Is Software Engineering Still an Oxymoron? Alan Kay at GOTO 2021

here's how this uh came about actually

it was last year fraben contacted me

and asked me if i'd do a conference for gotu's

conference last year and

well i didn't know

he gave me a title which is really

scary is software engineering still

oxymoron can you answer

this in any reasonable way

and it's for people who are in the software

engineering business which really is scary

and he wanted me to do it pretty quickly

so

i said yes because

a keynote is supposed to be contextual

it's not a question of

giving you straight information about today's

technology and this is a

an excellent question i think to put context on

a software engineering conference

and i have a one-line bio

which is no one owes more to his research community which

is the arpa park research community most

of us in that community feel the same way

it's what we did within that community that counted

my tombstone will be a single

curious and i'm basically

a systems guy so

computing is part of that

but i uh have worked in biology

i worked in human systems so

i like systems and

this talk uh is a

complicated one because of the

the topic and the

field because of the delays

i gave two other

talks on software engineering in between

i wasn't particularly happy with them the talks

were well received but i didn't feel like i

covered the ground well enough

so for this one i engaged a bunch

of my friends and colleagues

who've had a lot of experience in

various aspects of software

engineering for instance gerald holtzman

was responsible for the three

and a half million lines of amazingly

workable c code that

went into the mars curiosity rover a few years

o this is a

basically an assisted talk

and to deal with this

question i think we define

engineering as designing making and repairing things

in principled ways

and if that's the case then it's certainly not

an oxymoron that is what most people in

software do they have principles

and they design make and repair things

and many of the people in this room and

in the field will say hey software engineering is whatever we

do that's maybe an overreach

so i think engineering

is the term we have to look at

uh do we mean it like

the peruvian rope bridges which i absolutely adore

like the salute or even

the making of music which is done in a principled way

you can think of these as kind of crafts

at a high level

but i think we also have to include

things like the akashi straits

bridge how big is this bridge

well that's the empire state building so the pylons

on this bridge are as high as one

of the tallest buildings in the world

and the

empire state building itself was done 90 years

ago and from the demolition of the

site to occupancy was

built by less than 3 000 people in less

than a year and i don't know today

of any software projects of comparable size

that have actually been able to be put

together and built by 3 000 people in less than a

year so we have to think about this as

part of engineering

that means uh we just

have to open up this term engineering a bit

i think oxymoron is kind of insulting

so let's use aspiration

and i'm always interested in the soft

and software because there's nothing

less soft than a legacy system

it just doesn't want to be changed

and there's nothing more soft than a buggy

system so these are kind of a

way of thinking backwards

if you will so i propose

this as the actual title of this talk

and

we want to aspire we want to

learn we have to understand what engineering

is and we need to understand the hardness and the

softness and i'll mainly talk about

the the last two

so if we

go back before the wear bridges we can

see the historical development of all of these

fields because most things start with

tinkering and we share tinkering with many mammals

so we just do something any old way

we have

an aesthetic sense that tells

us whether we like something or not and by

iterating that we eventually wind up with principles

old days and in

many cases today are in the form of cookbooks

where many recipes

it's not completely understood why recipes work

but it's known that they do and they're collected

together mathematics was like that

also and it was used to

allow a certain kind of thinking

that did not require

structures and these all

co-evolved and then just about 400

years ago what we call real science

uh entered the scene very very

late because it's a very different kind of

thing from these other four things

it has to do not just with knowledge

it has to do with the fact

that what we call knowledge is something

we have to deal with in a very noisy and messy

brain with

noisy representations and so in order to

get real knowledge out of science we have to think

about what the knowledge is in a very different way

and in the 20th century

we had this wonderful thing happen

was engineering finally

got to embedding enough science

in it very very late hundreds of years

late but finally there and

you got this synergistic combination

of tinkering aesthetics engineering

mathematics and science

try to do the social engineering needed to get

uh these different personalities to cooperate when

you have that you have something more powerful

than has ever been done on the planet before

and that's what gives you something like

the akashi streets bridge and the empire state

building now if we ask the

question is where is software engineering on this map

i think it's a little bit like this now

it's still a lot of tinkering there's a little of real

engineering tiny bit of math

and i don't count programming as math here

