Alan Kay: Normal Considered Harmful (2009)

From Viewpoints Intelligent Archive
Revision as of 06:11, 6 December 2017 by Ohshima (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
this is a kind of a fortress set
kind of the professor against the  students
but the cable is short so
I  have to duck back there periodically
you all for coming out I believe  I'm the second prize
in the raffle
  this talk is
is about some of the ways I
think about our field it's not  particularly technical
it asks a lot of  questions
it's an attempt to be critical
  without being a rant and
it won't
be  successful unless I can get you
to  interact periodically there'll
be a  couple of opportunities early in the
  talk and the title of the talk
deals  with two
properties of frogs
one of them  is
that their nervous systems like all  nervous systems
are mainly interested in  differences
but unlike
us they are able
  to habituate very rapidly and
so it is  amazingly true
that if you put a frog in  a pot of water and heat
it up very  slowly they will not get out
because the  the difference
is that their nervous  system are detecting are
too small over  the unit of time that their nervous  systems work
on and they
will just stay  there until they cook and
of course we  aren't like that
and the
other  interesting property of flies which i  think is
one of the most amazing
  properties and it was first
discovered  in a landmark
paper called what the  frog's eye tells the frog's brain
and  you may or may not know that the
retina  in your eye is acts
  part of your brain is actually extended
  out from your brain and your eye  actually
does a little bit of thinking  right
in the eye for efficiency reasons
  for frogs their eyes
do a lot of  thinking thinking
that actually triggers  behavior and
one of the most  interesting things about a frog
this if  you take its natural food which your  flies
paralyze them with a little  chloroform
but peep them alive and put  them
in front of a frog it will starve  itself to death
because its conception  of a fly is
not that if you
take oblong  shaped pieces of cardboard
and throw  them at the Frog it
will eat them until  it's stuffed because
it's built
in  pattern for food is an
oblong thing  that's moving and
there's no limit to  the amount of cardboards
that a frog  will try to try to eat
so one of the  ways of
looking at this is that the  frogs both
of these conceptions of  normal that the frog
has are quite  harmful for it
under certain conditions  so it considers
normal food to be oblong  things that move
and it considers a
non  change in temperature as
far as it can  detect it to be quite safe for itself
so this is the famous Pogo cartoon
true of so many things about us  humans
trace back some problem
and very  often the
trace is back not to physical  nature but to human nature
and I think  this is the
problem in many fields a  little
bit later we'll talk about how  the biologists have
dealt with some of  the same problems that we have in  computing
first so here's our first  opportunity for interaction
who's that first guy Newton
who's the  second guy
Darwin and
the third guy the  most famous face
on the planet
who's  that first guy
who's the second guy
nope  who's
the third guy
who's the fourth guy
  and who's
the fifth guy pardon
nope so  first
guys John McCarthy
  never heard of him
these guys are all magic
people in  our field so
John did many of the most  seminal things to
this day in the
field  of reasoning and programming
languages  second guy is Ivan Sutherland
he only  invented computer graphics and
in a way more interesting than most  computer graphics
today probably if
you  had to pick the most powerful
  groundbreaking distance
from zero thesis
that has ever been done in our field  it's his thesis and if
you don't know  what he looks like  then
I'm sure you don't haven't read  that thesis and
Engelbart so
what did  angle Bart do okay
what else did he do
kay so when
I was giving talks in 2004  this is the question
I asked every  audience at every University
everybody  had heard of angle Bart in the mouse but
  most people don't realize that
Engelbart  was the prime
conceived of what we call  personal computing today and
the prime  conceived er of what
we think of as  hyper
linking ideas together over a  network with
collaboration between  people and the prime conceived
ER of  what collaborative work would be like  and
by the way he built all of this with  his research group in the
60s and it was  demo to 3000
people in 1968 41
years ago  so an interesting thing here
is to just  ponder the
fact that even though you've  heard of
Engelbart's name if you
had  taken the trouble to type e ngle ba
RT  into google within the first three
hits  some of them you will be doing it right  now you'll
find the bootstrap Institute  which have the
75 seminal papers that he  and his group wrote
and you'll also find  the videos
of that what
mother of all demo  in 1968
and yet in fact
the internet and  personal computing were invented just so  you could
do that and so now ponder
fact that you do know the name of a  famous person in
our field from the past  and yet you didn't have the curiosity
or  the energy to do that simple little  typing
job so if this were physics you
  would all be dismissed from the field
  because physicists would not tolerate
somebody being in physics and not  knowing
quite intimately what the  pioneers
and developers of physics  including Newton
and Einstein and so  forth actually did it's
not an optional  idea but in computing
seems to think  that it is so part
of my thing to get  you
to ponder is I don't think computing  is a real field
it acts like a pop  culture it
deals in fads  and it
doesn't even know its own roots  and worse
than that it does not know  about the really good things that
done in the past for instance  Engelbart's conception
if you were to  look at it you would see far exceeds  what
Tim berners-lee was thinking about  and
his demo far exceeds what you
