FutureLaw 2021 Opening Keynote: There Oughta be a Law

From Viewpoints Intelligent Archive
Revision as of 05:17, 19 November 2021 by Ohshima (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
so allen k is
something of a legend in the world of computer science
and artificial intelligence uh in his early
work at darpa and xerox he contributed
evelopment of object oriented programming
raphical user interfaces we all
use today so in recognition of his work
award in computer science the draper prize the touring award
the kyoto prize so
so we're uh extremely honored alan
and and really pleased to have you here with us today
and to give us your perspective on computational
law so with that i'll turn it over to you
thanks very
much i'm going to try sharing my screen here
and uh
if people are seeing a strip
of uh
other windows besides me down the right hand
side that's going to interfere with some of the materials
i'm showing but anyway i will just
start and i will let the
people who are running the
session uh decide
what what to do with that
thank you very much for inviting me uh
i'm a little uh rocky today because
had my second uh vaccination
shot yesterday but uh there's nothing
like the adrenaline from fear in
giving a talk
and
this thing started off
when mike contacted me a little
over a month ago and
he asked if i could do a keynote for this conference
and i didn't know what
this was about and should have something to
do with computers legal system society of the future
and all of
these uh interesting people in the audience
most of whose uh
professions i'm not very aware with about
and then he wanted me to do it very quickly
so of course i said okay
why not and
i'd like to i'd
like to thank the people who helped me on this talk
ahead of time because it gets lost in the shuffle at the end
so you see their names here
the connections from
mike to katy and mark steffik through tim
miller to bertram mala were particularly
useful thank you all
and of course i thought of this old cartoon
being an old guy
this is a series of cartoons that
were essentially gotcha cartoons
uh people say
one set of things and all of a sudden it turns out to be
worse or different uh later and it was
eries was called there ought to be a law
and the first thing
was asked about
this talk was well there are these nine people
who have a lot of
influence in what's going on
they're all lawyers they all have top
degrees from everywhere they're all really smart
in many
minds inside of our skull manifested by our brain
we could have as many as 80 or 90 of them
three that were articulated
uh some years ago that have been particularly
useful is a way
of thinking about things and way of doing things that
is through touch
of doing things and
another way of doing
things is through
our visual system and i include
in this are our hearing system they
both are basically configurative
and then
we have this very recent symbolic
system that's probably only a few
hundred thousand years old
now and these uh
not just these three ways of doing things
but are maybe 10 or 15 or 20 different ways
doing things conflict with themselves patron saint here
is jerome brunner
and
our theaters inside
of our heads are mostly not human
most of the way our brain
works are in common with other
primates and mammals
and this is not the way we think of ourselves so we're constantly
[Music] fooling
ourselves i'll i'll pick the patron saint here as
daniel kahneman from thinking fast
and slow very renowned nobel
prize winner uh it was basically
a behavioral psychologist
and then
we are immersed inside the play that
from the cultures that we are in
and it's the nature of our genetics
for us to become part of that play
especially when we're children but
all ages when we
go to another place we start acculturating
and we're much more shaped by
we think and i could pick a number
of people here i picked marshall mcluhan
because he was particularly interested in
how the media that we're more and more embedded
in now is
a environment
and a culture itself and is affecting us
uh very deeply i
put a halo on uh jerry brunner here because he's sometimes
called saint jerome and
one of the largest minds of the 20th century
this guy was interested
and extremely well versed in
many areas he was one of the founders
miller of cognitive psychology
but his deep
hobby was also anthropology and he really
spanned the gamut of
human culture in so many ways
so these five dimensions and i'll these
some of these characters will come back again
in this uh in this talk
and uh well this is puck
what fools these mortals be and we can see the vulcans visited
earth at least 400 years ago
and what what did puck
mean or shakespeare speaking through puck
well we're easily fooled we want to be fooled
we pay to be fooled
we fool ourselves and we pay to fool others
o
the world that we actually live in
is much more of a generated
symbolic world
in which uh story
is the central element rather than what might actually
be going on out of the world
