're probably less than

50 of them made in total so take a look at this

end user

system

first we erase a clow Arrow then

move the connector out of the way so that we may draw a box in

its place but recognize he wants

box and makes one now it's recognizing his handwriting

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 where

Modern Window Control came from literally then draw a

flow from the connector to the

Box attach a decision

element to the box and draw a flow from it

to scan we then erase

the flow arrows attached to the process post new

area and move the box to a new

position this allows us to

box so you get

it so

question that one could reasonably ask in the year 2003

is why the heck can't we do that on our

pdas and the answer is

that nobody who's making pdas has the faintest idea

that this is would even be neat to do

they just can't imagine that it's anything

but a little tapping thing

and it is truly amazing to me that

graffiti is not as good

as this recognizer was back then and in

fact this recognizer was published in a paper

uh in 1966 by Gabe

Groner for just anybody who wanted to know how to do

a single stroke almost perfect character

recognizer it's been in the literature since

1966 but the people who did

today absolutely did not we're not going to go back and read

anything that those old people did hey

we're just as good if not better

so this is to me the real difficulty

with getting software into the future

is that software can't even use the

past and so we basically have millions

of people now who are programmers and by the law the bell

curve most of them are average or

[Applause] worse in fact if you think

bell curve is 83%

are Caesar worse so if you

throw a bell curve on programmers and you basically

have a tyranny of the majority that

uh winds up you know full of Sound and Fury signifying

nothing so this is a huge

huge problem so one of the biggest

first things that we could do to get ahead in software is

to at least start with best

practice and the other thing was nice about this recognizer

was only about 8K of

code because it ran on a really small

machine that is the machine

was an IBM 360 Model

44 but it was pretty Tiny

from the standpoint of bits and

bites okay

and then sort of the

more indirect thing that arpa had a huge part

in was in trying to get real

computer science and engineering invented

now we claim this hasn't really happened

very well yet but there are some really

interesting uh stabs at it

my favorite one what I think of

as the best thing ever done

having to do with programming language was McCarthy's

invention of Lis and it was done

kind of in a typical way that these things get done is that

McCarthy was basically a mathematician

and so he was able to do this

when he was thinking about a universal way of doing

Computing this is kind of the Maxwell's computer

uh Maxwell's equations of programming

it's the bottom half of page 13 in the

list 1.5 manual and pretty much

everything that

uh is good about programming is in those few lines

so it really has a lot of the Maxwell's

equations then Steve Russell again the guy who did space

war was the person who coded this up and

made it into a real language called lisp then many other

people work with it um that

happened a few years earlier but this

year and I just sent an email to Peter Deutsch

uh yesterday to confirm

that this is indeed the 40th annivers iary of

a truly remarkable version of this which was the

first interactive uh

lisp on a standalone machine and

it was also its own operating system it was done by Peter when he was

only 16 years old so if you've ever

puzzled through trying to understand

the lisp eval

uh imagine a 16-year-old boy being able to

understand it and writing one of the most beautiful machine

code programs ever for

this uh for this machine and

many of the things that we wound up doing after this

were based on the fact that this stuff had been

done tiny little machine actually

ran in uh what was called a four

core machine which is uh 4,8

bit words so it's a little over 8K

9k or so so The

Interpreter was about 2k and you had the rest for doing

Computing in

ow for me

my background is in math and biology and

my first collision with this stuff was in

running into sketch pad and then simula seeing

the similarity to U biological

cells as a universal building block and

thinking up the idea of dynamic

objects then we tried this as an operating

system on this little machine called the flex machine

early desktop machine

the uh I later

saw this wonderful Grail system

and the same year I saw early flat panel

display at the University of Illinois

so we started thinking about the prospect of

putting the transistors in this machine on the back of one of these

displays so you could do this

and that same year

of 68 I saw Seymour paper

who had had a tremendous Insight involved ining children

that certain really important forms

of U advanced mathematics

particular uh Vector differential

geometry uh was akin to

the child's own way of thinking about itself in the

world so the child is the zero

the child goes it's at the center of the of the

universe the narcissistic coordinate

system and that if you know differential

geometry you know that's the coordinate system that differential geometry

uses it's always what is the geometry

like from where you are and so for instance

a circle is just go a little turn a little over and over

again you don't need any x^2 + y^2

= R2 or any of that stuff so when I saw

that it blew my mind completely and that got

me thinking about this little computer for

children and my image of

it was uh 12 and 13

year olds here would actually sit down and program

their own game of space War learn about

Newtonian Dynamics in the process and have fun

playing and uh those of us who are

around uh a long time ago will remember

and in the computer museum has it parked outside

its old building is the old SRI bread truck

how many people remember the mobile

radio packet radio stuff yeah John shock

does You' says this I think it

was a 40 conine van or something was packed

full of stuff and one of the uh frivolous

things that happened back then was one of the first

emails ever sent from the

beer joint called rosatis up in the hill

they drove this Econoline fan

up there they had not

even a line of sight I guess to but it was back at switching

to Sri and then into the arpanet into Washington

meanwhile they there swizzling down beer

with their terminal at the on the the one of the outside

rosatti picnic tables so we knew that

uh Wireless Computing was going to

come and this kind of packaged up

um the idealizations of all

of this stuff into this one idea I called the dnab

ook

and so if you take this arper dream