can  even do on the Met and the web today and
  why can't we do it today because the
  people who set up the web did not know
about Engelbert and did not take the  trouble to see if
anybody had been  thinking about this so this is the mark
  of a pop culture this guy is the world's
  greatest computer designer
designer  among other things of the Burroughs be  5,000
and that machine implemented in
  hardware what we would call a byte coded
interpreter for a higher-level language  today today
we can only do it in  software because Intel and Motorola  don't
know this is a good idea  the bharden
knew it was a good idea in  1961 and designed
and built with borrows  a computer that
could do that and a lot  more so
you can't be in computing and  not know what Barton did and
not know  who he was and this guy
  was part
of the for funders
at ARPA that  funded the fundamental
research in the  60s and he was also the
guy who set up  Xerox PARC his name is Bob Taylor and
as  they used to say in the old days
no  bucks no Buck Rogers so
part of the  reason the past
actually was quite a bit  more successful
in the present is  because the funding was done
better and  people who knew how to run research
  groups like Bob Taylor you
had to do it  so I knew you
were going to give that  response because
okay how about this  picture
yeah so both of you
were  right so iliac one
was a johniac
the von  neumann princeton
computer and was built  here at the university
of illinois and  they actually got done before
the  Princeton people did so there are  several of these
machines built all of  the same plans and the Princeton
people  are a little bit late
about that I
just  love this computer
this is one of the  first supercomputers to ever be done  happen
to be done right on this campus  called iliac 2
and it had the
register  pipelining
idea even before the IBM  stretch did
so this was a really serious  computing
machine done on this campus by  the professors and
the students kick ass
how about
play-doh what was played  up
it wasn't exactly the first  graphical interface
but so this is a  system that was built
by dom Bitzer and  colleagues
here at the university of  illinois is one of the largest
funded  computer-aided
education projects in  history
there are a lot of people worked  on this and
give you the scope of it  2000 terminals
connected to large  control
data mainframes and
  flatscreen displays on these terminals
were actually invented here at the  universe of Illinois so they could
make  flat screen display terminals and why  did dom
Bitzer want to do it because he  wanted to back project colored
slides  they wanted to have something that could
  do multimedia education so when you use  one
of these terminals in the top of  these things
was a Kodak slide carousel  and
so when you went to learn something  you get a slide
carousel that would have  the high resolution images that were  going to be part
of it put it in there  and then you had a 512 by 512
plasma  panel display this
is the first flat  panel display in the world
that was  actually practical done right here at
he University of Illinois and with  cooperation from Owens
shouldn't  this makes you proud to be in
University of Illinois computer science  department right you
should be because  this
place is one of the leaders of the  world but
guess what you
guys haven't  done a supercomputer for a long time
you  haven't done a project like this for a  long time
you're using vendors computers
  and vendors software and guess what you  can't
invent the future by using vendor  stuff it's all
looks back in the past  the best of it and most
of it is  exactly the wrong way of looking at the
  future so Illinois like
most places like  UCLA where I'm on the faculty
  most of the major universities
country have given up the future in  favor
of something else most places in
this country when it was really hard to  design and
build computers built around  yes
well no I don't
think so because this  assumes that there's something good
  about what you can buy from vendors and
  I think
when that's the case it's  worthwhile buying
from them but in fact  the
vendors are struggling like mad  themselves because
Moore's law marches  on and they
don't know the first thing  about doing parallel computing for
example somebody needed to be doing that  a university
is the right place to do it  so not
enough has been done and
most of the great inventions come out of  a kind of a
something as more like a  university culture and by
the way there  is a technology called fuel programmable
gate arrays in which regular students  like
you and me can sit down and make
a  thousand processor
supercomputer by
  writing not a lot of pages of code  because
the state of FPGA is of in the  last year
and a half or so has gone  through a knee of
the curve and is now a  serious contender for
allowing people to  do huge multi processing
projects by  themselves if you're interested
we can  talk about that a little bit later
me years ago my wife who's a writer  and an
actress and an artist and so
she  had one of these huge Mac cinema
displays on her desk and she showed me  this
display and
was kind of
odd over  here
she had all the applications
that  she used that were WYSIWYG she
really  liked them and here all the app the
stuff that she does on the web and none  of it was WYSIWYG
at that time and
she said in these apps I can see and
  do full WYSIWYG opera authoring and
the  web browser has all these modes and
I  have to type through a keyhole and I  have to
wait to see whether what I did  was ok and all these
stuff and and why  is that she
asked me and I said well  it's
because the stuff you like was  invented in
the 70s and the stuff you  don't like was invented 20 years later
  what do you think
about that how
could  that happen it's not that
people don't  make mistakes
but you like mistakes to  go away
yes and it's
quite  interesting
of course where was this
web browser was done
by  University of Illinois students a little  bit after
