Joe Armstrong & Alan Kay - Joe Armstrong interviews Alan Kay

[Laughter] [Music]

I actually went in and asked

Allen how they would like to be introduced

and Allan was very quick in saying oh

we're two old guys who' been around for a while

and to which Joe you know very helpfully interjects

are the ones who helped create the mess we're

in of course he's he's much

younger and

this is where you know a um a quote from

Sir Tony who was on this very stage

this time last year came up and it was that alol

was a great Improvement on its successor

and you know I think it's people who've helped

us get to where we are today who will have a very good insight

into where into telling us where we actually are today

and not where we uh who use this

technology believe we are and in a way you this is a theme of what

conference is all about and this is a perfect

way you know to wrap up so I'll leave it over to

Alan who will give us oh hello Joe

how are

you to the movie

that was such a hit maybe we should quit now I mean it's

of the best movies

okay so Joe asked me to

uh spend 15 minutes and I will try

and maybe somebody can help me

um to uh he wanted

to he wanted me to spend 15 minutes to say where we were all wrong

but um

that is a much longer talk so what

I thought I would do is uh to go back

into the past and show

a couple of ideas that

I witnessed this is not

a talk about my ideas but some ideas I

witnessed that had a big effect on

me uh I think many of them were

very important and a surprising number of them

have not found their way into Computing today

and this is partly because let's face it Computing

kind of a pop culture meaning

it's not that interested in the

past so for instance

you know we've had even some Nobel Prize winners

among the touring Award winners most people here doubt

have ever how many people here have ever read herb Simon

Sciences of the

artificial okay I rest my

case now 35 years ago

the question I would ask the audience that was

to a talk on Computing is I wanted everybody to hold up the

have with them right now

and they couldn't do it now I don't ask that question

anymore because at least things have gotten a little

bit smaller Okay so

this is uh the the fortune teller from

thoron Wilder's play in a

carnival put a quarter in and

uh if you ask her for Fortune she

will predict the future for you but she can't predict

the past and talking about

the past is hard because it there are so many

details great thing about the future is you can eliminate

all the details and ignore all

the difficulties so it's the best place

to be the present is even worse than

the the past because it's

so vivid so it's impinging on our

nervous system so if you want to think about anything at all you have to

escape the present

so uh world's

greatest computer designer Bob Barton liked

Gera and this is his favorite

Gera quote we should all share in the excitement of

Discovery without vain attempts to claim price

priority and this is a good

statement about especially how an

engineering field with some tangents of science

things there are just a lot of contributors

so we can look at some of the really great

ideas we can talk about a few of the people

who early articulated them but it's not worthwhile

trying to figure out exactly who did what when

my version of this is no one knows more

to his research Community than I do that

was really true virtually every idea

I had was extremely

derivative in the sense

of uh they came as reactions

to ideas that were already

around now

when we cooked this thing up I mentioned to Joe

that it happened that just about 50 years ago

today was when I

left what I thought temporarily

uh molecular biology and pure math

to take a year off and get a master's degree

in Computing and I did that by accidentally

showing up at the University of

Utah not knowing that it was one of the arper

research places

and I walked into the head of the Department's

office Dave Evans and before he would give you an

office he would give you a copy of this the

the sketchpad thesis which which was

just about 3 years old so this is in November

66 and this had just been done

a few years before and

I'd never seen anything like this despite the fact that I put myself

through undergraduate school as a systems programmer

for the national uh atmospheric

uh Research Center programming

supercomputers so I could program but I didn't know anything

does that sound

familiar I thought I knew something something until

I stumbled into this place saw this

well this was the invention of modern

interactive computer Graphics user

interfaces direct manipulation

a whole host of

things what Ivan called Masters and instances

we'd call them classes and instances

today and

instead of programming in an imperative form or

even a functional form you program in

sketchpad by telling the

system uh relationships that you must have

solved so this system

actually solved problems in real time and if I had more time in

this talk I'd show you the movie but it's available online

you should look at it because this

is not only perhaps the greatest PhD thesis

in our field it ALS many of the things in it

compare favorably to stuff that is being done

today this thing was just just completely

out of sight and then because the latest

graduate student gets the latest dirty task

uh mine was to get alol going

for the University computer and it turned

out it wasn't alol it was the first version of simula

that had been put on top

as a pre-processor to uh

an alol uh compiler written in machine

code and we actually learned about this system

by unrolling the listings down the

aisle of the it's an interesting

interface if you think about it if you have a 100

200 foot long Engineering Building you can run

a lot of code down there and you can get a couple of

graduate students to crawl on it and you can call out to each

other hard to do uh with

modern things and

this was a procedural system their Insight

was what's really interesting about

alol is almost nothing

except the alol block

an alol block is a thing that had

a bunch of variables so it

had a place to store State and it had

a main routine the variables could have procedures

and they thought of this thing as an entity for

computing and so one of the things they had to do

was to get rid of the alol stack

so that they could have many many blocks

around at the same time and

so what they called classes

and instances activities and processes

so to them an instance was not a

static thing that was

being more data likee but it was actually something

in in Pro process

whereas sketchpad had this beautiful relational

constraint program

simula required you to program a procedural form

but what was cool about they are not quite

objects was that you could hide really

ugly stuff inside they're not quite

objects and the objects would look nice from the

outside and

so wanting to do this you

wind up thinking about well maybe I can do

more than I can do over here

careful and like I

said I had a degree in molecular biology and another one in pure

math and so

those separate

ntities independently

in process reminded me of

cells and also reminded me of

time shared virtual machine

processes because that's the first thing you do when you have

make virtual machines out of that computer

so you can run protected computations

and that reminded me of computers on Network which

community was just starting to think about doing the

arpanet as the Prelude to the internet

and all of these things had this simple

bus structure of nodes on a

bus and any object here being able

to communicate with any other object and and of course

if I wanted to simulate those biological cells I'd make

up something abstractly like that and

so here's something we all know it's one of the

first things we learn about computers I knew it also but

I didn't really know it I knew it in a useless

way but seeing this I

started understanding it in a much better way

yes you can't go lower than a computer

if you want to do arbitrary things