and just a little bit of science

so it's like where uh engineering

uh medicine uh

were uh maybe

a hundred years ago or so it's evolving

into something where it is taking

on more of the ancillary knowledge that's

needed now when i was making up the

slide i needed a good example and

ferris bueller's day off happened

i mean facebook's

day off and

right when i needed it

a very very large company with billions

of users went off the

net for about six hours

to the point where they couldn't get in their

own offices and of course this has happened

to other companies it's happened to amazon

it's happened to microsoft

but it's never happened to

the internet itself the internet is much larger

much more connected yet somehow

it has not crashed

anything like this uh i know

a lot of the people who did the design of the internet

and they were much more comprehensive than most

of the computer people i know today

did a heck of a job because they

didn't just design for success

they designed for failure and in

fact the whole way of the arpa community

designing both the internet

and the ethernet is to deal with failure

as the main fact and make failure

not be a factor in the successful operation

well i got

one of these guys to comment on the facebook

thing and he said well the facebook engineers

are far from stupid or ignorant

in fact they're the opposite of that but

he internal architecture is such that certain very

small errors but only a few can have huge mystifying

ramifications that's this does not sound good

he said could there

be a internal rule check

that says we shouldn't ever partition

the internet on purpose yes

and what they've got there is

uh configuration checking programs

but they don't have even have

what my firefox browser has

close off something it says

hey you're about to close four tabs you really want to do that

you can imagine that facebook should have something

come up like this should we shut down the routers and get off

the internet well they didn't

have that so what

what they got what they did there was

something that was a part of the job

but i think with not a large

sense of what the job actually is for instance any

any large thing that deals with humans

needs to have a modern equivalent of asimov's laws

of robotics overseeing the entire system

there should be something that understands what

a good system is to oversee the entire

system they didn't do it they

have no model of the system itself and most

uh enterprises don't do this

what is the probability that no one at facebook

understands this zero i'm sure

understand this what's the probability

that facebook can't afford to fix this zero

so what we're looking at is something else

here's a clue to the attitude they named

the street that they're on hacker

way hacking is not what you do when

building big systems their motto is

move fast and break things

that's the voice of naive children vandals

drug dealers and other arrogance

and rebranding the whole thing isn't going to help

no matter what you put up there you're

still going to get the thumbs down

so the patron saint of

all this is a guy by the name of macarthur wheeler

here he is on the on the left there

robbing a bank without a mask

or anything else and uh

robbed several banks actually and

they looked at the cameras and they

got a picture and uh in a day or

so they had found him and arrested him

in jail he was incredulous because he said i wore

the juice i wore the juice they said well what juice

and he had doused himself with lemon

juice because he had read about invisible

ink made out of lemon juice really

and he thought well if he doused himself within

with lemon juice then the cameras wouldn't be able to see him

and the psychologists

that studied this case were named dunning and krueger

and this is called the dunning-kruger syndrome

and it's here he is

a low-ability guy

with a high estimated ability

but the important thing about the dunning-kruger

people is that they looked much

larger than macarthur wheeler

much every human

being overestimates their ability

and you have the basis of greek

tragedy and theater

up here at this end where you have superior people who overreach

and as the greeks used to say

hubris brings nemesis there are

also packed little packets of modesty

so one of the touring award winners uh

from my era tony

once pointed out debugging is harder than programming so don't

use all your cleverness in writing the program

but this is what most people do

they're as clever as they possibly can be before the real

work starts einstein said a bigger

thing we cannot solve our problems with the same levels of thinking

that we use to create them

great cartoon these smug

pilots have lost touch with regular passengers like us

who thinks i should fly the plane

here's zuckerberg who thinks

i should run the company not because

of the crash but because they didn't have enough

knowledge and perspective to understand they were creating a

legal drug and the worst thing

is they were really happy when they realized that they had created

a regal legal drug

and we just had this guy in the u.s for

four terrible years and we can't get rid of them yet

but

if you want dunning-kruger take a look at almost

all software not just the

facebook software but

this is the boeing 737 max autopilot

who also asks who thinks

the plane the answer is well we want it to fly the plane

that's what its job to do is but now

this is no longer funny it's no longer a joke

because everybody here is going to die because

the boeing 737 max autopilot

does not know how to fly the plane but is still put in charge