here's an example
mother
uh wants to crack some nuts
and her nine-year-old son
utcracker so she starts grooming
because
at some point she can demand
of him to groom her
it says come on
he starts doing it watch what happens
here as soon as he gets both hands
over there
he grabs your lock
very happily
mammals are con artists here's
a dog that looks like like
he's in desperate
state
when somebody
stops
tosses
out some food and whoops
o another
aspect of this theatrical approach to
life we have is we project our beliefs
world so the talmud has
a really good line about that as we see things not
as they are but as we are
this i think is the key here for
trying to deal with uh
modern inventions modern
world that is full of us
ancient human brains
computer people could help
things tremendously if
they were less interested in computers and more interested in
people however the psychology of
computing not a hundred percent
but it caters very very strongly
to people who are uneasy with people with uh
human beings the computers uh
seem to be uh much less threatening
so computer people have
traditionally had a really difficult time
understanding
the idea that humans are supposed to be using these things
so here's an example of projection
we take two poker chips here
and we organize them
so that one is
twice as far away from the other
if we look at the image on the retina as descartes indeed
did by scraping the back of an ox's
eye after the ox was dead thank goodness
what we see is the lens an
animal lens actually works like a glass lens
and the image on the retina of
the further away one is half the size
of the closer one then
we've got a kind of a pachinko machine called a brain
here and this
information from the retina goes to lots of different
including a place in the brain
where we believe things and one of the beliefs
about poker chips is that they're the same
size and so this is a this
information that's and it's on the retina
and another part of this complex
that we have inside of our heads
call the dream it's basically a theater
where
many many different parts
of the processing in our brain
we're not aware of are consolidated
together into something that
internally is is what we treat
reality and so
the compromise here is between
a half size on the retina and
the same size knowledge
that we have and so the dream decides
that the smaller one is about 80 percent
rather than 50 the size
and then we have this interesting thing is the dream
is actually what we think is out there
and what we think is out there when we actually look at this
we can do it with our thumbs too
but is actually out there we think
something that is uh
80 for the further away one
that's what we see when we look at our thumbs like this
uh artists get around this
by measuring by being very
careful about the angles that are
subtended and so forth but this projection
of what our beliefs are
and our conclusions out onto the world as though
it's actually happening out in the world
has gotten our species
trouble over
hundreds of thousands of years it's one of the major
things we have to think about whenever we're doing
anything that involves either
ideas communication or power
and mcluhan had
the great line for this it's not seeing as believing
but until i believe it i can't see it
is the way human perception actually works
you have to learn how to do it and your learning
might not be perfect so
the thinking fast
and slow book kahneman
proposes uh what he calls
uh two expository fictions
meaning they're not true
but pretending they are really helps
thinking about uh something much more
complicated so he just divides our entire
mentality up into six system one
which are all of the mechanisms in their
designed to go really fast
quick reactions
done by correlations
of the kind that machine learning does
they're very simple they're not cognitive
the system one system does not really know what's going on
and then there's this system two
which is not a very good thinker but at least it tries
it's slow it's pondering and makes up
reasons for things and one of the
metaphors that kahneman used in this book was
uh our our our
mentality is a little bit like somebody
riding on a on an elephant
and sometimes guiding the elephant but sometimes
the elephant just wants to do something and
what system two does is to make up a reason
for it as though it had come up
with the motivation for doing it so if the elephant
rears up [Music]
the person the the writer
of the elephant says well i wanted to do that
but in fact a
careful brain scan studies show that much
of the time system two never got to think
about it system one just decided to do something
and system two noticing it in this theater came
up with a reason for it so robert
heinlein once said we're not rational animals
we're rationalizing animals
here's one
of my favorite condiment
biases this guy
wants to catch a baboon
and you can find this video
on youtube if you want to find out why
here's a young juvenile baboon watching