that age of people that made  playdough terminals
and supercomputers
o in 1993
  here's mosaic and
one of the interesting  things about mosaic
was was it remotely  is nice at what angle
Bart had done in  68 and it was full
course they didn't know anything about  it Tim
berners-lee who did do the web
at  CERN was quite chagrined
when it was  pointed out to him just
the sweep of  Engelbart's ideas he was very
upset that  they did what they
did the way they did  it and even
more interesting
five years  six years before the web
was HyperCard  at Apple which
is the perfect model if  you think those of you are familiar with  HyperCard
it's the perfect model for  what a web browser should actually
be it  had full WYSIWYG editing
  right in the thing
and anybody who  understand this at
all would have  realized oh yeah this is the perfect  thing
it already has hyperlinking in it  there's an editing model
it's been  tested on four million users we know it  works
  nope a bunch of hackers
got together and
one of the ways to think about
this is  Engle Bart invented the wheel and Bill
Atkinson with HyperCard invented a  better wheel
that's really great and
and  unfortunately
I can't even give the web  browser
the flat tire award and
or even  the square wheel
because you can imagine  improving both of those to
good but you know when you invent a  broken wheel
there is it isn't
obvious  what the wheel is or
it doesn't even  work at all flat tire you can you
can  run it at the expense of the tire
figure out what the tire should be in a  square wheel you
can run it and figure  out it should be rounder but a broken
wheel won't work  and so attempts to
fix a broken wheel  produce more variations
of a broken  wheel and we've
had 16 years of them
ok  so
let's a man here you are today
let's  suppose you're at a computer
science  those you had an undergraduate degree in  computing
from back then and
you're  faced with this idea
that yeah there's  this internet that's going all
over the  world and
it's going to be not just
consumption mechanism for stuff that's  already there but
everybody is going to  be an author and a publisher and
so how  would you solve that problem
an idea  what
would your approach be to it  knowing what you know today or
even what  you would have known back
is a complicated problem right  there six billion
people in the world  there now a couple
of billion nodes in  the internet there
is zillion everybody  wants to do something yes
is how much work do you want it you want  to solve this problem really
well but  how
much work do you want how much do  you think your browser should actually
  know in order to solve this problem  really
well because
think of everybody  has a different interest in the
web  there's many
kinds of media you don't  even know what kind
of media is going to  be invented and for any given kind of  media
there  like video there are dozens of different
  codecs and just think of the what
problem is so how would you approach  this problem
well but
suppose they want to write  machine code because they know how
to do  something really really fast
suppose somebody wants to implement  their own codec
but you're you're  partly partly there but
I'm talking  about letting people
exercise their  creativity and
sometimes they're going  to need the
most every resource on the  machine to be
used but
of course we  can't allow them to take
over the  machine so what would you do
so it's closer it's
almost a good  idea after 16
years including see Google  has some real interest
in this because  Google
did v8 because they couldn't
  stand to have the normal conception of a  slow
JavaScript and they're
doing the  thing that you just mentioned because
  they realize that
people who have ideas  also
so ok
this is an interesting  thing
it really should be like an  operating system kernel right
  operating system kernel controls address
  spaces that can confine computations  completely
therefore you can actually  allow
any binaries to come down and be  used you can
completely control what  goes in and what goes
out and all of a  sudden
your browser is simply something  that allows
bitmaps canvases that you  give to
these things to write on to be  displayed so
this it's there's nothing  you have to do because
you don't want  it's like the
the the good
idea in UNIX  was hey
we don't want a big operating  system we want
the tiniest kernel we can  have and then we want to use
address  spaces in order to protect everybody  from everybody
else so this is this is  operating
system 101 this is well known
  all the way back in 1965 and 1966
should be the first thing that occurs to  anybody so I figured somebody
young so
  this occurred everybody in the 60s and  70s knew
this is the way to do it we're  all surprised it wasn't done
this way  but I figured somebody in
the current  generation would have figured this out
  and so I went looking on the net using
  Google looking for downloadable binaries  and
stuff like that  and I
found this grad student in Cornell
  in 1997 Oh far Erlich
son who is from  Iceland wrote
this paper in this paper  if you just type in
the title of that  paper you'll find it's an HTML document
  that he put on and on the net in
1997  and his problem by
the way was hey I  want to run a Kodak I want
Kodak and I don't want to have a system  bin worried about
whether it's an  executable or not so it's this thing  should be like
an operating system so of  course I
wrote an email to him he's at  the
University of Iceland now but on  leave to Microsoft up
in Mountain View  and had
a long correspondence with him  on the
pathway of these ideas and
he  started a company that Google
acquired  and
is using a couple of the ideas but  interestingly not
all of the ideas in  this paper
which actually solved the  problem very nicely so
this is an  interesting thing where it
wasn't that  somebody didn't understand what
he problem was and what the right  solution was it's that
the larger mass  was