of it and boeing let this happen not

just once but twice

so this is a terrible thing it is worthwhile

studying to understand both some

of the states that engineering can get into

and some of the states that management can get into

there are many articles about this

many forms of lying now i was in the air force

so i of course was interested in looking

into this more deeply and i found this picture of the

electronics bay on the 737 max

now i can tell you having been in the air force my reaction

was what this

is the electronics bay it looks like a trash

heap and sure enough a few months ago

boeing notified airlines that

there's a potential electrical problem because

some of the things that were in this trashy were not

sufficiently fastened and it was up to the airlines

to find them so this is an

example of artificial stupidity meets human stupidity

in greek this is happening

more and more and software is getting more and

to get a little of a bit of understanding

about this

let's think about what we're doing right now

you're out there in the audience prabhan

didn't supply a fire but

basically we're at a campfire right now

and if you think about it what we're talking

about are things that were invented because

writing and mathematics and

science got invented and yet we're talking about

them orally around a campfire for

just a few minutes whereas

everything that's important is something that

is hard to talk about orally and it can't

be done in a couple of minutes but somebody

who understands this is our cave

guy from 100 000 years ago he

understands completely what's going on

here and what he wants is a story and

i'm driving i'm trying to

make a story and that pink brain

he has hasn't changed much

so if we give him glasses a tie a briefcase

and bring him up to the presence he still wants a story

and the problem

stories most of the things we

have to deal with today have systems natures

and systems have no start and no finish

they're a bunch of processes uh interlinked

with each other but he still wants a story

so one way to look at is human genetics

has changed very little over the last hundred thousand

years as far as we can tell

there was a place where our invention started outpacing

what our genetics give

us directly our culture gather these

together and we start increasing our wisdom

from what human genetics gives us but

it's so much easier to invent than

it is to get wise that we have this

complete mismatch today and the result of that

his guy with a hundred thousand year old brain

a million times the power of his

fee of his spear and we give them a million

times the power of the campfire to organize

people so this is incredibly dangerous

and we have hundreds

of flaws not just these that i'm

showing and the result

of that is something that is very well

understood very noticeable

like the covid pandemic was

completely botched in most places most

citizens in most governments couldn't handle it

even though there was nothing new about it

as far as contagion

and mortality so that is a very

bad sign because in the background

here we have something that is much more complex

much more devastating which is the climate problem

and right now the general

not in any good

shape to have any perspective on it we can

see it just recently

so 60 years ago one of the great

people in our field an engineer's

engineer by the name of doug engelbart

decided

that we have to augment human intellect

we have to build things around

the human in the general

human intellect in order to be able to take

care of all the

problems we've had especially the problems we've caused

so he wrote this report in 1962 it's online

it's really important

o understand he basically says this explores the nature

of the system composed of the individual

and the tools concepts and methods that match

his basic capabilities to his problems

that's just what we need he says in 1962

one of the tools that shows the greatest immediate

promise is the computer when it can be

harnessed for direct online assistance

integrated with new concepts and methods so this

hadn't really happened yet but he could see

that it was going to and one of the things he describes

here which he saw

an early version of

is a kind of a

thinker of the future's augmentation

system that includes this very large three f

meter by meter display pointing

device a system in which you can give

advice to semi-intelligent

and so forth so that was the image in the early

pages of this and he realized

as with

the k person in this in the spear

that when we have a pink

brain and we're given power

what we learn from the power is power we

don't learn wisdom from it and this

gives us a rather wrong view

a dangerous view of how to use power

so the first thing he called for is

have to intermediate that with education

and training in order to get

a more reasonable balance

and power if we do that

then we can use modern methods which are extremely

powerful and modern ways of representing ideas

especially on computers and we get

quintuple which is an augmented

human then he said most

things that people do

that make a difference in the world are done with groups

and so what we want to do is we want

make up groups of these augmented humans and these groups

also so it's something

like that and then just four years after

making this proposal they gave

what is known as the mother of all demos

in which they showed

many aspects of modern personal computing

and networking today an important

here is that these were not the precursors

of many things we do today