so going to data structures is

meaningless you can't go lower

going to procedures it's

meaningless so I started thinking about yeah I don't want

to go any lower than a computer and what

was the problem well the problem is that especially

the time shared virtual machine processes like

Unix uh the overhead for each process

was about 2k bytes and what

that meant was you couldn't simulate the number three

the Unix process and get away with it for too long particularly

with the memory sizes back then so there was an engineering

problem here then the other thing is is if

you think about scaling and this is what biology is partly good

for we tend to think on the scales

that we live in in Computing but biology is on scales

millions and millions billions of times

larger and

once you scale out as we were

thinking about in the internet you have to uh get away

from gear-like meshing of procedure calls

to go to something like messaging so this is just

simple stuff here and the

other cool thing about tissues

is they all have the same DNA so we

have about 250 different kinds of tissues

in our body but there's one class

object if you want there's one

thing that every

one of these things holds and every

special thing that each one of these tissue cells can do

is not a new thing it's something that

is actually comes out of the

uh organism itself

and my math major was in

algebras it's one of my old books that I had

to use and the cool thing there are many cool

things about Vector spaces but

one of the ways of looking at it is what you're trying to do

in algebras is to emphasize the similarities

over the differences to get

generic things that are powerful

and worth proving things about

that's the whole point here and so I started

generic messages

back then

and then a little later I met Marvin Minsky

was hugely impressive

guy on the strength of that I got

book I commend this to you this is one of my top 10 books

I believe in technical fields in the 20th

century hardly read

at all in Computing because it's actually about old style

Computing was a class he wound up teaching

was basically 50s Computing of finite

automa and the roots of

computation but boy no

book was written more

beautifully every part of this book is it's just

absolutely beautiful and I got

a glow from it

and one of the cool things he did in there was when he was trying

what girdle did

um instead of doing the girdle

way of represent in

symol using powers of primes he made

lisp

powers of primes it makes a perfect computer memory

so why not put cons pairs in there and he made

little lisp language in there I've never seen lisp before

I learned lisp in this really weird it was

just so goddamn neat I was talking to David Turner earlier

is it just boy when you can do it that

neat you're you have an urge to do

the rest of the stuff that neat and

that got me to look at an earlier book that he'd done where

basically his students PhD

thesis I commend this book to you just for

the fun of it because the way I looked at AI back

biologist I didn't think they were going to get to

AI uh anytime

soon little Phil philosophically I think it's

definitely uh a possible thing

but what I went batshit

over in AI is they were the only

about meaning

everybody else was worrying about how to make simple

little machines and debug those simple little

machines where the AI people were interested

in systems that not only could hold larger meanings

could reflect on those meanings

and in that because I Minsky

had mentioned McCarthy the last couple of

chapters in this book are m one of M

McCarthy's earliest uh

papers called programs with common

sense and I don't want to take too much time here but you

must read this paper this is

1958 conference over here in England

and a lot of people that Joe knew were

also at McCarthy was the guy who made up the term

artificial intelligence and he partly

made it up because he had looked at

a sage air defense

terminal uh in the 50s

big thing with a guy with a a gun

to point at things on that

still around in our in our

uh uh Aviation

support system McCarthy looked at

those and he said every house is going to have one of

these and he didn't worry about personal

Computing or anything because what he thought

what the thing that occurred to him this is like getting your power

from the outside this is like getting your

water from the outside it's going to be a utility

it'll be a universal utility be like the telephone

everybody will have one of these things and there will be

national uh Computing centers

that everybody can tap into

and he started thinking about it and he realized oh

problem is nobody can

uh deal with computers on the computer's

terms we need a common sense

way of dealing with with computation

and so he proposed that there be an agent called

the advice taker that you

could have back and force with and

uh not just ask it questions but tell it things

and have it reason and there's a whole bunch of

interesting stuff in there

so

what I got from this is the future and if you think about it

going to get this stuff around to everybody future has

to have common sense interfaces the one did at Park the guey

was also a common sense interface because it mimicked

what much of what people knew about the world

and didn't try to explain how machine code

worked and the other thing which I think is a

more important one for the conversation we have later

is that the future has to

be some form of relational

advising we just

cannot continue to survive

by making very comp complicated machines by hand

in fact nature doesn't once you get into biology

uh completely different set of Dynamics

rules and

also in this book is a paper

that had just gotten written as a memo

and this is one I want to say just something about

because McCarthy wanted everything

he was a really good mathematician

so he thought about things that way which means

relationally and he thought about

things in terms of having

some sense of what it means to make a deduction

or an inference

but he wanted this had to

be a robot because it had to take on new knowledge

so it had to reason differently in the future than

it was reasoning now that got him thinking about robots

in general and the problem

is if a robot can be in Philadelphia

uh on Monday and in New York on

Tuesday that defeats normal

uh uh predicate

calculus because you have two facts

contradict each other and

so this paper called situations actions and causal

laws should be of great interest to ER

longers so a

situation for McCarthy

was the set of

states in a world

that are consistent with each other at a particular

time

we're not going to worry about

uh anything except

a situation

and we advance things by getting to the next

stable situation and he wanted to do

that by uh deductive

inferential means and so we hit on a great

idea too bad we don't use it very much today

and that is to every piece of what we'd call

data or fact he added

the uh the timestamp of

the situation this fact was valid in

got it right so this is label the labeled

States idea what this means is you can

go from situation

three to situation 4

by purely functional means you don't need monads

you don't need any of that stuff

because it is exactly like what tail recursion and lisp

does for you tail recursion gives you the

effect of evaluating all the

you're going to use in the next frame and then doing the

effective a simultaneous assignment and That

next frame you've walked into is

a situation that is bound

but it doesn't have to be bi a stack discipline

think about that so stry

by the way one of the great computer

scientists the UK has had

uh grabbed onto this Landon grabbed onto it

uh and the sad thing was is uh

from the US standpoint is people over here understood

this much much better partly because they were better in

math so this paper is really

interesting to look at and you

know we use it in Computing in a few isolated areas

it for uh

transactions on data bases

want to munge