with interest but he's digging a hole in this termite
yield there
and he's putting some seeds in the hole
and then he backs
off
primates
what is in there
puts his hand in there
grabs onto those
seeds and now he can't
get his hand out because he won't let go of the
seeds he could let go of him
and run away but in fact he's not going to let go of it
so the hunter
just comes over and catches
him ties
the seeds
the seeds this reminds you of anything that you've
done a hundred or a thousand times
o kahneman and tversky
put together uh
and came up with the term cognitive biases they
started collecting weird
where we aren't thinking at all reasonably
they came up with i don't know a few hundred of
them there's a good wikipedia page
for these
um
so
yeah so this one was loss
aversion favorite one
here is dunning krueger
these smug pilots have lost touch with
regular passengers like us who thinks i should fly the plane
we just had a president kind of like that
but if
you want gunning krueger take a look at pretty much every computer
program in the planet it is the
maximum of being more confident
than it should be about what it's able to do and what
another one confirmation bias we should all
know that one being researchers and academics
we look for things
that support what what we think is going on we're totally
story based even when we think we're trying
to be scientific
so
basically what's going on
here is an enormous part
of us is atavistic
and in fact what's going on right now
uh through zoom is
campfire where somebody's
up there telling stories here's mike checking me out to make
sure i'm going to end this on on time
[Music] and
we're wired for stores we use stories for so
many different things even when it's not a good
idea and uther
horse in here this guy from a hundred thousand years
ago everything at this
modern campfire would be completely familiar to him
and this
is because he has pretty much the same brain that we have
he wants the same things
and if we give him a haircut and
a briefcase and modern version
of a sphere of a spear which is a tomahawk
basically
the same thing obtains
and this is something we have to remember that
and it's something that is used all the time
in the courts and in
every kind of advocacy and [Music]
prosecution and defense
is turn it into a good story and you
don't have to worry about the truth
because we don't care whether the stories are true we
and to be the way we think things
o bertrand
russell wrote a book about power
and one of the major ideas
in it is hey no matter what system you
devise you need to arrange to allow the most
able energetic and aggressive to prosper
or they will game it to death and eventually undo
it there's a lot of contention right now
with people simply
trying to get their own way by gaming
things and
the law has to be above that but it isn't
his
pamphlet was key
to the
uh direction that our country took
and it's a joke because
it in fact is not about
what common sense was at all common sense was
you should have a monarchy but
what he tries to do is come up with a new common sense and one of
the great principles in there he says instead of having
the king be the law we can have the law be the king
meaning we should be able to
invent a better society than we
just blundered our way into
and so
this pamphlet which uh
900 000 went out to 1.5 million
colonists in the first six months right
before the declaration of independence was one
of the most influential pieces of writing in
american history then this great
book by uh anthony
amsterdam who is a supreme court litigator and
a distinguished lawyer and legal
theorist many of you will know of him and jerry
brenner uh jerry lives along
jerry lived just slightly past
of 100. and so he had many
different careers and he spent 10 years in
the nyu law school as a professor
this is one of the books that they
came up with
and their aim is to make what is familiar strange again which
number one job i think for any university
uh the toughest people to
are people who know something
because they often know it only too well
and it talks
for a person who doesn't know a lot about the law this
the best introductions
to what it's really about
but what was interesting is that the
thrust of this book what they looked
at to make up the book was
what the supreme court
actually does and did and has done
how much what they've done
has not been particularly logical
at all but has been organized around
different kinds of story lines
and narrative
i think understanding
narrative being able to generate narrative
to be able to deal with narrative
on a human scale is critical
attempt to be serious
about computing in the law
so here's an example
of what is sometimes called the private universe
from nsf a few years ago
right after the
harvard commencement they went
around and asked some questions
heart
despite a lifetime of the very best education
students in our classrooms are failing to
learn science many of these students will graduate
from college with the same scientific misconceptions that