for example uh

this looks the top pain there with

the guy talking and showing looks like what i'm doing

but in fact in the englebarts system

every participant had his

own cursor every participant

could deal with the information simultaneously

and that was built into the operating system

there's no such thing today today

is actually a subset of

many of the things shown in this demo in full

glory and their system was so useful

important that even when they

had face-to-face meetings as you can see in the bottom

they did them using the system as

part of the representation of their

ideas and their communication

so this was one of the great landmark groups

and if we again go back to 60 years

ago to one of the inspirations for englebart

we find

a supercomputer back then which is about

the size of a football field

part of the air defense system it had uh

all of 150 000 instructions per second

and all of 330 kilobytes

of memory but it was this enormous

thing and down at the end

at three o'clock in the morning is just one guy

a graduate student at mit that they

like they gave this machine to him for three hours every morning

and they

had an oscilloscope display

and a light pin on there

and here's what he did

so the display could only plot points so the first

thing he had to do was to make it be able to draw

lines uh do characters and everything

that so these rubber band

lines were his invention and look what he's

doing here he's saying i want these guys to all be mutually

perpendicular and the system

problem so you program

this system by giving it desired

goals and it will figure

out how to do these goals so again he wants

uh to put a hole in this flange

so that he says make these guys uh

parallel and mutually perpendicular

and then he says i want

these gu lines to be collinear

with the guidelines i just put in

ow he makes the guidelines transparent

and notice that this is the first clipping

window ever done is actually drawing

on it on a canvas about a third of a mile on

its side the zooming

is done with knobs

now he wants a rivet

to put into the

flange there the reason it's called

sketch pad is because of the constraint solving

you can just make a rough sketch

and then again he does this thing

saying make these guys mutually perpendicular

this will drag the center guy and that will drag the

top of the rivet

and it generally finds the closest

solution to the sketch that you made

so here he's doing modifying the sketch

and the solution

is a different one that obeys the symmetries

there but he wants something that looks like a

rivet so goes back

to this now here's another

idea done for the first time here is that

this thing that he did is actually a master rivet

what we call a class in object-oriented

programming this is one of the very first systems to

ever have objects

so here's an instance of that rivet

again the rotations and the zooming are being done

by knobs that he has so you can see the system

is rather efficient even compared to ones today

so now he's got that in there and he's showing

us he can make more instances of the rivet

hey're

all independent different sizes

he says whoops i forgot i've got those cross

pieces i should make those transparent in the master

and we can see all the rivets feel that

change

now he's got a flange with

and he says okay i want to make that a master

i want to make instances of those

okay now i want to make that

into a master

make

instances of those

by the way for each dot that the oscilloscope

is plotting there is a code word in the memory

of this machine so about half of the computer

is actually used just to do the simulated graphic

display okay

now let's take a look at the bridge

example this is a recreation of it

from his thesis because the movie

has gotten mislaid

of it but there's enough information in the thesis

that this is pretty

faithful so again you like

making the flans you hook the

thing together but now you can turn on

a downward force

which you can think of as a constraint that tries to

make the difference between everything and the bottom smaller

and

the constraint on the beams here is

a spring constraint and if you think about

it a spring is a thing that tries to maintain a constant

length and the constraint

solver here

shows the non-linearity

which is what you want to achieve in a bridge

everything ripples across everything is

connected together and

the solution here shows that the bridge will handle

these weights so

when i first saw this i saw this

uh in 66 about

four years after ivan had done it i was completely

blown away it changed my life

uh

this way of thinking about

all systems the important thing here is

that sketchpad didn't have to be prepared with any knowledge

of bridges it could do electric circuits without

being prepared with any knowledge of electricity

this is a general way of thinking about

propagate things between

their parts and

a mind-blowing thought i had was wow this

is the first thing i've seen on a computer

programming where the programming is

not like the machine code of the computer like

algol programming is like in

when you when i say algol think c today

it's really just a slightly

sugared version of assembly coat

most programming done today is like that

but what sketchpad did was to completely replace

the notion of a von neumann computer with

a computer that is a goal seeker

we have three different problem solvers

and a little operating system that

figured out which problem solvers to use all integrated

so here

he is and

so interactive computer

graphics and computer aided design clipping windowing