databases but basically for most programming

various kinds of logical brinksmanship are

done to try and deal with

the fact that functions are

nice but you need to advance the state of the system

over time and McCarthy

early on had a good way of looking at that well that

o look at uh

so simulate time along with process basically just don't

let the CPU be your uh be your

clock and in between McCarthy

had written this paper in 1959

uh which is the paper on list and he did

list because he wanted a programming language for the advice taker isn't

that great so he thought of the

application the big application I want to do an artificially

intelligent agent to deal

with time sharing interaction

and I need I better invent a language

to to do that so one

way of thinking about it what I took away from this it's

all about Dynamic meanings not static

meanings and these meanings

can be negotiated it

is not remotely like programming in Fortran or C

any of that crap

and then just

quickly the other uh thing

that was going on in many different ways was

trying to make things happen by matching patterns

and doing transitions so I'll just mention

Envy in 1958

system called Comet for doing natural language

parsing one of my favorite have you seen

that I I was going to ask you a question about this paper

well yeah I love this Dan Engle said if you read that paper

it's a month of your life will vanish and it did it's

my it's yeah cuz I implemented

that yeah everybody does it's a

test did I pass or fail

yeah you passed no but this is this is

test for graduate I looking at that 10 minutes ago I love this

this is just and it was done on an 8K

bite six bit bytes 1401

yeah that's going to ask about meta and meta 2

yes yes so so this is just

a beautiful thing but this basically uh

you know the earliest thing of what is called uh pegs

today programmable grammars

and there's just a lot of this is it

yeah in its own language that is all there is

read to make itself it's so

freaking cool isn't it it is it's a Dan Engles told me

about it yeah so he told me not to read it because

if I read Because if I read it I would stop what I was working

on and a My Life Would vanish and it was completely

everybody what's cool about this

is besides this in here he he also has

the definition of two other languages he has a

small alol and a fairly large alol this paper

is only about eight pages long with a couple of it's

greatest little thing

and the thing that's great about this is it gets this

is where the rubber hits the road because it is kind of

perfect compromise between all the theoretical

stuff you might want to think about and what it actually

means in some pragmatic thing hey

all I want to do is write a translator I loved it because

it was so efficient that you could think of it as an active

interface language on objects that isn't

he way Val uh thought about it

then imp was is worthwhile looking at

January 1970 of the communications of a

ACM because irons was the guy who invented

the syntax directed compiler and his interest

was in doing languages where the procedural

headers were actually the grammar

of the of the Syntax for

that particular procedure so you extended when you wrote a procedure

you got to decide whether to use some

standard way of doing it or extending the language

adding a new kind of control structure in

there uh that that

work was mainly done in 1968 and

the amazing planner language

which was really the uh

affected uh prologue

even though the prologue people say no but it did

um

and so two big hits

out of the many here shows how messages

can be parsed on the Fly it's just

so cool and

planner was powerful enough

so its matching and the way you did matching could

actually do negotiation and to me as a biologist

doing something the size that the internet was going to be

you can't you can't know

exactly what you're going to call something at the other end of the

internet so there's going to have to be a looser

binding that where meanings are negotiated

programmers would negotiate

and so I love this just because

and and there many reasons I loved it

and then just quickly uh

Mo's law this is Mo's original public paper

in electronics

1965 lead to such wonders as home

computers automatic controls for automobiles

personal portable Communications equipment

the electronic wristwatch needs only a display to be

feasible today etc etc

so and because anybody who's can deal

in this domain and by the way back in the old days you

had to be good at both hardware and software to get a degree

in the arpa community it just wouldn't let you

out if you couldn't do both but

once once you know enough physics to understand

his argument you start seeing what the possibilities

are so here's something not made

with hardly integrated

circuits 1962 but it was the first

personal computer and that box to

the left hand side is where all the electronics is

right so but this thing this

computer is a wonder it was interactive

had many wonderful things about it you programmed it in

a kind of an assembler but it was a late bound

assembler uh these ideas

the the Advent of a 512 bit

ROM got me to

do this machine for my thesis so a

desktop computer with a tablet and windows and

operating system

and there I used some of the some of the

irons ideas and some of the matat 2

ideas and the hardware emulated bite

coats and meeting Seymour

paper and looking at

kids I didn't want kids tied to a desk so I started thinking

about the further future and yep More's

law allows one of these things to happen I thought

by at least 1998 and

that was uh well within the bounds of it

and this one had quite a bit

of Blue Sky stuff I wanted the

objects to have a planner interface because I wanted the kids to be able

to do Ai and deduction as part of their programming and

blah blah blah blah blah so these this is

the easy part at Park we got to actually build it

so translate this the dyab

book into the interim dynabook Bonk

Bonk all of 128k

bytes we used half of it for

the display so just to take you into

some realm that you may never have thought about

operating it what would be like to do a complete

Macintosh kind of interface

where you only have 64k

bytes to compute in that's

what we had and

the interest here is that small talk

72 was

nowhere like what we

wanted but it actually would

be completely understandable to people who know

erlong and the actor languages

completely understandable because it had

this idea of an inter interface grammar it did parse the

messages coming in and so

forth okay so one way

when we get down to here reality is often

inconvenient particularly to a person who's more of a scientist

engineer the other hand it's Illuminating

and useful sometimes because you actually get something

you can do things with and these tradeoffs

are the things that you have to keep

alive when you do something there's a tendency to

be proud of it and the truth

is you can only be proud of that of the fact

that you did it you should never be proud of the thing

okay and last set of ideas

here by the time software engineering of

a language gets in good shape the language has become

obsolete in its level of needed

expressiveness does that sound familiar to anybody I mean this is

this is the of it is not that

it's easy to design a good language there are

thousands of Lang languages that prove it's hard to design a good

language but man it is

really hard to make a good language run fast enough

to be generally useful and by the time you've

slaved away on all of this stuff things have

moved as Butler Lamson used to point out he was one

probably our greatest operating system designer in

the US a touring Award winner did

many things he said look the problem is is

it takes minimum

three years years to do an operating system and by the time you have

it done and successfully going the actual design rules

have changed from Mo's law

the kinds of memories the tradeoffs and memories

the kind of memory architectures you can do so every