um it took shape for me when

children came into the picture because

um I really didn't know how to design for a adults I wasn't very

adult myself I didn't have

any real contact with that world but

the improving the child's condition for learning

something that really appealed to me so I got very interested

in it and

I have to thank Bill English once more it's great that bill is here because

I rarely get a chance to thank him in public

but um Bill actually took me under

his wing when we started here at Xerox Park a long

time ago and and one of the things

me is Alan maybe maybe you should write a budget

and I'm afraid I really did say to Bill Bill

what's a budget remember

that Bill said well you put

this number here and you put this number here

um so this is actually the 30th anniversary

Chuck ther's Alto

here it is and so

Milestone it went along with a whole bunch of other

stuff that we did and

historically it had an interesting parallel with the

pdp1 lisp thing because it

was the the strength

of what McCarthy had done

um because I'm I'm basically a

mathematician not a programmer I program a little

bit but I don't consider myself a programmer

but I felt I could do the same thing for an object-oriented language

that McCarthy had done for lisp and this is

one version of it and Dan Eng Les who is

your hand

so Dan uh was the reality guy

he took this little paper thing

and made it into uh a succession of small

talks that uh uh lived to

this day and Chuck

uh uh fashioned this uh first

and from our standpoint what we're really trying to

do was to get

machines built so we could take them down into schools

and start working on this is Adell Goldberg

Here and Now up to the

present I'm just going to show you a few of

things that we've been doing

um so squeak is

kind of an outgrowth of some of the stuff

did at zerox Park and we

don't think of systems like this as

goals they it's hard to

Define uh but it's relatively easy to

Define uh computer systems as

vehicles or media so you sit down

because you're basically trying to make

models in it and you get ideas

kinds of models and so you have to use the modeling material

to make the uh different kinds of modeling

material to make the models and so forth

so this system uh is a wide

it's been implemented on

many many different kinds of platforms bit

o explain a little bit of that because I think it's

very very important for

future software especially having to do with the internet

uh it's also of interest because it was basically done

by this small group of people here here's Dan

Engles again Scott Wallace Ted Kaylor John Maloney

and Andreas Rob uh

so it's an example of a system that involves its

own operating system and many many things that

uh we think of as applications done by a

number of people rather than the hundreds of people

that you think of uh needing it so there's interesting

questions about well how is this actually possible

has a very flexible

object system so if we take a this

is kind of a cute little demo here if you take a paragraph

doing word wrap um and you're in

a real object system then

uh you can just let those guys

wander around here and if you think

about what do they have to actually do

they kind of have to follow the

leader

all right can get them to go

faster

and the algorithm we usually think of is kind of

the one that is done between frame times

so it's actually doing the same thing there but between

frame times it looks like it's happening instantly that is

programing because it's basically just four little

rules to do it there's many kinds of media in

here I'll show you just one other uh thing

here

and so

we live in a world of applications

but if you think about it applications are one of the worst ideas anybody

had and we thought we'd gotten rid of them here at Park

in the 70s because

applications kind of draw

a barrier around your ability to do things

what you really want is to gather the resources

that you need to you and make the things that you want

out of them uh so for instance

one of the ways of thinking of desktop publishing is

that it's just uh Graphics done

right right because you basically

to be able to make almost any kind of graphic construction

on the screen you want it to be able to react to

things and you want the elements to be

uh not confined in the bounds of an application

but be uh be part of something else so

you should be able to uh have primary

and secondary reactions so here I'm

drawing through text just

as you would expect but I should also be able to hop

the fence here and for instance drag this guy

out like this

and but I should be able to do this without destroying

so you see if I'm SC scrolling here it's scrolling

there and flowing through

these objects interact with each other to the extent

that they need to but no more so and you kind of get desktop publishing

for free or any kind of media

composition and actually

one other thing I should show you here

as long as we're talking about this is an idea that I spent

12 years at Apple intermittently trying to get them

to adopt which is the idea of unlimited desktops

and Small Talk At zerox Park

had this and the squeak system here has

this and what I mean by that is that

each time I decide I want to do

something I really want a separate work area

to work on and I like these work areas to persist

over time and we've been looking at them so there's no

separate presentation uh thing

there's nothing like uh postcript here

like uh PowerPoint

so if we look at the projects that are in

uh in this guy here we

can see the presentation that I've given so far

right here's where we are right now

here's the next thing I'm going to show and I decided

i' really rather show this thing

and each talk that I give is

a sorting of the projects that I have so the

idea is you never have to go to anything weaker like PowerPoint to give a

presentation you just link up the various things that you've been

working on and U and show them

are also the things we ship around the web

instead of uh web

pages okay

so just a word

about how the porting is done it's kind of

cute so here's the the

thing that I I think

is actually known but is not

well dealt with in software and that is

that um for

decades software has been specified

using paper documents people have tried

to implement to those paper documents and then they've tried

a benchmark Suite of some

kind to validate the Imp uh implementation

and to my knowledge it has never succeeded in producing

compatible implementations

across platforms Fortran is

famous for the exceptions

and even C

is not exactly the same across platforms

so that is really a shame

so a different way of doing it though is

to make a model of your kernel