pointing iconic user interface

object-oriented uh design

with masters and instances many interesting

internal things like the inverted pointer

recursive garbage collection idea

automatic dynamics simulations from the

designs and this programming

in terms of what's rather than house

so you basically put in the requirements and

the automatic problem solvers find the solutions

and give you this bottom

so my thought from this in 1966

is wow all interactive computing could be like this

i mean this is kind of like the way it all should be

it all should be like this for everything and i asked

ivan a i guess the next year

or so i said how could you do all of this in just one

year all by yourself and machine goat

and he said well i

didn't know it was hard

that is where uh the

main cad that we have today started

was at lincoln labs with this system today

an electrical cad chips

are not made before they're known that they're going to work

because it's not just designing

something it's simulating it using a supercomputer

necessary you know the thing is going

to work to

know one part in 100 million or something

like that before you ever do the fad

recently

people at stanford were able to model

turbulence at close to the

atomic level and got a lot more insights

into why uh when you can see this

the nacelle on this

engine is trying to make the wings stall above it

don't

build a building without being able to simulate it

biocad simulates

nanocad simulates we've

come over to computers we've got this what

simulated punch cards

what is going on here

what happened to ivan

sutherland and sketchpad what happened

to engelbart what happened

to integrating science and math into

engineering in most

places in the country today

in the world today none of this has happened

and this is particularly

unusual because now

we know why because software people

think they should be doing fab much more

than designing and simulating

the rest of the engineering disciplines

know they need uh cad

and sim so they get software people

to make better systems for them than the software

people have themselves for the most part

this is crazy but it's actually what's

going on so mantra

here is look it's cad sim fab design

simulate built you've got to deal with the entire

system these days in order to make progress

i'm basically a researcher but i've spent

the last four decades also consulting with

large organizations that

only yielded anecdotes

and so i engaged these

friends and colleagues to give me more anecdotes

o at least have a lot more anecdotes than i

have and one way of putting

these anecdotes together into a picture is to say

if you go to an organization it will

know certain things and if you look at what

it does it will do a subset of those things

there are a bunch of things the organization

doesn't know that computing knows

and there are a bunch of important things that nobody knows

they haven't been invented yet

and joy

is when what is actually needed for a project

is something that the organization already does

those are the most straightforward ones

it gets much more problematic because really

most projects these days at least need

things that computing knows that the organization

doesn't know often the organization

reinvent those and may wind

up reinventing the flat tire

instead of the wheel

there are things that they do

that they won't use there are things that they do

that they shouldn't use this

one might say hey you should look and see what regu

what computing has turned up over the years like if

facebook had done this they would find the r1 system

that deck used for

modeling complex configurations

products this one was a

real model and it was of

enough strength to actually prevent

the what happened with facebook 40

years ago

ibm had a great initiative called autonomic

computing which took off

more in academia than ibm ibm used it for

a few just a few small things and eventually

abandoned it but if you look this

up you'll find some really great principles

for thinking about making an autonomic nervous

system for the system

that you have something that keeps the large body

of the system running

i've just talked about the need for complete

cat sim fab idea and i

just threw this one in because this is a terrific book it's called

engineering a safer world it's done by

nancy levison who is a aerospace

astronautics engineer

at mit and

this book is about systems in general

and thinking about the whole system

and thinking about what it means to do a safe

system first 75 or 100 pages

hould be mandatory reading for any computer

person because she is able to debunk

most of the fun theories about what

safety and danger are and provides

a much much better picture of how to think about this stuff

and of course

with moore's law and

requirements getting larger and larger

a lot of systems require things for success

that nobody knows yet

computing knows that

it should be doing research funders haven't

quite got it now they used to

so i've this is where i've lived my entire career

and so my knee-jerk response

to this kind of thing is something like this we need to

invent a live super high level language cad synthetic

wobble notes

the it's the right advice

but it's misplaced it took me a long

time to realize that the advice that you would give

researchers is not the advice that

works with most organizations

and the clue to it is most organizations

don't take the trouble to even learn what computing knows

they tend to stick with what they know

and this is very much like

complaints about classic engineering disciplines they're late