time you start off you're starting off with much less

knowledge than you normally have in

engineering so one of the ways of thinking about it

is the best languages of an erer should be the assembly code

for the next languages

and and so here's assembly

shouldn't put algol and Fortran in there that is really

disservice to to me they're kind of the

really significant

uh Fortran uh Fortran

is amazing for its ability to

uh optimize itself in its day

now this picture is ruined by sketchpad where

Ivan just jumped into the future we haven't gotten

to yet by eliminating the bad

kinds of coding we do even when we're being good

CU he went to something where you're programming basically in

terms of requirements and if you think about

that boy if we could program in

requirements and we could debug the requirements we

would have something like what other engineering

uh disciplines have in a serious

CAD system a serious

simulation of the cad designs and a serious

Fab facility

to deal with the real problems of doing

programming Ian sort of just jumped there

because Ian Ian actually Ian was a

great programmer did sketch paded one year by

himself but he didn't care about it he absolutely

didn't care about it what he wanted was to do something

that would allow people who were designing complex

things like Bridges and cars and linkages and everything else

to just sit down and think through the design

aspects and try them out

so came about

roughly at the same time as Fortran a little

bit after but of higher language

simula uh really was

built on top of alal almost and

strap made C++

specifically to be like

what simula did with algol he said in

his first paper I want to do a pre-processor for

C that does to see what simula did to alol

and that's what C Plus+

is list spawn planner

and small talk

what maybe it was a Unholy combination

of a bunch of ideas but we were not trying to be

original we were just trying to be as clean as we

could and actors came out of that

I'm going to leave uh prologue came sort

here

prologue is complicated because I just

really don't like programming languages you can't make an operating

no come on got to be

serious about programming but I do

like some of the little programming prologue

of things that would be nice to have in a much stronger

way for instance concurrent prologue by Shapiro

solves many of the semantic

problems that prologue has then I'm going to draw the

line here for the purpose of ending this

talk but point out two fun things to me happened

in the mid 80s one of them was Linda and the

other one was erlan I remember first

seeing erlang and thinking oh

God you know and

I didn't worry about whether it was like small talk

72 or not because it's totally

irrelevant the thing was they had a good set of

problems that required

uh independent objects to be relatively

protected against each other and shitloads of them running

and the system had to be fault tolerant

you have those things you're either going to do a really

terrible job or in the llong case

it got the thinking

to what I thought was was of a level

that was worth it in the 80s the 80s

had more resources particularly as far

as memory and it made thinking the early on thought so I

thought that was really good the Linda thing I really like not

because of Linda but because of the idea behind

Linda and I'm sure people have looked at it

here so for instance if you squint

at Linda the right way you will see

something like meta

2 because you get and if you did it

did Linda better you would be making

a a kind of system for making

transactions between processes that's really

cool and if you look at the second order version of it

you wind up with something like prologue

right because Linda matching

is incredibly uh rudimentary

compared to what the matching actually could be it

could be doing a lot more and

so yay for ear long it

kept on going and somehow garer couldn't

get people to pay attention to what was good about Linda

and so what we should talk about

is maybe what should long be an assembly

code for but since you're running it now

you can talk about well I anything you

want yeah

I I think the term interview was wrong I

kind of it should have been a conversation I think sure because

um kind of

um I should tell you the story of how alen came

about uh I was using small talk

and uh I never I never

actually got got through this boundary of sort of assimilating

structure in my head so I could program I found it rather painful

use the change log to find things that I G it

and found that and it was so it was at a

sun workstation and it was so slow that

I'd take a coffee break and while it was

garbage collecting I go and drink coffee and um

I ordered the first tectronic Small Talk

machine and while

for I thought there's no point in programming small talk

because it's so slow but I built an object model

really I built a model of communicating processes the object model

drawing bubbles on the black on the Blackboard

with arrows and these were my I wasn't thinking in classes and methods these

are messages they were flying all over the system that was the

was interested in and then a guy called Roger scarel said

do you know prologue because I've made an algebra to describe

this I said no I don't know prologue so so he

he took me in his office and he said look at this this

algebra you've made is a PR program you just move a few

brackets and I went oh my

God and from that moment on I wasn't interested in small

talk anymore yeah and and and and I got this prolog

I and I I got this prologue system

thought yeah let's let's let's do a message passing

object model there and the other INF was the CSP because i' read Tony

hor CSP so I want to kindy of prologue message

passing in CSP I didn't know about actors at all then I met

Carl huitt and and he said you know said well did

you you know about act I I didn't know about that but when I learned about it

later I thought yeah we were having exactly the same ideas I think

CSP was too rigid yes I think that was

a big big problem it was too gear-like and not enough

bio biology like but yeah and then

I thought hang on I'm going to ask you a few questions I was scribbling around here

and I got this notes and and I don't have to ask you you've already

answered them yourself by talking about meta 2 and thing and I was going

to ask you ask you about Oma and and

this stuff but we we might get onto that but so now go

back the start where where I wanted to start and that was

the the idea of This was um

about 3 years ago I heard you were going

to give a keynote at strange Loop and

somebody told me that you didn't didn't happen no it didn't happen

but at the time so so I mailed

said I'd like to I'd like to talk to you on stage because I

what our paths are sort of virtually

physically crossed and I thought it' be just nice to

talk to you and I thought maybe a few thousand people might want to

fit in and listen to this and and the reason

for that was was actually an interview or

a conversation that that Jules Holland had um

Jules Holland had this late night with jul Holland

and he was interviewing Elton John and both were

pianists and and the nature of this interview was completely

different to the nature of a normal interview with

Elton John because the first question um Jules Holland

asked Elon joh was you know I hear you were highly influenced by

somebody who nobody's ever heard of well he went to

the Royal Academy yeah and and a pretty good and

John just sort of lit up like a candle and said oh

that somebody's asking me sensible questions was so I was

going to start off by asking asking you about about

test your first question this would have

the same effect but you then preempted this by

well that's that's the job of any good do

ther thing I thought I I was wondering if you're the modern

embodiment of B Cen because

he was this 17th century Jesuit monk who

who was a master of the aphorism and he wrote these books

aphorisms I don't know if you've ever read them but you you certainly should read