that can be debugged so

it's not a paper model so when we decided to do

squeak um there had been some experience

in

um doing various kinds of small talks at zerox

Park and afterwards so the idea was to

found an old Mac that would run an old Apple

version of small talk

and we'd written a book 20

20 years ago 20th anniversary of this book

now this book had in it among other things

the VM

of uh Small Talk written in

itself so if you type that in

and get it running you have a simulator the VM

you could treat that simulator as the spec

and say okay that's the only

uh uh validation

and specification of the system that

we're going to have and we'll debug

this and when we get it running we will have the definition

of the VM that we're hoping for now of

course this doesn't do you any good because it's running glacially

slowly so next thing you have to do is write a

translator of that and this work was primarily

for squeak was primarily done by uh Dan Engles

and John Maloney um

a translator and this

translator should

with without changing the meaning of what the simulation

does translated into a lower

level form that can be put on another machine

like we we chose a subset of C

as the lower level form to go for a Target

the subset of C that tries to be compatible across

platforms so all of a sudden now you have a power PC

VM and this top

stuff uh is machine independent

so all of a sudden uh we were now off

the old Mac and onto the power PC

and since it contains all of this stuff you can use it

to improve itself um about

a month and a half after after we put it out on the net

guy in Germany we'd never heard

of before sent us uh a version

for all the Intel PCS and that's what we run

today and it now runs on 30 platforms bit

identically because the key idea here

is the debugging of this complex thing and

then having a mathematically sound translation

that pre preserves the actual

meaning so in many

think of this is an old technology or new technology

but very few people are doing this today

trying to run on many different platforms

internet I'll show you the relationship of

squeak

so there's this nice malleable stuff called

CPU and memory and then there are

bad defacto standards called operating systems that

take away most of the degrees of freedom

and the internet with tcpip

is really great then there's a really

de facto standard called the worldwide web

you want to do

so we actually when we're running on a machine like I'm running

and we have a couple of different PCS

here we actually have our own operating system

but we contact the existing operating system at a single

point equals about a thousand lines

of code because we have to share the display

we we want to do this without killing the existing operating

system but all of our own tools are inside of

squeak and these are go portable with the rest of the

stuff and the same thing with our

socket system for handling the internet this gives you a completely

portable uh system

that uh can go across a wide variety of platforms and

usually takes just a week or two at the most to uh

the port something to think about for

software

okay so let's take a look at the

present I found uh doing these things

howing some of the kids stuff

uh makes most of the points in a way that's easy to

understand um

nice thing about working with kids is

um when you give them something

they feel quite free to reject it

it's hard doing tests with adults and companies

because adults have been trained trained for 25 years to

take for

money

but most kids haven't learned that yet

so what the kids really want to do is have some fun so

like a fun project for them is to design a

car that they can learn how to

drive paint it it's

like a 10-year-old kid

here put in some

nice little specular Reflections that we learned

Disney

and we found that both boys and girls love to

put on big powerful off-road tires on

their vehicles so their vehicles

wind up looking a lot like this

usually and we got a little graphic object

and in this system

and this is an Insight that goes back a long ways to the days of Xerox

Park was I think after one of our beer Buss

at the black forest or something

I was blury looking at the screen in my office

and I realized that to an end user there were only smart

rectangles floating around on the screen all

the crap that we had underneath all of the inheritance

all of the other stuff was totally invisible

to a naive end user it was just smart

rectangles if you start thinking that way you start thinking

about gee what I really H want to have is uni

Universal objects that can wear costumes

and I'll give them a little bit of idiosyncratic behavior

and try and build my system out of that so here's our little

car

and it's a graphic object but

it has another view of it which is uh

symbolic I'll call this

car the kids like this and

to teach the math so we want to get them interested in this

so for example if we look at this property

called cars heading and start it counting up

car turns

heading changing

here's a behavioral property forwarding

binging turn

binging make a script just drag out a

tile hit the

clock start it running

and of course I can steer by just changing the numbers

here the good one is zero Which

car goes straight negative numbers it

turns the other way that's not really like driving a

car so we say okay make make yourself

wheel

I always do a blue

one it's just another little costume

one of these guys

look inside of it call it

wheel it's got a heading

you look at that number there on heading it

positive and negative as I turn the wheel

positive and negative numbers here influence the

invitation to pick up the name of the numbers coming out

of the wheel just drop them in there

and I should be able to control the car

now so it's sort of for the kids it's

one step learning about what a variable actually

is generally one of the harder concepts

for kids to grock but it's very easy

here because it's operationally in their

space and the system is just recompiling

and recompiling and recompiling as we go every time we make

a change so these scripts actually run as fast as

underlying squeak scripts they are the underlying squeak scripts

very fast things as we'll see in in a little

bit okay

so once they've done some of that stuff we can

investigate the real world using

our ability to make models now

say sound please keep the sound

on

both hands oh do not pay any attention to

[Music]

else who's got

Apple

that's a

jamer

what you get what' you get try use

stockes

together so put

spongeball I think we should do the shot foot

and the spongeball because they're

two totally different weights and if you

drop them at the same time maybe they'll drop at the

same speed drop

[Music]