quite happy to plunge into a de  facto standard
even though most of the  original people to
use the web we're  computer scientists they didn't protest  that
this was a very bad way non  scalable
way of doing it  they just plunged in and sixteen
years  later Google is and other companies are
struggling to actually use the internet  the way it was intended to be
that's  pretty interesting to me
now another  story is
35 years ago the
kind of  personal computing we think of today was
  done at Xerox PARC in 1973
of the paraphernalia we have today  was done in about ten
thousand lines of  code Smalltalk code
written for this  purpose and
so another
question you  could ask is well jeez
the browser is a  terrible
design but it does have  JavaScript in it and
it turns out  JavaScript even though it's slow
today's standards it's faster than an  Alto was back
in 1973 so hey why doesn't  somebody
just do a whole personal  computing system in Java
  like everything Windows
interface the  overlapping
idea is two and a half D  graphics like every
damn thing in it and
so basically what you're doing is just  drawing a line and saying
this stuff is  crap but
it runs and I
should be able to  write 10,000
lines of code so
the basic  idea take the rubble that's the browser
  make an arch out of it get
the guy who  did the job originally at Xerox PARC
Dan  Ingalls to do
stronger classes in JavaScript make  a better
graphics system make the
widgets and stuff to make an application  development system and
and an  interactive
development environment and  if you want
to try this out I
don't have  time to do
it but this is what it looks  like running completely in a
browser  took about five people six months to
do  could have been done any time but
nobody  knew how to do it has
who had done it before in the past  because nobody
knows how to do a system  like that in
ten thousand lines all
right so this is a craft that was lost  because
people have been used to just  building on the stuff that's
there so when this new thing came along  it wasn't anybody
around who was working  on it that could just say okay
we want  to have something like the WYSIWYG
interface and authoring tools and  everything else that we enjoy
on the our  regular personal computers and we want
o be able to do all of the stuff in the  browser so
is called the lively  kernel it was done while dan
was at Sun  you can look that up so
here's another  way of looking at this is if
you take  the
minimum bar I would say the general  practice
today is below that minimum and
  take a qualitatively better bar
I'd say  that the best that was
done by the ARPA  park community like the Internet and the
thernet and personal computing and  stuff like that was better I
think the
  best knowledge of today is better
than  that it's not that everything
is back  slid absolutely not it's
not that the  people are
not as good as they used to  be I'm convinced
that there are more  people of high
ability available today  than there were 40
years ago there's not  a complaint about the people
but I'm  complaining about the outlook
of the  that
produces such poor results
now the  problem is that the real bar today
is  above the best of what
ime ago and it's above the best  knowledge today this
is the problem this  is why companies
are writing millions of  lines of hard to deal
with code because  the ability
to scale is potentially
  there yet you have a
question okay the  ability to scale is potentially there
but the knowledge of how to do the  scaling is not
there by anybody so the
real way of looking at this is we need  something like that
today  and we
have to think figure out  something that's a little bit above the
best knowledge today that has a  stair-step built
into it to get up to  where we are also
things are just going  to get worse than
they are now so this  is a very very difficult
problem it's  been fomenting for years I
you are aware of it in one way or  another I'm just
I'm pointing it out not  so that you believe
it but just so you think about it  yourself
back to a couple of ideas so
  here's the printing press appearing in  1450
and whenever
a new idea appears you  get two things
you get news
news is  stuff that's incremental
to what we  already know this
is why you can tell a  news in five minutes
hey a train just  crashed we
all know what that means so  the
reason we want to know is we're  excited by changes like that
but we  don't have to have it explained
to us  there's no epistemological hurdle
we  have to go through to understand that  and
so the original idea for
printing press was to do what monks do  but do it cheaper
and there are a couple  of other side
things but when ideas
are  really interesting is also do
  and knew is
by definition not like
what  we already know so
there's no news about  new there's
nothing you can tell  somebody in five minutes about
what new  is and
for the printing press was a
  huge change in Outlook for our planet
  and one of the changes was
believing to arguing in a new and  special way and
it took 150
years to go  from the Gutenberg Bible to Galileo
two  hundred
years to go to Newton and three
hundred plus years to go to the American  Constitution
but one was the driving  force for
the these other things now  interestingly
about 50 years after the  printing press there
were people who  were writing in Europe like Erasmus
indicated that they understood  completely what the printing press
about and what was going to happen from  it but
the larger mass of humanity took
  150 to 300 years to actually gradually
  change their thinking around to start  thinking that
ok Leonardo had a high  IQ but
he couldn't invent a
single  engine for any of his vehicles so
if he  had an IQ of 500 and you were born in  10,000
BC you're not going to do a lot  you
can outwit everybody before they  burn
so Henry
Ford was not nearly as smart as
Leonardo but he was born in a better  century so
he revolutionized  transportation
so knowledge almost  always Trump's
IQ and