the the perilous wisdom for per practical wisdom

for perilous times and the Art of worldly wisdom I mean

they're brilliant I'll confess not in French but

no no but but um and and what graen

did he he was the master of the aphorism and

I think you're the modern embod embod

because the nice thing about these short piffy quotes is

I I rather like Twitter because you strip away

the nonsense if you if you you've got a quote that's a few

car you know couple of sentences long 20 words long

well so it reduces the thing to its core

yeah it's I think it's it's something that

seems better until you examine it more closely

well it's something like in so you can you can M over it and

iscuss its deeper meaning so it turns out I don't think about things

in terms of those quotes at all

a paragraph and

Page guy rather than a sentence guy

but and I'd never had

to worry about

compressed Communications until I went to Xerox

and we had periodic visits from Xerox

Executives one one kind or another

and in the middle of uh of a three-day

meeting early on there I realized these

guys were not actually

even able to remember five sentences

in a row that is

the because they're they were basically

reactors they're reactive

types and so I

start so I thought my God these

these people think in terms of slogans

I'd better start making some

up and so I started and

one of not the first but one of the first was

way to predict the future is to invent it which was

more or less yelled at the Xerox

planners in 1971 because they were worrying

way too much about what uh

uh univac and IBM

were were doing they wanted to know about

Trends and we said on Trends we're

not into Trends here we're we you know

personal Computing and worldwide networks

we're just going to do it but in

fact in our attempts to explain this

to them over the first couple of days it made no effect

so I think

there are several quotes I I I would call them Bon you in

quotes when you hear them they Bo why didn't I think of

that and and they yeah you can if you spend

some I'm I'm got

so I don't just come up with them I I think you think about it

very carefully yeah right I do the hard short

writing is the hardest writing there is difficult so so one

I liked a lot and I tried out on the Ericson management because

I just loved it so much and it said if you don't fail at least %

time you're not aiming high enough and I thought that was really nice

it is and so do you fail 90% of the time of course not

95% no

99% no the important thing about see these things are

gestures once you go into slogan

land you've gone into fiction you've

gone into theater you've

gone into uh uh gestures

you've gone and my wife is a writer and

she has a zillion books on writing my favorite

one it has the title how to tell lies for Fun

and Profit oh I've got that as well it's a great book it is

it's brilliant TR truly great book and

so the thing is is that you know in our in

our world

uh we can only lie to the computer

being stupid

that is we make stupid things that are

consistent but you know we wind up

getting a kind of consistency that is not really

aspired to or even thought about in the outside

world in the outside world the theatrical world

what you're trying to do is get them to pay attention at all and I

I got this from a book by Neil Postman

guy wrote amusing ourselves to death and The Disappearance

of childhood I commend his books to you

and he and I became friends and

I found out that when he was a grad student he used to travel

around with mclen just to see

what mcluen would do he and

another graduate student mclan didn't sleep

so very often uh you know they would the

two graduate student mclen would be in bed in his dressing down

and smoking a big cigar and the two graduate

students would be there getting and mclen

talk and talk and they said one of the things they noticed

was that when mclan was up on stage

if somebody asked him a question or somebody argued

with him he didn't bother

answering he would say well how about this one

he'd come out with another one of these mclan Zen

Cohen he wanted the master of making

these things mclan

and Neil said at some point he and his

uh friend realized oh mclen

actually doesn't care whether people agree with him or not what he

cares about is whether people are paying attention at

all that's what his goal was when he was

on stage was to just get people to even

register it because they're thinking about about

other things and being in an audience is a tough

thing if nobody's getting killed on stage

you know it's it's tough and so mclan

had evolved this uh this thing that

is going Beyond slogans and Beyond aphorisms

to things that are paradoxical

because he wanted you to think about them again

so so so one thing I'm thinking is is

that does this

kind of what I get faing

90% had to you on the

head moment I'm Wonder but for you know

failing 50% of the time it just doesn't have any

punch in slogan land

right so you have to exaggerate because

you're going into a low pass filter yes CU I I

I I I I was in in Eric

ER management meeting and and I I quoted this the typical

low pass filters manag meeting but then I said you know come on

we're doing project planning and I said look

you know I'm absolutely fed up with projects that are guaranteed

to succeed why where all the project you know can't we

can't we just do 10 projects that are guaranteed to

fail and and and then I I said you know

if you do projects that are guaranteed to succeed they're

going to be horribly conservative things you've

got to just do things that you know that's just fun stuff

know people and all the management said yes yes so

that's a really good idea and guess what they did nothing

yeah and so they never did any projects

my aism for that is everybody loves change except for

the change part

so I I the other thing you have

must have thought carefully about the titles of of your talks because

sure you know the the it's almost provocative one the the

Upsala 1997 Kyo I thought very nice the

computer resolution hasn't happened which which I thought was nice

because you're talking to a lot of people who think the computer revolution

has happened and here you are staying it hasn't happened it's a nice

catchy title yeah exactly um and you're into

that and you you and even when it when it does

happen we should keep saying it yeah when I wonder when it would happen

hasn't happened yet but um it's like the human when

does civilization happen so there's a bonk there's a there's a I don't know if it