so so the average of galileos

per class is usually about

one for 30 kids or so

this little girl cut right to the chase you realized the stopwatches

bad because you couldn't tell when the janitor is dropping it's hard

to whole thing takes about a second or so

and so she just thought well all we have to do is

listen when they hit if we can hear two hits then they're falling at a

different rate if we hear one hit so this is Galileo's insight

back when uh it was hard to time things

so children can actually think about this stuff operationally

very well and as we'll see in a second

they can do better than most college

students so the way to actually investigate this is

to take a video of the dropping ball

here and

um hard to see what's going on by

looking at the video but

um these frames are actually part of the objects

can do things so we can just pull frames out

we can stack up every fifth frame here we can

stack them up vertically and we can start

measuring them so

one of the ways of measuring these is to

take the make the height of a

rectangle

go from the bottom of the ball

and one frame to the bottom of the ball in the

next frame that vertical drop

is actually the velocity because it's a distance

traveled in unit time

you can see the velocity is

increasing we'd like to know how the velocity

is increasing here doing this a little bit too

quickly but so one of the

ways we can get a qualitative look at it is simply by stacking

these guys

up Galileo had

a very interesting way of doing this himself and

if we we look at this we see oh the change in velocity looks

constant change in velocity is acceleration

it looks like constant acceleration so the little script

that they write looks like this they make

a simulated ball

and have a variable called speed and they say let's

increase the speed by minus5 a constant

each time and then let's change the ball's position by

speed right so that's a second order differential equation

done by 11 year olds and we

let Tyrone tell you how he did his and to make

sure that I was sound please just right

I got a magnifier which would

help me figure out if the size

was just right after I had done that

I would go and click on

the little basic category button and

then a little menu would pop up and

one of the categories would be geometry so

I click on that and here it has

many things that have to do with the size and

shape of the

rectangle so I would see what the height

is and I kept going along the process

until I had them all lined up with their height

I subtracted the smaller ones hyp from the bigger one

to see if there was a kind of pattern

anywhere that could help me and my best guess

work so in order to show

that it was working I decided to leave a DOT copy

so that it would show that the ball was going at the

exact leaving a little cookie acceleration

behind to show that it matches up with where

the thing is in the frame another one that the kids did is running

the movie against their simulation like

this

to show that they've got the

nailed

that cool

so if you know anything about this this is one of the most studied

difficulties that

in science there

are literally hundreds of papers that give

the every kind of statistic you

could you ever hope for but the most interesting one is that 70%

of college students who

encounter this in college uh uh can

conclusively prove that they don't understand

it so this is one

of these point of views is worth adiq points thing

because the form of the math that's used in college

is not the best kind of math this is a really great kind of math

for this because it's a state space math

at gets rid of all the multiplications and makes them into additions

you've got the looping adding up

the additions into the multiplications that are in the standard

formulas and it's just a simple

two stage if you

our old enough digital differential

analyzer um that does this stuff and

it's just right there and immediately the

kids use it to start doing things they can shove things off

cliffs they can uh fire water

balloons they can make uh a lunar lander

game so here's the little

two-stage thing drops the

spaceship down there

so gravity eats velocity the motor

of the spaceship

uh U produces velocity

this little script makes a flame appear when the

motor is on this is the one that crashes you if you're going

too fast so lunar lander game looks

like this so I can

toss the thing up on its

jet

down people used to spend money

for that game

and of course in the spirit of this thing

uh have to do space war and again we have

a way of doing the sketch pad

kind of prototyping so here's a here's a

spaceship with its acceleration

variable because now we're not worrying about

just vertical stuff we have to worry about vectors velocity

variable the joystick and so forth and

want to be able to make several of them we can just copy these

Green Copy button

here put these spaceships into the

Universe here we can differentiate

say this guy by

U

color and we can start them both

going right so this is kind of what I was talking about

hat you have to write in order to do this and that is the

only only program that you have to

do so the interesting thing about

most software

is that

he

mathematics you can think about what the

actual Prime relationships

are of

what you're trying to do

uh you realize they're tiny almost always compared

to the code that you wind up writing there are many

reasons for this and we should talk about them a bit right now

I want to move I just want to move you through

a few more examples and then show you some Edge of-the-art stuff we've been

working on uh and then

have some time for questions so here's a

fun program now thinking about feedback done by

these two 11y old girls the idea is

have the car uh be able to stay on the middle

of the road as a robot car if you look at this little

script you can see uh when the sensor is

touching the middle of the road color it's it's going to go

forward and no other condition

when it's touching the green curb it's going

it's touching the yellow curb is going to turn the other

way so these are

they an analyze this into three separate cases

and you can see it's quite smart because when it

runs into a sharp turn it really doesn't go forward it just

turns its way around and when questioned

the girls realized said that yes of course

this will work even if we don't draw the lanes

well because the plus4 and the minus four will cancel each other out

and so even if all three of these guys are

firing uh it'll work just fine

the kind of mathematics that you're hoping

children are actually able to do

okay here's another example

remember I said these things have costumes so

here's a bunch of objects drawn by Sam

and he's increasing the cursor

by one over and over again so we can see it

going through there and then he's telling this guy

to look like the the costume that's

down here learn a little bit about rates

by putting like for instance a two in

here

1.5 is

interesting

0.5 and the kids realize

immediately well gee that's all a movie is movie players

just two lines of this

tuff then there's

didn't suspect that when you say something into a

microphone the system gives it back to you as a bunch of little

rectangles in one of these very same holder

guys here and here's a cursor

when we move the cursor we have we have a little program

that's going to move the speaker this graphic speaker

here is actually connected to the real speaker

in the machine so if we turn this

on as the cursor goes

by we're reading the numbers here and we're animating this thing

you might say well why can't I hear it the answer

is it's going way too slow

but this stuff runs as fast as the adult

stuff so I can actually speed

it up by a factor of 10,000

Tom

Tom and I can play the same

games what if I put a two in

[Music]

here

0.5 and 1

point2

come and say

1.5 so the

the children very quickly realize

that the $300 they just talked their

parents out of for synthesizer is actually paying for two lines of

code and some recordings

and they quickly just make buttons

and the button has the the action

of putting the magic number into a variable and

over again they have a little synthesizer so

the the point here I'm not trying to get you interest of course

I am trying to get you interested in the kids stuff because not enough

adults uh realize

that the kids are going to be the ones who invent this

stuff the last 20 years have proved that adults

are hopeless at invent inventing this stuff so

we should pay more attention to the kids but

what I'm trying to show you here is that the models

what you might think of is the thep mathematical

models for things that are considered to be very complicated

phenomena in programs are actually extremely

simple so if you actually make a software system

is able to deal with those models directly and abstracts

into a different place the

uh optimizations you get a very very powerful

very very simple way of

programming okay just a couple more

um the

if we take the feedback stuff and apply

it to looking for Shades of lightness and

darkness we can do a salmon swimming

Upstream what it does is it

circles until it finds

something darker so it's continuously acquiring and

losing and it only has one little piece of memory

from the last time it it looked

in the water to get it to follow this

gradient clownish does the same thing with the

circular gradient from the Cen enemy

okay and once you

that

then you can use this these ideas that we'