to the party as far as making use

the knowledge that is not just available

actually needed and here's a great

little cartoon from the 50s about pollution

that applies to a lot of things we've met

the enemy and he is us and the

problem is we don't understand us computer

people are especially bad at this i think one of the reasons

many people go into

computing is because they're much less messy

than human beings are

but there are people who are trying to understand us and here's a great

called thinking fast and slow by daniel

kahneman uh amos taversky

was his partner and died

before kahneman wrote this book and the book is

dedicated to him the two of them together came up with a

lot of different ways of thinking about things so

first here's a really simple idea

that inside of our head we've got lots of things going on

many many things and

they even got a handle on this they said let's make

an apo expository fiction that

we'll call system one everything that's

done for fast effect

and we'll use system two

for uh the what we call

thinking or slow pondering

and they had a

metaphor that uh inside there it's like

maybe a person riding on an elephant

and when the elephant does something

the person to feel and control says

wanted to do that but in fact

we know from pet scans and functional

mri scans that

in many cases

it's system one that wants to do something and system

two sees this being done and quickly makes

up a rationalization for it

so this brings into question what the term i

actually means if inside of our head we're

and in that multitude we have hundreds of cognitive

flaws that distort thinking most of which

people are unaware of so i'll just

show you a couple because they're really important to

help with the task ahead

so we take two uh things

that whose size we know could be our

thumbs it could be oranges

could be poker chips and

look at them

have one twice as far away as the other first

thing to think about is you know as far as

our brain and our

eye is concerned there's nothing out there

because all we get is photons photons

go through our lens the lens

works like a glass lens descartes showed that years

ago and so on a back of our lens

the further away one is half the size

one but inside the pachico

machine that's our brain we

have the knowledge that these two chips are the same size

so this sets up a conflict

we know and what we're actually looking at

now you'd think that what we're actually looking

at should dominate what we know but it doesn't

what happens is there's a compromise

done in this little theater that i'm calling

the dream it's the thing that

consciousness the compromise

is that the

further away one is

acts as though it's closer

actually is and it winds up

being about 80 percent the size

the dream the way we

think about the dream is not

inside of our head we think the poker

chips are out in front of us and so what

we actually see out there is the further

away thing about 80 the size of the

of the nearer one even though that isn't

actually the case what one of the things

that science is about is to remind us

that we're living in

mcluhan said until i believe it i can't

see it that's true

that's the way our perception system works we have to learn how to

see everything we see it through

modifying

our beliefs with sensory information but when

i believe it i see it wrong which is also true

so we live in a waking dream

a delusion an hallucination that is only

influenced by primitive signals that

our sense organs can pick up if we can keep this in mind

we can be semi-sane

and we don't need

uh photons from the outside because we can pay attention

to our own thoughts like we can come up with the idea

is no god but my god cost this

into the dream and depending on where

we are we wind up with a reification of it

and again reification seems like it's out there

instead of inside of us

and for hundreds of thousands of years

been no science to remind us that we're in a dream

and there are

about 3 000 religions

because like languages

the manifestations the particular manifestations

of these ideas are local

but the larger category the category

having gods is global to the human

okay here's a fun one

this hunter in

africa wants to catch a baboon

so he's poking a hole in a termite

nest here

and here's

a young baboon watching him curious

the hunter puts some

seeds

into this hole and backs off

well he's curious

a hunter is a

decent distance away so

what's in there

seeds he's got him

can't get his hand out because

the seeds he could let them go

any time but he won't let them go

the hunter could come over and kill

him you still won't let him go

so the you can find the rest and rest of this

video on youtube and i can tell you

no harm comes to the baboon

here the seeds

the seeds well we have that problem

also it's called loss aversion

we can make up

funny cartoons about not being able to

give up terrible old tools

our field is riddled with things

people can't get rid of because they don't naturally rot

they just keep on going on

and going on so some

elevated level of perspective is needed

to know that you shouldn't be using these

so if you go to wikipedia

don't do it now but you look at cognitive

bias you'll find hundreds of them

in a nice chart like this

o we just looked at loss of version

learning

aversion is related to it

because we usually have to give up something we think is

true in order to learn some new thing tunnel

vision is a big part

it's important to understand that all of these cognitive biases

in the small are actually positive for us