was planned but there was for me a bonk me on the head moment

when you said every object should have a URL oh yeah

and and it already said that objects are like little servers

yeah once you I mean I think the thing is once

you think biology yes of course it's

triv that's why I say I don't take any credit for these ideas I just

woke up because I've seen

these ideas over and over again and I just the thing I hadn't

realized was the level of power they

had so in preparation for this I was watching some of

your lectures and oh God I can't stand to watch them

e in the head moment and I thought yes I shall immediately

because I said on the eling mailing list I posted this and and said

you know so every every Ling process

in the universe should be addressable and introspective and we

should all talk to these so so I expected that the be a great

enthusiastic response and everybody would immediately stop what they were doing

to implement this because having implemented everything else would be easier

yeah guess what yeah they didn't

yeah why don't they do it

well because um it's very

strange why don't you go out and do this stuff I I

said you told you yesterday make everything remotely controlable

you you have no idea how close a thing it

was for personal Computing and the internet

because you have to realize that the man's

amendment in the early 70s from the Vietnam

War shut down arpa

to an enormous degree

zerox Park only was put together because

Taylor wanted to finish off

previous eight years

and if he hadn't done that the critical mass wouldn't

have been funded to

get to you know a practical Alto

personal computer and a practical ethernet and a practical

internet cuz it it wouldn't have happened

these things try doing that how about today

Suppose there wasn't an internet today

this is years later right but think about

hey guess what I'm we're going to do something

where we don't get to control it

there's going to be no central control it's going to have

uh 10 or

uh maybe even as many as a 100 billion

nodes on the thing it's going to

be self-balancing we have to be able to change everything

in it while it's running because we don't get to stop

it it has to be like biology

and most Engineers just check

out there because

but why do we do it

well partly because the original funder uh lick

lighter known as lick was a psychologist

in 1963 wrote a memo to

the members of The Intergalactic Computing

Network

and somebody asked them

why do you call it the intergalactic Network he says

well Engineers always give you the minimum I want a

that covers everybody everything on Earth and so

I'm asking for an Intergalactic one and when the engineers

scale it down we might get it

hat memo one other thing that's important

about reading that 1963 memo

it came out of an experience that he

had which he was a

psychologist he' used was used to the

Computing uh at MIT and

he was out in on one of these enormous

Sage computers when I say enormous I'm

talking two software pitches of vacuum

tubes that's

big and

they had the first time sharing Terminals and he was sitting down

doing something and he wanted to run run a statistical

package that well he was in jovial

and uh they didn't have it

he said what this computer is connected by telephone to

the computers at MIT they didn't have the arpanet yet but they

had that and he said I've got a package and

Fortran there that I could use

said well you can't do that

so I have let me just finish that the story

and so this translated this little story was in

there but basically what he said he said look

if we actually succeed in building The Intergalactic Computing Network

our main problem is going to be learning how to communicate

with aliens and he didn't just

other people he meant

between us and software at the other end of the world and software

Computing with software at the other so every form of communications

mode between active entities was

going to be like communicating with aliens because

there would be far far less basis for common ground

and how do you solve that

so we took that to account that was a real help

in getting the internet to turn out the way it was because

doing something much smaller than what lick ligher

wanted we only did tcpip we

didn't do what the thing needed which was to

make consistent objects across the internet

this has been done since but it's not deployed

on the internet in a standard way it was a as a layer

to tcpip so there was a huge

stop once TCP IP was successful

and people uh have done almost nothing with it

use a few local you know even Google uses only

a few hundreds of thousands to millions of computers that's

nothing compared to 10 billions or so

so my complaint would be most computer

people a think way too much like their first

programming course which is almost certainly in algorithms

and data structures and that set of ideas do

scale at all forget about them

and the second thing is most computer people think about the scale of

things that they're working in now or had been working

in and they have no sense of scaling so

computer people have been unable to take care of

of uh Moore's Law and third computer people

don't care about like if you saw a sketch pad

from 1963 done by one guy as

a PhD thesis that's like the greatest thing single

thing done in our field because it has every part

what we want it's got simulation it's got modeling

it's got programming that can't go wrong what's

great it doesn't use a theorem it actually

uses an engineering Theory so it has Tolerance

on the solutions it doesn't require exact

I mean the ideas is in this thing you could do a dozen

projects starting now and they'd be

hailed as revolutionary so

you know my my general thing when I

talk to University students is

hey get off your asses and learn something

you know nobody asked you to become a computer person

but if you're going to be one try and think more

like a real scientist rather than just some

you know well engineer I mean I I started

off as a physicist and and so

you know when I'm dimensioning things I so so how you know I ask people

atoms are there on the Earth was 10 to the 50 you know so so I

just number and name every atom on the earth well that'll probably be good

enough or do you want to name every plank volume of the universe

something yeah and and then Dimension it and if you Dimension it

's going to work fine you know then you can scale it down

to one or two people if you start with

people and try and scale it up then you're going to have yeah exactly

scaling I mean scaling isn't the only thing

no but man it is Big

yeah it is a big it is a big

thing to hook in to think about the scaling

aspects even before you're trying to solve your problem

and also Computing people

don't break the laws of physics in your programs I mean

there know thing is simultaneity you know there's

backwards and forth electrons bouncing up and down and photons

hitting things you have to be aware of that you can't just

pretend that we can have simultaneous knowledge of dat in two different

it's not true was Dick fan was

once in my pool and we're

admiring the sunlight on the water and the waves

and he said isn't it just wonderful that

what we're seeing is the superposition

of a gazillion different messages

off the side and just look looks so chaotic

and yet with the right little detectors which we can make with a little

float and a little spring and a little weight we

can just pick out one frequency I mean

it's just it's just the most Wonder

superposition of enormous numbers of

things and what does it mean to separate out and isolate

and what does linear and nonlinear mean these are the

really help

uh get from where we are now into

something that actually has some life to it rather than just

being a legacy anchor so I was thinking

in in a conference like this the the normal format

of the conference is people