adapted from star logo

which is to

this is like a zillion little salmon but they're ants

the ants wander around randomly when they find the food

they pick up a particle go

they leave a scent trail behind it's about 10,000

patches Computing in parallel here

uh diffusing the Fones

evaporating them and this is kind

because you see very quickly

uh a large percentage of the ants have been organized

by this loose coupling this is a loose coupling

architecture

even though the ants aren't communica ating directly with

each other here's one where they' run out of food

but they're still scent so it's kind of like

Street that's kind of interesting

learned about epidemics if you take

the particles and apply gravity to them you

get buoyancy this is truly

beautiful little things so they have thousands of particles

here and if you ever wondered where does the upward Force comes from that floats

things it comes from the force of gravity

on a liquid or a gas that's

confined you don't confine it it just spreads

out so it's the putting it into

something that has walls of some kind including a

circular spherical Earth

that makes it impossible for the

particles to spread out sideways they start piling

up and they start exerting

uh an upward Force that's uh

proportional to how many things that are piled on top of

them and you get this model very very nicely and we can change

the mass of the sphere here make it

a hundred times more massive so I'll change this to

thousand

and actually it's not going to sink all once it gets into equilibrium

it'll still be boyed up a little bit

it's not quite massive enough to sink all the way to

the bottom and if I turn Gra gravity off I get

another interesting

thing which is when it

stabilizes I get Brownie in motion and we're less still less

than a hundred years away from Einstein's

original paper

which was written in 1905 back in those

days they could only see the blue guys so

I one of the best Arguments for the existence of atoms was

Einstein showing the kinetic theory of heat

on small particles

um moving around would cause larger

particles to move in the way that was observed

it's a per perfect way of actually

understanding this

tuff okay

so now let's uh we're going to look at some collaborative

stuff going to go to some so here's where I want

to uh Booth switch over to

this guy

okay what

up

oh it just wasn't showing

up

okay okay just a

slight pause here

so I'll tell you a little bit about the

problem which is that

when the internet

was worked on it was trying to solve an N squared

problem

and which was how to be able to

get any node to communicate with any other node

without having to make the crossbar switch

in between so that was done by a peer-peer endtoend

way of doing things and one

with that solution was

that it uh uh

yeah you can connect me up too

one of the problems with that solution was

that it made

it very difficult to do what angelart had shown us how to do in the

60s which is to do immersive

collaboration because the problem is is that there

for n people on a network there are two to the end

subgroups and you really can't allocate a server

as engelbart's group was using

to solve this problem so this is a

catastrophic exponential that you have to

deal with so various people including Dave

Reed years ago

uh started thinking about how could you do a peer-peer

uh solution to this which

gave you the equivalent of something like EverQuest

but without having any Central server at all

and we'll talk a little bit about the solution the solution

is basically you have a distributed

replication of objects

that has uh uh realtime

uh transactions over a slow Network

I we'll sort of give you a sense of of

uh what this is maybe we could dim the lights a little

bit

and this is uh my friend Dave Smith uh

who's the guy who did the original first 3D

game on a PC long time ago and Vera

walkth through and many other kinds of things and

uh uh this system which is

called croquet was basically done by three

people with a little kibitzing from me Dave

Smith Andreas Rob and David Reed and

Dave go why don't you I'll be your slave

here hear

can we turn me on somehow just

a switch here hello

you're on good um

c as said is collaboration environment

what you're looking at right now essentially a 2d

display uh

looking at is not what you think you're looking at

back away from

this I'll back away from mine

so Alan's the the White Rabbit and

I'm I'm Alice and if you look I am the guy

on the right hand side Alan's the one on the

left

okay hello hello sounds the same to me oh there he goes

wow all right so I go over here and

show off a few things yeah I'm going to follow most of

time here I'm going to follow and see what Alice does

um this is a mirror uh and

uh so I'm actually looking at myself

and the white rabbit standing behind me but it's a mirror that can move

it's a window floating in space

uh I can resize it

uh I can pick it up and move it around

just like you'd expect I should be able to right there's nothing

special about it uh in fact this is a portal

into the current space we I can grab it too so he grabbed

it over there so essentially what's nice about is any of

any activity can be done by one guy can be done by any person

so these are essentially objects that you know are

xhibiting a certain set of behaviors but uh

we we we can use those objects in any way we want to

uh any person can use them that way here's another

object this little uh pyramid

this is actually a zarinsky pyramid so I can

actually model ify its uh

depth I'm

sorry oh yeah yeah that's

a that's that's easy actually ask you do

the time right

that's a good one to ask I I can uh modify this thing

also give

it but essentially it's

it's totally um

that's me right

another thing over here is Allen again

and uh just putting a little bit

of a just so there's no distinction between 2D and

3D first accurate rendering of my actual state

right and now we're going to go into

a uh another portal but this time into a portal

into a different space besides the one we're we're actually in

this is actually a little a martian

landscape now as I spin this

around it's like picking up a shoe box and looking around inside

of it now Alan's going to watch me walk into it so

you pay attention kind of to both of those what's going to

happen is this is actually think of this as a hyperlink on

the net by the way exactly this is like a web page

better so I just dropped

in do you see I'm inside

that world if I turn around now

and I watch Allan come in there

is okay it's a

door over here we have a little robot

that's in both of our spaces and what I'm going to

do I'm gonna get up on the Rock here so I can see you

is uh have it start moving around

so I'm controlling and giving a little bit of a velocity

and I can I can turn it's kind of like the car that Allan

showed uh it's got inverse

kinematics yeah notice the wheels are following the

terrain which I I I find very

cool and I can attach myself

to that frame or I can drive

it from behind

uh so I'm literally like sitting right behind the thing and driving it

around around

Mars uh actually yes let me show you that

I can get on it now Alan

able to see me

now how cool is that all right

I I'll get on again

there we go H something's

uh I keep hitting something no

it's uh it should I should be able to stay on it

I'm not but that's all right who cares it should work and we'll show you something