with knowledge is most of us know a lot  but it's
not it's not good knowledge
  might even never have been good  knowledge
but sometimes it was good  knowledge and now it's bad knowledge
so  knowledge is this double-edged
sword  very powerful when
we have it right and  the thing that gets it right for us is
  when somebody is able
to change the  outlook the fancy word
is  epistemological stance the way we  actually
look at things and Newton was  one of these people
so he changed what  knowledge actually meant and
he changed  what engineers
and more practical people  could do and so
the line there is point  of view is worth ad
IQ points you go
from a weak way of looking at something  to a strong way of looking
at something  that's like getting an extra brain and  Newton
is a perfect example because  before Newton the
smartest people in the  world couldn't do things that a high
  school calculus student can easily
do so  calculus is like getting
an extra brain  and it makes you much
much more able  than a genius
who doesn't have it
so these outlook changing things are  really critical
so I think of knowledge  of
silver outlook as gold and IQ as a  lead weight
I believe one
of the biggest  problems with computing is in
a sense we  have too many smart people
it attracts  cleverness and
you can do clever hacks
but the clever hacks don't
scale well  and it's very hard to build a
into software it just stays around  forever and
so what what's actually  happening
is kind of something like  building a large garbage dump
that makes  it the odor of which makes it
very hard  to think about other things so
if we  take news
or normal we can think
we can  solve problems avoid obstacles
beg every  once in while we
have an outlaw thought  but
we went to school we went to church  we
have parents we live in a world of
  normal and for most human
beings the  world of normal trumps any weird
new  kind of idea
try having one when you're  going
out for tenure
think of all the  different reviewers who
already have an  idea of what computer science should be  about
it's crazy
and people start  winding up gaming the system and
doing  lots and lots of little papers that mean  nothing at
all in order to get
the  approval of these reviewers whose
ideas  about computing might be completely
off  but every once in a while
in an  unguarded moment like taking a shower
getting up in the morning or something  you
get a Kirpal
not that this kerpow is  true most
ideas are mediocre down to bad
  but at least you've got a new Kirpal
and  these career paths
are the things that  need to be tested out science needs
as much as art because you have to get  ideas from somewhere
so when you have to  be in a state to
get these cows and then  you have to have the tools to be
able to  discard the Kirpal if it turns out to be
  a really stupid idea after all
  so think about this is
if the pink plane  there is reality
then what do people  think the blue plane is
like insanity  you
are insane
and learning
something that your brain isn't well set  up for could
require almost as much  creativity as inventing
place because you have to invent it  inside your
head and so
you can say that  normal is the
greatest enemy with regard  to creating
the new in the way of  getting around this
is you have to  understand normal not as reality but
  just the construct and a
way to do that  for example is just travel to a lot of  different
countries and you'll find a
thousand different ways of thinking the  world is real
all of which are just  stories inside of people's heads that's
  what we are to normal
is just a  construct in to the extent that you can  see normal
as a construct in yourself  you have freed yourself
from the  constraints of thinking that this is the  way the
world is because it isn't this  is the way we are
changing people's minds is hard  because of this
so I'm going to give you  just two simple dimensions
so this is  the whole
human race just
5% of us are  intrinsically interested
in ideas 95%
of  us are interested
in new ideas and tools  just
so far as they contribute to our  current goals
so once we have a goal we  tend to be fastened
on it you can  imagine why so
mething new  we
immediately evaluated in terms of our  existing goal structure it
doesn't fit  into our existing goal structure forget
  about it 5% of us
though are get  interested in whether
this is an  interesting idea or tool intrinsically  so
let's take a different dimension
this  is one that has been
well studied about  85% of us generally
do things for the  approval of others who
are social beings  after all
people use
the Internet  very much for confirming
disconfirming what they should be  thinking about by finding
out what their  friends think about so
this is  completely normal to human beings about
  15% of us
have many of our motivations  be internal
so if we do something that  we like we
don't really care what other  people think and
to the extent that  these two dimensions
are independent you  get
something like this and
so 1% of us
things for their merits and we
  don't care what people think
you can  imagine what do those people
do with  their time and the
other extreme is  eighty percent of us are
instrumental  and
our goals
are primarily determined  by whether other
people approve of them  so
think about what these one-percenters  actually
do and think about
what happens  to those ideas
when they try and get  society to change
and then
the other  extremes are outer
ideas and tools and a  very
dangerous group of people who are  inner and instrumental
there are a lot  of managers and politicians
in this  group
dictators of countries and so  forth
here's another analogy it was an  old one and
has been shown to be a good  one that when it
rains it's somewhat  random as to where
a little gully starts
  but the gullies have the property
of  being self optimizing once
you get one  started they are more efficient at
getting water through them and so they  erode faster
and so pretty soon