they had on entering grade school to test
how a lifetime of education affects our understanding of science
ask these recent graduates some simple questions in
astronomy consider for example how
cold it is at any given time of the year has to do
with the closeness of the earth to the
sun during the season the earth goes
around the sun and
now
after i saw this i was
had taught a class at ucla so i went
and asked some of the students if they've come outside
with me to talk about
both the phases of the moon question
and the seasons because
to ask for instance this
guy uh well
do you know when it's summer up
here in the northern hemisphere do you know what the
season is in the southern hemisphere and
all the students i talked to knew it was
winter and all of a sudden they realized oh the
distance the distance to the sun can't
be the main thing about the seasons and
to those who thought the earth got in between
the moon and the sun to cause the phases
of the moon i asked them if they'd ever seen the sun
moon in the sky at the same
course we have on many occasions
and the moon is usually in face
and so the earth cannot possibly be between there
so i think we can get two ideas
this that the major curriculum at harvard
at least is confidence 101
and from the time i spent at stanford i would say
that that's true there also
but the other thing
is that the what nsf was exploring as
a knowledge problem is not just a knowledge problem
a cognitive problem and we have to thank
the harvard students for not having the knowledge
because the cognitive problem is that they had
seen in their lives every single one of them
examples to what their
main theory was and none of their
training in their entire life had
set them up with the heuristic to try and compare
know about
uh the earth and the planets
so this is a much bigger failure than simply
not knowing something
so this gives rise to this private universe
idea that we have
very different worlds in our brains
we project them out on the world
and we have trouble communicating so a mane and hooves
well
not necessarily a horse
and in fact in med school this people who
come up with the zebra from that description are called zebras
people who come up with the most complicated
diagnosis rather than the one that's the most
but
for these two kids uh they can
eventually get to agree on what
hamburger is that comes into their shared space
and we can see that that words are
kind of uh an
invention that allows us to point to things
that are in our jointly shared imagination
so language came after the gesture
so if you go to an upgrade of this campfire
because the theater
grew out of what happened in campfires
we noticed this interesting thing about
storytelling in general and
audience's intelligence
is relied on to make sense
of what's happening
even in a movie
so the the only
intelligence in the room is the audience there and you have to do
something in this shared space to get the
understand what you want them to understand
and the theater
is sometimes called the magic mirror precisely
because uh
its job the job of this
kind of media is to beam the audience's intelligence
out at them and we can see that
this is one of bruner's mentalities
iconic in fact if you take acting
lessons the first thing they'll get you to do is not
just pantomime but to learn how to pose
because posing is the the
center of what the theatrical art
is about there are words
in fact we can get rid of the
everything except the words by inventing
writing and this puts a real burden
on the audience the audience
has to do a lot of learning but the learning they do helps
them think differently and changes the way civilization
turns out when a computer comes
on the scene it's inherently
a backstage kind of
thing nobody it doesn't have a story to tell
so you have to provide some kind
of a story this is what we did at xerox park which will be
familiar to you today
and it uses both the iconic and the
symbolic mentalities of bruner and
the fact that we can touch it
and we even made some interfaces
that could touch back
so the idea here with relevance
conference is pretty much
i think what comp law should be about right
primarily user interface
for something more complicated than what we were
xerox park and you make progress by inventing
entirely new environment not just doing
little experiments
this interface worked because
it its
semantics was grounded on itself
it was about itself it wasn't about the outside world
and the uh object-oriented stuff
at xerox park was able to create
an ontology that allowed the system uh
to be very self-explanatory and exploratory
in various ways but if we take something
like uh montesquieu's spirit
of laws and put it on a computer
as a book but for most of
this book years ago and it was a
real struggle most of us want
an ai or something that can really help
us learn so the user
interface is not just about
access to functions
it's also about learning and exploring
and getting more versed in what