in a little bit that uh illustrates that better

so next thing

is find the way out of here oh there is

um by way this is another one of those windows this is

identical to the ones that we saw

move this around just like we did and we jump out

here's another portal to yet another space

we'll jump into this one I'll go

the other way and what we're going to see over

here is another portal into this

world what we have going on here is this

uh neat little flag now this is actually the first thing I

ever did in in squeak uh

and uh this is actually a full mesh

uh physics trans uh it's it's a mass spring

model and just to prove it that we're not this isn't the canned

simulation I'm going to remove release the top

of the flag so I'll jump in here so I can watch you

doing this and then I'm gonna start pulling it back

like it's a rope back onto the thing

now this is a lot like the particle system stuff

except the particles are constrained now I'm

actually standing on that little um so I

can grab it here and give you a little toss

so I'm I'm on the little Carousel

now spinning around

as you can see what's really neat about

that to me is that he's actually controlling my

position in the world remotely

I mean it's it's literally uh yeah it's

like he's changing my state in a dramatic way it's

just a fundamental thing that's just really

fun um but dizzy

um over here we have a a little tribute to Scott

fiser and and U Warren Robin at

the the little escalator and this escalator actually works

I can get on it no hands look I'm being picked

up and

if you recall

let me uh close this

guy

um what's next do we go close

this guy should we go to let see we should

go to the viian world

oh that's I gez I knew it was something I forgot

um this is our good friend Andreas Rob

one of the co-creators of the system and what we have

here is an underwater land now what's interesting about

jump into this

uh I I actually turned it into fish but what we're gonna do is watch Allen

do the same whoops you went to the wrong place

ah

that

now watch as you'll see Allen

uh come through see him

into a fish um

now what we're going to do is populate this there's

a few fish swimming around but we'd like our own yeah

go up go up closer up here yeah go up there

yeah if I go too close it'll the fish will show up under the

rock ground so what I'm going to do is

draw one and literally

paint package

and uh I'm just going to do a quick and dirty fish

here so this is a

u derived from a master's thesis by

Teo igarashi who's now a professor at

uh University of Tokyo

was originally called Teddy you can look it up on the net

is really very cool this is

one sort of one more advanced version of it

so what I did was I drew a quick and dirty M of a fish which

any child could and should be able to do but guess

what he's uh

inflated and what you'll see in just a

second if he's not already there

there he is both sides so essentially we're able to

collaborate I can pick him up and so Allen's moving

around uh so alen's gonna make some seaweed

so here's

a this is

a real challenge for this thing that none of us

but I think it was uh

Dave's what 5-year-old son or something thought about

making something like this and this if you know what

algorithm does this

is kind of

frightening because the the algorithm has to

figure out the major and minor axes of

things and

then once having done that it uses that as a way of intuiting

what the 3D shape is going to be so this is going to

uh compute for a little bit

Yeah figuring out what the

3D

I'm always amazed when this uh that this

works there it

is there it

is there it is a't that neat I I just think that's

one of the yeah what's what we've got is the

simplest 3D modeling tool ever built which

is uh just for me just a joy

this this thing is so much fun you have no idea when you start building

you start making yeah this stuff is scriptable just

the way the the 2D world

is okay let's chug our way out

and of course when we go out if you watch me

we uh return to our

our normal normal forms okay there you are go ahead

and I should be Alice there you are

and you're the rabbit okay now we have to

go to uh David's World right this

this this little world do kind of show

um uh a very

Hefty kind of 3D

environment there's a little bit of a

texture problem with this I just got this machine and the drivers I don't

think are quite ripe yet

but uh still you see the the environment's

pretty

rich what I'm going to do is go on top of the

um this Aqueduct I'm literally going to walk

there

and walk up onto this

and Allan should be able to see me there you

are one of things that's cool here I just

a little side comment is my machine

I'm using for this is an ultral light from Japan

so this is a two-b

computer doing its

stage of uh this

tuff and so we're basically at the point where two pound

computers can uh do all of the Computing

that uh a child could ever

hope for to your right if you just

look on the so are you

going to jump off actually what I'm going to do is one of the things about 3D

is distance doesn't mean anything uh

in these virtual environments so actually I can jump to that window that

you see way way down at the bottom of my screen

I'm just gonna just immediately go there like

so I just did a a quick and of course I

can jump right back up

hey there oh what am I doing up there

there you are okay anyway so I'm GNA jump

through the window again and out

and back to where we started

and what's neat

is I can seal Allen in before gets out

you got me in the little world

okay so let's switch back

okay last couple of things

here okay so here

a couple of things about this system so

parts of the system were done by this group of about

five people and

virtually all of the 3D system

uh that you saw there including the

network stuff was done by just three

people so the message here is besides

the fact that this is fun at least we think it is

uh and that we think especially this kind

of uh highly scalable

uh stuff that doesn't require servers to do immersive

stuff is a is a next interesting layer for

the internet that if you think of it as a kind of an

addition to tcpip as a way of handling

N squared problems it's

interesting to people who are interested in the future

of software that uh this

tuff was all done by a very small number of people

and that means that there's powerful leverage

underneath in this way of building software

whatever way it is

one one of the things we might talk about a bit is what way that is

my favorite uh

statistic here because this is a dynamic object

system um I can use the

dynamic objects as though they're a database so

have a retrieval basically a

a an expression that's asking

all the comp compiled methods in the

system to uh

accumulate their size and bytes and add them

up so this is so this will include the operating

system all the applications that you saw

the 3D stuff uh development systems and

everything this is an interesting number

so it's 2.8 megabytes for everything

so that's a thing that's really

worthwhile thinking about the number of

methods or things that are like subroutines there

is uh

uh about 50,000 which is

too many if I divide

these guys into these guys

I get an interesting number

uh which is an average of 56 bytes

per uh

average of 56 bytes per method

in the system about 230,000 lines

of code including the operating system

and we think it could be a factor of 10

smaller and one last

little yeah see this

is this is an interesting now this is a beer game

because um

um I I only mentioned this a little bit

because the one of the things that's fun about Computing

is you don't have to uh do a lot

of bullshitting because why not

just write the the code and do the

demos so so the interesting thing

is that this system is 2.8 megabytes

and 230,000 lines of code

but you know in

having Chinese uh lunches

with the groups one of the one of the interesting things is to

sit around and think about kind of what is the actual entropy

the code so for those of you

who are like an easy one that's very

suggestive if for those of you who are familiar with 3D

graphics and the math of 3D Graphics math of 3D

Graphics can practically

be written on a page

because it's highly repetitive it's

basically uh has to do with being able to do a

couple of different kinds of Matrix Transformations

and uh the most

complicated part of it is the rendering

stuff but if you write down the actual math of it it actually comes

into this embarrassingly small part

and almost everything else that you wind up writing in order

to do this stuff uh is optimizations

of various kinds so it's an interesting

question of

uh what does it actually take to prototype

things if you could write the direct relationships

and have those run fast enough to be interesting

what if you could separate out the oper the

optimizations in a way that they weren't co-mingled with the

the actual meaning of the code so U

one of our one of our uh projects

that HP is going to fund next year

is to try and take a whack at uh

this actual question of how small could

it actually be if you're able to actually go right

at the thing and uh you had

a a nicer architecture than say small talks

for doing this

um I guess the last slide

is

um I believe so software engineering right now is still

an oxymoron it just