you've  got something human
memory is rather  like this
basically once we get  comfortable about something
  we are extremely homeostatic
about it we  love our gullies
and they take a long  time so
a little bit like the Frog in  the pot occasionally
our gullies will do  us in because
we can get awfully  comfortable and then get surprised
by  something like AIDS which takes a long  time to
really show up and
we humans have a hard
time forgetting  things we don't have the men in black  stick
for when we have a bad idea
or  learn something that isn't actually not  good
for us we can't erase it and so
have this interesting this is something  that's been noticed in
programmers there  very large percentage of programmers
  think like their first programming  language for
the rest of their life no  matter what programming language they
learn afterwards if you look at their  code it looks like that you
what first language they had because  that was the first
time they got fluent  those were their paradigms for thinking
  about problems so
human beings are  actually quite anti learning
in fact  most of these things are
built into our  genes
for example we are a coping  species
not a product not a species that  likes to progress we're
basically social  we all have language
so more than three  thousand cultures have been
studied by  anthropologists and every single one of
them has about 300 things in common and  so
the determination is that these  things are not things
that were invented  but are actually
built into the genes a  very human  so
once you've and by
the way if you  look at these things you can realize if  you want
to make a lot of money build an  amplifier for any of
these using  technology and you'll win
see why so
you  know when television came out the
movie  people didn't think it was going to sell  but
they didn't realize it was a fantasy  amplifier that was going to go in  everybody's
home of course it's going to  win
but they couldn't understand it for  what
it was and
you can take a look at things that have  not been found
in every culture and a  couple of the interesting
ones are at  the top
progress as it was called
18th century is an invention  it's an idea was
first articulated not  so many hundreds
of years ago because  back then was the first time when people  realized
that the world that their  children are going to die and
to be different than the world that they  were going to die in and
so the people  who wrote the American Constitution even  suggested
that there'd be a new  Constitutional Convention every 50
because things are going to change  fortunately they didn't
carry that out  they were unusual
compared to our  current puppet politicians and
the  things in
blue they're the non  universals are harder to learn the
ones  in the pink because they're not as
trongly built into us they all had to  be invented and
so an accurate picture  of
us is that we're pretty much
cave  people with briefcases and
there's  actually a group of people who study us
  as cave people with briefcases called  behavioral
economists so
if you've  looked at so one of the books is called  nudge
by Richard Taylor another great  book is called
predictably irrational by  Dan Ariely
these are people who study us  as we
are and why we actually make  decisions even against
our best  interests and
most of these decisions  are made because of
things that are  built into us and this
would be a  humorous picture except that the cave
guy with the briefcase has an ICBM in  his briefcase
and a lot of other things  that are
too easy to invent for  technology and the
wrong kinds of things  for K people with briefcases to
be  carrying around
a lot of what's going on today in  technology
in general not just in  computing is basically for
marketing  reasons we automate the pleistocene
so  if you look at what
is being sold and  used on computers there's almost nothing
that wouldn't be completely recognizable  to a cave kid
there's hardly any  new media on computers
it's all  imitations of old media and most
has to do a storytelling of one kind or  another
little a little bit has to  do with those powerful
ideas in blue  okay
here's another thing I think you're  all familiar with that in
a short period  of time change appears linear
often  that's just because you're looking
at an  exponential curve at two finer grain
so  these are big surprises for many
people  so
for example with silicon
in round 65  we had Moore's
law and that
allowed us  to have a very different look at
what  was going to happen and we use the
Gretzky theory of hockey anybody know  what that is
yes that's
right so they  asked him why are you so much
than everybody else and he said well  everybody just goes where
the puck is I  go where it's going to be so
this is a  good idea because it takes time
to  invent things so if you invent things
for where the puck is you're going to be  behind by the time you
get done so
what  you'd like to do is to know
what that  technology is going to be out there
and  the wonderful thing about computers
is  you can get that
by just spending money
that's all
it is that's the easiest  thing for Americans to do we just love
  money and we make tons of it
so for  example that's what the University
of  Illinois did when they did iliac 2 they
  wanted to get five or ten years in the  future in
computing so they built this  supercomputer neuter enabled
probe into the future and it cost  millions of dollars so
what if they
built something that was not as a  compute supercomputer
at that time all  they'd learn is obsolete stuff
there was  not going to be usable later
so a park
  we looked at about
12 years or so to the  late 80s and
said we want a computer in  1973
that will compute like the  computers are going to compute somewhere
  between 1985 and 1990
those cost
$22,000  for us to make we had to invent them
build them we built about 2,000 of them  and
so 22,000
in those days was worth  about 80,000 today right
so it costs  80,000 bucks to have a 1989