um they say well they had some problem and

then they've used some technique to solve the problem and then

the solution to the problem in great detail and forget to

problem is and we don't take the opportunity to talk

about problems are unsolved so usually the problems are boring

too because they they are picked up

too easily you know an easy problem that you recognize

is too much INR in 1900 David

Hilbert defined yes came out with 23

problems the 23 problems that that hadn't been

solved and in one conference he

was in in uh Paris somewhere he he came out

with 10 of them so so so what problems

do you think should what problems should we be looking

what are the most important unsolved well that we should work

on again slogans it's not big data it's big

meaning there's a

Logan so so one of the ways

this kind of thinking like I say you're in a different

world you have to get out of the Practical

world for a second you have to get into

extremism crazy world

so crazy world is you know the fix

the idea fix a

of I'm going to be a monomaniac

on something and you know and suppose I'm just

extremely

sensitive and in

pain that is one of the sensations

that people like me have

basically pain

and trying to

find the sources of the pain so

the so a simple

is

this wonderful thing about language which also

should help people understand about language every sentence

you can make up in English you can also

toss a knot in there

and so even a simple natural language

like English is bigger than the universe

you can say things that aren't there and this

should bring home Einstein said to the extent

that mathematics uh

uh gets that truth it does not get it

extent it gets it reality it does not get it truth

he understood the complete difference between

the two and the problems of using a symbol system

for something that isn't like the symbol system

in any important way many scientists

get confused on this because they tend to think of the math

as the reality and it's absolutely not

so so to me the the first thing

is uh well yeah let's just negate

everything let's take let's write down the top

this would be a good thing if we had this thing at the beginning of this

conference which I was kind of hoping because we give

people something to do here

that like write down the top 25

beliefs in Computing and then put a

knot in front of them and see what that is

just the hell of it

big data not big data

well if it's not big data what is it well big meaning of course

it has to be about meaning the problem is the Big Data

people are confusing some of the means

to the end as do an end

and this happens all the time in Computing we

turn the sub problems into the problems

get mired in the sub problems

so so just going through the things that

you personally believe I do this maybe every three

to five years I sit down

spend a week writing down things I think I

believe and I spend a fair amount of time trying to see

whether there is any reasonable reason to believe them

still some of them I have

like uh I mentioned the situations

actions and causal laws thing because I think McCarthy was getting

at one of the Prime cruxes

in Computing which is that you want to be able to

advance State and you need to do it absolutely safely

so you'd like to be able to reason perfectly

about how you're going to advance the state and you'd like to organize

system so you get the best at both worlds monads

is too much of a clu

that's trying to save functions and missing what

McCarthy had already gotten to uh more than 50

years ago that kind of stuff drives me crazy

it really does it's like where did you people

come from how blind

can you be why why are you treating this like a religion

well Bob Barton the great computer

designer when uh I had him as a

professor and one of one of

the things he used to say was systems programmers are high

Priests of a low cult think about

that so he had a he had the other

thing I kind of worries me is is I I get the impression that

an awful lot of the things we have are are done because

we can do them not because we need for I call that inverse vandalism

yeah making things because you can it has

unpredictable side effects yeah

consequences so so one of the things that worries me

is um this silly idea of the cloud and just you

know if we if if we just store everything in the cloud

the the the unintentional side effect of that is we'll probably

lose all our history and I'm kind of worried like

years time we'll look back and say oh just sorry we just

lost the history of by the way another guy

of all these things invite Vin surf here

next year and talk because vent is very very concerned

about this concerned and has been doing a fair amount

to try so so so what I wanted to ask you were of

I mean I think losing history I think hijacking

our attention systems by by bombarding us with crap

yeah of of the things which do you think are the most dangerous

which should we address first kind of rank them in

order of city so we can knock off the bad ones

first so I had the

unsettling uh two days ago I had tea with a young

Oxford professor of

philosophy nice

guy and U somewhere in the middle

of it I mentioned

uh Francis Bacon's four Idols

from this book he wrote called

the in English it's the new organization

of science new or

Nova Morgan 1620

he'd never read a book a word of bacon

this is the philosophical structure

that underpinned the actual invention

of science which really happened here in England because

scientists don't give you science

it's something like the Royal society that gives you

science because science needs other other people to debug

you have to have a social system in

order to debug your ideas it's not

enough just to have so Archimedes was a great scientist

he Greeks didn't have science they didn't have

the thing and bacon pointed out in

this book a lot of things but one of the

things he said look

the simplest way to think about us humans is we are born with Bad

Brains and

our Bad Brains are partly because of genetics

our bad thinking is partly because of genetics partly

because of our culture partly because our languages are terrible at

representing what we want to have as ideas

and we have Academia that's

constantly recycling bad ideas into the

into the young and he said we what

we need he said in this book is a

set of heris for getting around our Bad

that's what his his definition of science was

it wasn't just to find out

uh what forces might keep the moon in the

sky and the planets around the sun no

what bacon saw science is as a

set of methods for thinking more clearly

in the best ways that we know how so

when you think of it that way that should be the center of any

schooling system in the 21st century

and it certainly should be a big part of any

kind of schooling system for people who are doing technology

which because of the IND Industrial Revolution we can do

this inverse uh vandalism of making

something and spreading it by the billions and the problem is our

systems are overwhelmed by

things that seem to be an environment we start accommodating

to Total Environmental things like television

and and the internet

so I dare say most people in Computing are not at

all aware of this stuff never have read a word of mclen

have not looked at cognitive psychology

design a user interface to save their ass

or any of these things so what we've

got is what I think is

a to me one of the

most beautiful things that humans have come up with

I just love

computers and B we've got

a basically a

no nothing culture who's trying

to do things with the computers

it's just to ad hoc it's like if you wanted to be a physicist

Middle Ages you only had to get a pointed hat so

so it me because they didn't know anything I I the

last time we spoke um we had this very nice

conversation it was simpler there wasn't an audience yeah there weren't