they help us focus on things they help us take

chances for instance gunning krueger in the

small is very useful because it

makes us less timid about trying things it's

when we try and scale these things that they become

huge uh problems

confirmation bias helps

in focusing but when i was

a molecular biologist

one of my professors

said beware you always find what you're looking for

this is a big problem in science

is you love your own theories so you

tend to look for evidence that supports the theories rather

than evidence that says

the theory is bumped most of them are bunk

and we like stories because

we're this hundred thousand year old cave

person stories are

the way we tell ourselves lies that

really get us in trouble well

i was at this place

uh in this talk and

this is the way i do talks is i actually

make a kind of a whole

territory of ideas in this case is

like 350 slides

every time i get an idea i try and figure out a way of explaining

it after i've got a bunch of them i try and

see if i can

make a campfire talk because a

talk like this in the end it has to work in the campfire

it has to be turned into a narrative

but here's the way to

think about this at best the story is an ad or commercial

for studying the real thing so i haven't

shown you the real thing i've just shown you that there

is a real thing that you can study

and the real things today are usually systems

and systems have almost

nothing to do with algorithms and data structures

and so if your first way of thinking

about anything computing is an algorithm data structures and

it is for most people because that's unfortunately

the way they're taught exactly backwards

you're starting off in a

behind the eight ball

yeah so while i was doing this i had one of those john

lemon linen moments life is what happens to you

while you're making plans and for me it was

scans not plants those scans

got me under a bunch of robots for

hours there i am

and when i was recovering in the hospital

from this big operation

weeks ago i had a

very happy thought about this talk

that is i should have been talking

about medicine using it as an analogy

because it's hard for people to look at their own field

and see anything they're so used to it

but a medicine is a branch of engineering

and we have this idea

from helping hands which is like writing a few lines

of python to open heart surgery

which is not like writing a few lines of python

and we can think of some funny

like here's a surgeon

should we ask stack overflow what to do next

happens all the time in computing

so we take the server offline meaning the heart

so

we go back and try band-aids

so my old colleague butler lampson

e of the large mines of the 20th century pointed out that

yeah the early pc and mac os were made from matchsticks

they were kind of tinkered together by people who didn't know about

operating systems but the fix was really

bad because vms and unix which

into service were not set up for

modern computing needs and

so looking at this you see every kind

of bug that we should be familiar

you better not have to ask stack overflow about

anything important

you better not confuse server with service this happens

all the time you take a physical

server offline but you have to have service

band-aids are not the solution to the most

interesting things in medicine

the way the early pc

and mac os's were done don't scale

and even

better ideas don't often scale

so medicine

says there cannot be a continuity

between the

band-aid part of our field

and the open heart surgery part of our field we cannot

stand it we have to have stages

we have to have high fences

we can't allow jumping the queue

and if we do that we're

starting to follow in what

he other engineering fields did and what medicine

eventually did about 80 years ago which

is to integrate science and math

and trying things and engineering

and the whole shebang

into one larger

discipline that uses knowledge in the best

possible ways so i'm very thankful

that i was able to wait for my operation

until today rather than having it when

i was five years old

and medicine has always had a thing called a hippocratic oath

one of the parts of it is

ays to the physician

the physicians promise to

no harm of course they've done a lot of harm

but they thought they were doing well

but this is a really good one to focus in on because if

you really think about avoiding

doing harm you start getting

a better picture of how things might be in

our field so here's a

bridge done by the romans about 2

000 years ago which is still in continuous use today

here's a plane

done about 10 years ago not by boeing

flying safely

and engineering has

something like a hippocratic oath also

the pledge of a real engineer is the bridge must not

collapse the building must not fall the

plane must not crash and now

the software must not harm or fail

and software is reaching everywhere and has the most degrees

of freedom it's the most dangerous

new set of technologies invented

and it's starting to kill people

so the engineering pledge here has

even higher quality it has to be much

much higher than what is being put into it

and you don't get to a place here

al civilization is not a place you get to

it's a group of people try striving to become more civilized

it's a process not a place

engineering is a process not a place it's people

striving to become better engineers and

i take from that that whatever we are

doing each day we also have to

ask ourselves are we also striving to become better engineers

thank you very much and enjoy the rest of your conference

[Applause]