Mac but if  you had in 1973
just think of what you  could do
that can convened all the stuff  that the 1989
Mac was going to run 1989
Mac would have to wind up being like  what you did
because nobody else would  have the time to
invent this takes years  to invent good things
so when I come to  a university
and I see students using  current day laptops it
drives me crazy
ou're in the past  you're
not even in today and
just think  about the scope of the stuff
that's  already been done you need a lot more  leverage
the way to get leverage is to  get a ton
of computer power and write  very simple
non-optimized languages
that  run very very quickly
and you should be  able to do them every couple of days
  to get power you should be able to just
hrow them away use them like Kleenex  because
it's rather easy to write things  like that if you don't have
and the way you don't have to optimize  is make hardware that
runs a hundred  times faster than what you've got today
  so unless you do that you can
only do  incremental stuff today
so a lot of the  incremental ISM today
is the lure of a
  cheap computer when it's actually
  standing a lot of people's way okay
  and so
I think the last set of ideas  here is about
problem solving and one
my best friends who died just a couple  of years ago it's
one of the great  problem solvers in
fact was awarded the  mechanical
engineer of the century award  in the 20th century  it's
one of greatest awards any engineer
  can do it this guy was completely
magic  about solving problems
so BAM powered  flight is
an idea that's been around for  a long time and
so Henry Kramer's idea
  was well I'm going to offer a large
  prize like $100,000
but people
tried to  win this prize for 40
years and couldn't  do it
MacCready his
brother-in-law  wound up with
$100,000 debt and Paul
  being a nice guy he decided to assume it  to
make his wife happy all of a sudden  Paul was a hundred
thousand and dead and  he was driving across the Arizona
desert  with his family one day
and he had been  the world sail a sail
plane champion a  number of times it was a fantastic pilot
  this is driving along looking at a hawk
  circling in the sky across the Arizona  desert
and he suddenly realized that he  had seen the
exchange rate of the  British pound that morning
in the paper  the current exchange rate the
40,000  pound Kramer prize was
worth exactly a  hundred thousand American dollars
so he  started thinking boy if I could win that
prize I could play off my  brother-in-law's death so
he started  thinking about how could you actually do  man-powered
he said the problem  is
we don't understand what the problem  is really good people
have been failing  for 40 years so this must be hard
so his  idea was forget
about man-powered flight  first
thing we have to do is to try and  understand what the problem
is you look  to see what does why aren't
these people  doing this so what happened is people  were building very
elaborate complex  designs
for man-powered flight and they  would take it up and they'd
have a crash  it would take them a year to rebuild the  thing
so they're getting like one flight  every 8 to 14
months and Paul said well  we need
to we need to be able to do 12  crashes a day
so his first design was  this
contraption here that was literally  made out
of baling wire and some plastic  tarps
and some aluminum struts
and stuff  but you could fix it in a
few minutes so  in just a few months
he had flown more  had more flights and more crashes
all the rest of the people in this prize  put together
they started
getting an  idea of what the problem actually was
less than six months after Paul took
that drive across the Arizona desert  with his family they
won the Cramer  prize it was easy
just easy
much more impressive than that
  was the second Kramer prize which
is  flying from London to across
the English  Channel and he won that a year later
one  of the most thrilling videos you
ever see is called the gossamer  albatross you
can order it every child  should see it
it's one of the great  romantic videos
of our time it's done  beautifully
but so this was a
mile-and-a-half  on land and this was 22
miles over water
and when asked about it
  Paul said well everybody else was
trying  to make an airplane
said we were  trying to do a human
powered flight so
he didn't start off with an airplane as  the idea and
started off with the fact  that we had a third of a horsepower
at  max to
run this thing and we had to do  this and that and the
other thing we had  to find out what that meant and that  wound up
extremely different than all of the  other things and
bingo let's
take a look  at one
way of looking at this
so here's  the effort and
we can think of reward  going
up that way and there's
kind of a  threshold there
so here's what tenure committees
and NSF  loves
they love problems you can do and
you know what they don't care whether  the problems are worthwhile or not
  because Congress doesn't give a damn and  can't
understand the difference between  any and Congress oversees NSF
I'm on I  can say this because I'm on to several  of
the NSF advisory committees but  basically if
you want to get an NSF  grant you
don't write it for science you  write it as an engineering project here  is
how we're going to do it here's where  we're going to be successful this is
why  almost nothing funded by NSF over the  last twenty
years has been interesting  just hasn't been and
these big  inventions that we use today we're not  funded
by NSF at all they were funded by  our PI
the first thing we have to get  rid of is this
idea of looking for the  keys under the lamppost
which is not  where we've lost them
all right
you can think of there being a curved
  something like this that you put effort
  in if you're lucky
it might have a peak  like that and
the McCready principle is  spend
a lot of time finding
that tiny  little place there
it's just above the  threshold
of being interesting but
way above the threshold of being  interesting so