isn't here because there's nothing comparable

to

uh for instance the Empire State

Building was put up by

uh less than uh 3,000 people in

around 11 months that included

demolishing the old Walder Historia which is on

that site so we couldn't organize

3,000 programmers to do some massive pro

project in less than a year if we had to

so there is just nothing if you use the

word engineering as it's used today in civil engineering there

just isn't anything comparable in what you know the

kind of engineering we do is more like the Egyptian pharaohs

did of making large structures by piling

up Rock and then Plastering it over with

Limestone so but I believe

that the Saving Grace for doing this stuff

even more than the abstraction mechanisms is the ability

to do late binding because late binding

has this property that when you're

just at this part

of most effort in the in the system you get this horrible

crash you finally understand

what you needed to know when you did the system and

if you're in an late binding system you can actually

go back and make those changes right now

and if you're in the way most people do

software you cannot go back you say well we'll do that in the next

system and you know what next system never happens

because that system that took so much time and effort

hat people decide to patch it for the next 35

years rather than doing this so I believe

in order to deal with a learning curve in a reasonable

way the number one thing and it's a thing that

uh uh I think the world first learned

from lisp the greatest

single idea in programming languages was

Lis for sure because of many many

but one of them is this notion of being able to late bind as

many things as possible including your meta system

and then the next thing

works for you is the abstractions

and uh maybe above all is

to take heed from what the engelbart programmers did which

is they just really wanted it to be that good

they were willing to work their asses off to make it that

good so that's a very important part also

so um time for questions any questions

about this stuff what we'll do is we'll

set up uh one microphone here one microphone

okay up

right mic

on yes uh

the question that I had um I think that everything you're saying is

right on the money I do video games and I believe that objects

should have behaviors to be able to communicate and

I've been writing books about it for a while how

do you see though changing the entire Paradigm of how everybody

else thinks to try and teach them and how long would

that take see

I I think one of the

again I love these historical references

I I just at the risk of telling a

maybe two one of the most brilliant things ever done on

personal computer was a thing done by Warren Robinette

called Rocky's boots anybody

remember Rockies oh okay

just one of the greatest things ever done on

an Apple two and what it was was a little

was basically kind of a maze of rooms

and you started off in a room and you had a couple of simple components

you had a little Thruster and a little sensor and you had

a couple of and and or Gates and the

objective was to get to wire up a little robot

that would find its way out of this room and so if

get the robot to do that then you're in another room and it

was harder to get out of that room

and you just kept on doing this and by the time you at the end

of this thing you were pretty darn good at

uh digital logic it was just a fantastic

fun thing and I have an apple too solely

to be able to run this old software

it's just great although these days you can get good

emulators so the um

a tremendous game that was derived

from this that turned out to be a failure

but just because of one slight flaw was called

robot Odyssey so the idea there it's an adventure

game you're stranded in this city and

have three robots to help you and you can program them

and the adventure game is paced a little more

like football so it's not continuous action

you can use the robots to probe the next barrier

you have to get through robots can communicate with

each other you can program the robots so you can make a strategy it's like a football

play get the robots going and get to the next

level and uh children

can learn a lot from uh doing

those kinds of programming

um so

the um the the stuff that we are primarily

interested in is what we call hard

fund so we're not as interested in

the game that's arriv

as we are in the game that the child is

a co-creator of

what's going on so like for instance

a game that we would do in this system you know other

this system is quite General you can do EverQuest

in it if you want to but a game that we

might uh do would be one where the child

has to create things give them behaviors

and stuff in order to get to the next step of the game and

interact with other people in various ways so

I think um if you look

at that slide that I did about uh what happened

with the printing press the

it was Unthinkable in the year 1400

uh when one person in 100 in Europe could

read and write and the Vatican Library which is the largest

library in Europe had 372 books

and they knew there were 372 books in the Vatican

library because you can count 372 things

accurately but

nobody knows how many books are in the Library of Congress

right so the the change that happened was qualitative

nobody expected that most people

uh would actually learn how to read and write right and

think in a different way because of this stuff so

I I so my simple

version of this stuff is that

I think two things have to happen we have to

uh help the children not invent television

when they start taking over this thing

because the biggest problem with the inventive skills of children

is they tend to be trivial so like if you give them a

piano they'll invent chopsticks on it

but it took 200 years to develop real keyboard

technique so you have to help the children there but

you don't want to help them so much that they become

mired in your ideas

the idea is to help them enough so that they have a sense

of taste and threshold and then

hope that the children will take it on to the the next

level because I don't think that we can actually

quite imagine what the next level should be our our Theory when

were here at Park was let's try and guess

what the next literacy is going to be like and see

something like it and see if the kids can take it the rest of the

way in some sense that's what I showed you today is

not what I think the next thing is going to be but our guess

at the best thing we could do for children so that

they can take it the next level

it may not be a satisfying answer but that's my

answer other questions

yes as we go back in

as we go back in the history of computing I'm always

astonished that it was the young kids that started it seeing

photographs of Bill Gates at 16 and so

on yeah and you're working with kids yeah

is that environment possible today in 2003

that some kids in a garage

will again create our next level I know well this

there are two myths there one is that Bill Gates had anything to

Computing he had a lot to do

with the economics of it but I'm not aware

of any advances that he ever made

um and the other thing is there's a myth about the garage which is the

favorite American Myth of all times

and the thing to realize is and this

is a hard thing that Americans hate

but like the two Steves in the

garage why were they able to do what they did in the

garage because of all of the phds at Intel and

Motorola who understood solid state

physics it absolutely

wasn't the way people loveed

the

garage thing the stuff that

has happened happened primarily because of

phds Park was full of phds

and we worked with children

because it was a perfect balance between the two

things but it was not done by naive

invention of tinkering things

together at all because it just doesn't

just doesn't work that way right

so it was done by people who had as much knowledge as they could

possibly put together and tried not

to remember it most of the time

but the the garage myth I think is

uh I think it was true for certain kinds

of invention a long time ago but it certainly hasn't been true since

silicon became a critical factor

and since higher level languages became a higher fact

so the interesting thing is is what the

uh the children were able to do with real

knowledge so for instance the Mac f ER

Mac user interface was actually done by Bruce horn

and Bruce started with us at Xerox Park when he was

12 and when he was six years later when

he was 18 he did the VM

virtual machine for the Dorado for small

talk which was a pretty hefty piece of work

for an 18-year-old but I I still think

Peter Deutsch's feet of understanding list

because most software people today don't understand lists

of any age so to have a

16-year-old boy be able to understand it and implement it

is to me one of the the most interesting things

ever done by a teenager in Computing

so other questions

yes um getting back to the kind of

the technology behind all this stuff um I've noticed that the

theme seems to be kind of higher and higher levels abstraction and

US simple things like scripts and sketches to leverage these huge algorithms

um underneath and so I was wondering ing um

how you see almost inevitably Artificial Intelligence

coming into this to kind of fill in the gaps even more so things like neural nets for