people there and just as you left we started

talking about Douglas Adams and I didn't know you you work

with Douglas and then I thought well Douglas Adams he came up with

like the Babel Fish and things like this you stick it in your ear and

was thinking you know do we need computer people to figure out what we're going

to do because half of the you know this Internet of Things

Douglas would have loved this all these completely

stupid things that people are coming up with yeah

and and and sensible things like a Babel Fish so so what was

your relationship with with funny

Douglas and I got along uh you know we met

a long long time ago when somebody

told him uh Did You Know The Hitchhiker's Guide to the

Galaxy is kind of like Alan K's din book and

so he so he got interested in

by the way uh if you don't know the

story U Douglas was quite a drinker and so

was I back in those days

um and one evening he'd had

a ton he was a big huge

guy he's like 65 66

really nice and he just sort of wandered

out in Cambridge and

asleep in one of the greens

out there and he woke up in the middle of the night not knowing where he

was looking up and there are all

these Stars up there and

a phrase popped into his head Hitchhiker's Guide to

the Galaxy you didn't know what that

meant but he started thinking about what

so he would have loved the B I mean there there is a b

fish in almost I the

what's interesting is that

so Douglas uh Douglas

and I uh you know my wife and I used

to stay at his home and we're here he used to stay at

our home he he and

his daughter came for for dinner just

two weeks before he died suddenly

terrible thing he played basketball

didn't feel well went to a doctor the doctor said you're okay he

didn't test him next day he was dead

49 years old that was bad

really bad

and the thing

about so M here's what said said look

most people think they live in

reality they're completely unaware that

everything they do is completely conditioned

by what they believe about the world this is easy

to prove to people but most people short

circuit it's actually part of the what Buddhism

is about is this this

distinction between efficiency and

what's actually going on and

so the

problem there is

you are aware of

what's going on you can see

the present as kind of a construction it could have

of different ways and

mclan said said most people can't see this at all

and so people as he his phrase was

people are driving faster and faster into the future but steering

only by looking at in the rearview mirror

that's a beautiful one because

they assume that uh the president of

some form of reality but he said artists can see the present

construction that is what they were reporting in their

art great artists are all look at what

guica is just to take something relatively recently but

they are seeing what is actually going

on because they have this pain

their pain comes from reacting to what's actually

going on and they artists are people who have to report

that to people but the cool thing

about artists and Douglas was one of these

is and Douglas did have a lot of pain

from this stuff but what

what a person like this does is if you

can see the present as a construction

you can have much better ideas about

the future because

what is an idea

that isn't incremental an idea is escape from the context

you're in well what is a joke

a joke is taking you down one context and revealing

your in another so the simplest form

of a creative act is a joke what is a

discovery in science it's

looking at the same world that you saw the

second ago and all of a sudden you see some relationship

that was right in front of your nose it's a joke you start

laughing what is art

art is the reminder that

there our brain is so limited we can hardly deal

with even getting from one moment to the next and

art forces us to realize there's more there

and Kesler who wrote a book called The Art of creation

uh the Act of Creation I'll commend it to

you pointed out the emotional reactions to

jokes is Haha to science

is aha to Art is

ah and each one of them is a

different kind of emotional disar discharge you

get from escaping from the context that

you for a few seconds ago you thought was reality

so this is so what I believe so and Douglas's

communicate with anybody

of course it takes a special person to do it his idea was

always do it with humor yeah sure so like the most bitter

book he ever wrote was called the last chance to

see it was still quite funny

but it was basically about how we're

uh killing off species species and how unique

each of the species is it's actually a plea but he wanted

to get people to the end of the book and so his

theory of writing a pmic or Kurt vaget

Von Kurt V every book that Kurt vonet wrote

was an attempt to deal with the fact that he

had been a prisoner of war of the Germans in the fire bombing

of Dresden which far

more people died than in Hiroshima it was one

of the worst things the allly we did it the British

and the Americans we incinerated 300,000

people by just

uh making a con config

vaget had to clean it up so now

we've got all the I mean we've built these billions of computers

and it strikes me we're doing all these stupid maybe we shouldn't

have we do on the other hand have severe

nvironmental problems and things and we do need to plan the

resources using these things so so in fact you know

there are I I I collect uh

sort of good news story because so many bad news

stories you know what people people with Parkinson's disease

somebody made a spoon that with your handshakes

of little accelerometer and a little computer it so it shakes in the opposite

direction these are brilliant devices why aren't

we building these things instead you're like

these right this was

this was one of the first declarations that

uh biology is not Destiny

that's what human beings are about biology is

some of our destiny but it isn't all of our destiny and it's up to us to

figure out what to do with the part that we can't

change biologically and a lot of that is through education

you know from the ideals of the people who are my heroes

and I didn't talk about engelbart I didn't talk about many of the people

I talk about in longer talks

well engelbart had been in World War II and

his his thought was man the world is in a complete

mess

uh and particularly after the German concentration camp you know people

were sick about this and engelbart

was I guess 19 or 20 and he de decided to dedicate

his life to helping humanity

and eventually this turned into real

realizing just as you say

that a lot of things that

positive effects can be done by people working together

and the computer should allow people not to

work by themselves but to actually coordinate large Plant

because we can simulate things we can deal with

threats and everything else these were the ideals of the 60s

that were the mother's milk for graduate students

like me well guess what

it happened it just didn't happen to the general

public virtually everything being done in science and engineering

today is just like what we thought

was going to be for the whole world

but only the scientists and the engineers took it took

took it over almost no scientific thing

is done today without having things that are far

far beyond classical mathematics so nonlinear

uh models of very very complex

structures similarly in engineering you know

bridges used to collapse like the

Tacoma Narrows Bridge one of my favorites so so Franchesco

is kind of making signs which I I think

s that the guys here might be want to have a beer

learn I thought you'd learned how to ignore those

signs I'm I'm not even I'm not even looking

in that direction

so I I think perhaps we should draw it to an end yes I actually

a dinner party but

I would uh you know we're here now

so I just like to thank you for coming you're welcome it's a pleasure

thank you [Applause]

[Music]

[Applause]