you're looking for the  minimum
thing that you can do that's  qualitatively
better and most
good  problem solvers are incredible at this  they have
a homing instinct McCready  would spend a
long time thinking about  where he should put 80% of his
effort  before he did a darn thing and he
never  had a failure and he solved an enormous
number of problems in his lifetime just  because he worried about
this because  this is where you learn about
what you  need to know to do the real deal
as soon as you get to that doable  thing you've
actually changed this curve  and you
brought the thing you wanted to  do into
range and all of a sudden you've  got
to break through so most of these  things are kind of two
stages effects  the hardest thing is to get the
funders  to pay for that first stage because
it's  not cosmic so one of the reasons
AI  research has not done very well the last  25
years or so is because the new are
  Pro DARPA is not interested in funding  actual
progress towards AI they're only  interested
in funding the robot  combatant of the
future something that  nobody knows how
to do but they can sell  that to Congress
and nobody can
here's a couple of punchlines Susan
  Sontag I love this one
all understanding  begins with our not accepting
the world  as it appears this
is the hardest one  this is where
you have to do with this  exercise I do
it every day literally not  getting
normal considered harmful  because
of course we have a zillion  mechanisms we don't want
to have to  think every time you take a step right
so you can cripple yourself by  questioning everything
that you do yet  so you but on the other hand every
once  in a while you know take
fifteen minutes  and instead of doing meditation on a  flower
which is also a good thing  meditate
on all the assumptions you're  making about the
world that you're just  taking for granted for efficiency
  reasons and try and think of which ones
where the world would be very different  if you made different assumptions
  reason even people who are good at this  stuff
have to do it is because we all  have the same brains
our brains are  defective nature didn't
make those  brains for doing great inventions it
made those great brains for survival and  coping
so we have to get around our  brain
science of course is one of the  great ways of getting around human  brains that's
what it was designed to do  but individually
we have to get around  what's wrong with our brains
so it's  best if we think of ourselves as
actually being unsane and that we're  dynamically
trying to be stable with  respect to all
that stuff out there Red  Queen said to
Alex to Alice  try to understand three
impossible  things before breakfast
we have a blind  spot in our
eye which if I had many
audiences I do this with I get prove  this to the audience
by having them do  the experiment where
a dot this big on  the paper will
before your eyes and be filled in by  your brain
by material gathered from the  surrounding stuff
because there's big  area in our retina that just doesn't  have
any  optical
neurons and
the realization  that we have that blind spot in our  brain is
filling in that helps us be  saner
and I guess the one
to leave you  with is this idea that if you realize  you're in
the pink plane then you have a  chance of escaping
it but if you think  the pink plane is all there is there
will be no thought of escape because how  can you escape all
that there is thank  you very much
- any questions
comments yes
well I you
know it's
so you get old they  give you
gold medals and the proper  response to
getting a gold medal is to  thank the funders because
they just gave  you gold back before
you turn the gold  medal so a good funder
who is willing to  back
something without trying to  rationalize
it as well as the scientist  is going
to do is the most critical
  thing in our field because it
takes  money like I said in
the old way that  Buck Rogers was this guy who lived in
25th century so the old saying was no  bucks no Buck
Rogers and so what changed
in the 70s the thing that forces the arc  spark to come into
existence was the  great ARPA funding of the
60s was put  out of business by Congress because of  the Vietnam
War it was no longer  possible to
do that stuff in  universities funded that way so
Taylor  decided to take one more shot
at this  and convinced Xerox to fund about
two  dozen of us to finish
dream of personal computing and  pervasive
networks so that's the way  that worked and
the funding has not been  good since
1980 there's a lot
of money  out there but the money is tied to
the  perceptions of the funders as to whether  you're doing something
worthwhile or not  and that is missing the point
  the real way ARPA did it was to
say hey  we're willing to tolerate 60%
failure if  everybody is working on
really important  stuff because that 40% is going to
  change the world so this is a baseball  model
we're not going to fire you if you  don't bat a thousand
now if you bat
368  like Ty Cobb did that's good
so that's  the way they did it and back
in the Cold  War Congress didn't over over C
ARPA the  way they do now is the Vietnam
changed every  and not
for the better their so that's  the simplest explanation and
what but  once this thing winds down
getting all of these factors that I was  showing it's our tendency
comfortable  it's
our tendency to say hey if
I can  buy a compiler I'm
not going to make one  even though that compiler might
back some thinking I need to do about  programming languages because
compiler I get I'm only going to compile  the thing kinds of
willing to compile it already has a  conception of
what it's trying to do and  that's going to
tie me to past  conceptions and I might have a good
idea  so Park literally built
every bit in  every atom of
all of its computing  machinery
build all the hardware and all  the software and it wasn't
you know it  was a lot of work but we realized
it was we thought it would be a better  trade off if we could pull