pattern recognition and fuz logic to make things more organic and all

that you know well I think the I mean there's two

things one is uh one of the things

think we were all hoping for was

at least uh

one of the one of the ways we thought computers could

be different than a book and I still think this is true

but it didn't happen and it's not not being funded

now is uh

um AI tutoring

because you a book should I mean the computer

should be the book that helps you learn how to read it

right it can reach out to you in a way that a a regular

book can't and uh turned out to be

a hard problem uh

with some interesting special cases we're done

but uh no General solution to it

because uh

I think in certain cases it's worthwhile doing Brute Force

versions of it like for certain things about physics that

you could spread out over billions of people why not spend

you know uh a couple of million bucks

for each little thing but that's been a

real disappointment on the other hand I think

that you know SE more pait's dictum

was uh the question he asked was uh

the computer program the kid kid or should the kid

program the computer so in our stuff

the in spite of the fact that we

abstract the optimizations away

we don't abstract any of the mystery

so basically everything is totally

you know in the sense that math is completely understandable the

that we do is completely understandable there's no

there's no nothing like an AI ghost

helping behind the scenes so the children

are very very anchored in the

cause effect relationships that they're dealing with with and they

are the ones that do the heuristics like the little so when

they they will program the feedback routines and they still know what

it is that they're doing so it's not like the

uh the invisible machine where or

uh a much worse program like Sim City

where it does things and you don't know why it's doing things

you can't find out what it's doing and

cannot change those things those things are very bad

lead to Superstition

like the real world

other questions

yes

so going back to the example

of the car so I'm

wondering about um the I don't know how

advanced the you know constraint solver or relationship

maintenance uh system is

um but how often I'm curious

how often children must have some limit I mean there are

fundamental we we don't use that at all

the programming of the car is done just like logo it

doesn't use any constraints at all right I mean they're

writing they're writing the relationships themselves right

you can think of them as oneway constraints if you want

okay so that's and do

so and and do children ever try

to get more find more advanced or or establish

advanced relationships that that bump into limits of the

system sure because the

you know the the essence of computing is

nonlinearities whether you want them or not right

are temporal some things

are batched in such a way that you don't see some of the ones

you might see in other systems and

sure and the

but uh in the curriculum

stuff that we do we made up you know like

40 Projects of which

12 worked out to be kind

of the ER projects

that are the first 12 that children generally do

almost regardless of what their age is from

10 10 on to even some eth

graders do about 12

projects establish

a couple of powerful ideas

having like the the the three

powerful ideas that the children work with

directly that are outside of the program pramming domain are

this notion of increase by

and that goes in state so that's a first order differential

equation and if you do that into a variable

that variable and increase by into another one then you get a second

order differential equation and second order

differential equations Model A lot of interesting

phenomena so those and as I showed you you can

use them in data structures as well as out in the physical

world so you can do sampling like the

first order thing gives you sampling s syis if you think

about it a second order guide gives you FM

synthesis right you have to modulate

that second guide because you have to speed up and slow

down within the wavelength

of the sample that you have but that's

basically what FM is and if you look at

it from the standpoint of phasers which is the way I happen

to like to look at things then the entire

synthesis mechanism is just piling phasers on

phasers and each one one of them is

uh basically one of these incremental

guys so so the idea is basically that

um to take a few simple but powerful ideas

and go deep on them in within

these projects uh second one is the feedback

idea which is a very nice one because it Bridges

the mechanical and the biological world

so there's a million feedback besides the gradient

followings finding lights so feedback is kind of

a general technique for making progress when

you don't have complete information then the third one

Randomness and probability

of being when you don't want to know what else to do you know

something like it's kind of like Monte Carlo there's a

whole set of heuristics built around that

if you think about that those are three if you had to

pick three that Encompass a lot of what

science thinks about the world those three are

it then of course there are powerful ideas

from the programming itself like like the notion of a conditional

the notion of a loop notion

of a variable those are ideas that the

students haven't really encountered before

um it's usually quite late

before we reveal to them that every little

object in the system is a vector

so if you plop one of those guys into its own

little thing all of a sudden you can find that you can add

these things to each other so the thing you thought so you can add

a car to a fire hydr and you

get the vector sum of it but that isn't

introduced early so you think about and this is

approach because in math you don't want to

as aam said you don't want to multiply entities

unnecessarily so the idea is you try

and stick to a few simple operative principles

that have great range that's why we call them powerful

ideas and

um but the people who

very very often run into to barriers of many different

kinds are adults because adults when

they adult teachers for instance when they learning if they've

had a little bit of hypercard they try and do what they learned in

hypercard and it's very difficult for them to

sit down and just learn a new system as a new system

they want to learn it as part of some other set of

knowledge they have regardless of whether that knowledge is powerful or not

so they are the ones whose questions we answer all the kids are

perfectly happy because from their standpoint

hey're basically playing creating their own toys

and they're happy to work within any system that is powerful

them

yes um

one thing that I get a sense of you working with children um

as adults we I guess when we get

to be 20 25 we get very serious in life

and I get a sense that by by working with children

you brought play back into life at

some level and and and that's

an important part of the whole creative process plus enjoying

life one of the things that uh I was

sensing when you showed all this I could very easily see

these these mirrors each different mirror let's say

one of the mirrors was Art and you you walk through the

art art mirror and then you have all the different types

and you could you could visually go to different

uh impressionism different types of art same thing

with music so I could see very easily this could be

whole kind of knowledge base that you could walk through

depending upon your level interest and start

playing with these things in a way and even

because it's so visual you can start connecting

these things maybe art to music to this and there's

whole connectivity factor that could take place

that in a way could make for an explosion in creativity because

we're all kind of focused in a particular Paradigm

whichever we're in I just like your comments on any well I think that's

a good articulation of what uh

the um augmentation of

human intellect Center was doing at SRI

right look look at what these guys were doing

and you can also ask Bill since he was there

and responsible for a lot of it

but basically the the idea

was to try and make connections

that when we did

this system one of the things that was

kind of both amusing and exciting

was at some point I don't know whether it was you

I think it was David because uh David

had done a demo that was one of the greatest