is an entire world that you're giving the user to live in
and the grounding
problem is well put by einstein
when he compares mathematics which
is not grounded in anything except itself
and science which has to be grounded
reality and so you
can use math and science as long as
you realize your grounding is not going to be perfect and that
your main job is to find the groundings that
mathematical logic to come to conclusions
about the real world
well okay
so
in order to point to something
in our shared space we have to imagine the same things
it's very likely that's what these cave paintings were about
and it's not that we don't
uh imagine things
we can imagine angels 77 of us believe
there are angels 94 percent of us believe
there are supreme beings many
too many of us believe
in a human sacrifice still going
on about half of us
pretty much everywhere believe in witches and demons i mean
in demons and ghosts 21
still believe in witches
you can look up the great cat massacre
when they decided the cats were the witches in france
in the 17th century
we believe in war we can imagine war
and we can
imagine something going wrong when we're vaccinated
so we have all of these things we can imagine and talk about
and people do but most
of us have had a hard time imagining
tiny little viruses and
the devastating deaths are
precisely the result of not
understanding something that a small percentage
of scientists and anybody who understood their
grade biology course would
understand so his current schooling has failed
disastrously and because there are second
and third waves societal learning from the first waves
and thinking have also failed
and of course even smaller is the carbon
dioxide model uh
molecule and we're flunking this
badly this is the
okay uh
i had a rant here but i decided i would
just ask you in this conference to consider the
the following two questions how weak is the current
state of the art of computer generated explanation
and computer-assisted learning especially
compared to what it was 40 years ago
it got up to a certain state 40 years ago it's gotten
slightly better since then but i
couldn't find anything like what i expected
and what seems to be going on is
somehow a lot of researchers
are like this guy who lost his keys back
here and
well he found a skate
key has machine learning
and semantic web on it and if you keep on working
might get a better skate key the problem
is you're not going to get the access to the vehicle
you wanted in the first place so
uh yeah this is
this is the symbol i put on my slides to quit
talking know more about that
okay so now i'll finish up
some some bright
prospects first uh thanks
to tim miller for bringing this
book by bertram to my attention
i think it's terrific
he considers himself a cultural psychologist
he was inspired by brenner
and it's a book about the way people
try to explain things to themselves
to others and gives a kind
of what you might call a common sense
uh framework for
humans actually go about
doing things and there are other books
associated with so it's all
about what can we agree on
how do you find meanings with the
private
universes and
many many different topics
that i've not seen treated quite this way
um
our field i think only a few
actually been treated
in a.i so there are many many
other ways to look at it and
an example of a an enormous
effort over decades with
tens of millions of relationships
and ways of inferring things
course psych by doug lennon
and this is worth looking
at again
ot because it's the answer to today's problem so much
but because of what's in it
is hard to escape having to have
answers for and it's the only
large-scale artifact of its kind that i know
of that has uh that could actually be talked
about and many things have happened
about the theory of causality over the
last 40 years including this
the work of judea pearl
and then finally there are
new uh user interface
people on the scene the user interface is probably at its lowest
level
uh in computing uh that it's
been for a long time except there are people
like brett victor
here who is
just an amazing talent
so if you look at this these
four things here we've got
a framework for thinking
we've got ways of dealing
with uh the
areas of logic that we can deal with we've got examples
of huge artifacts
that are going to be required in order to do cognitive ai
t's not that there isn't going to be machine learning as
part of it but having machine learning
central part is the tail wagging
the dog and then finally there
are some great user interface thinkers on
the scene so
i think you need to make a platform
think these are four starting places
to do such a platform
i think these are the four places i would start i'd look for
some sort of vector sums of these
dimensions uh get everybody
involved in a new grand challenge and synthesis
and make the center of it dealing with critical
are considered to be important
thank you very much
i'm sorry for running over a little bit but that's thanks
alan thank you so much for this uh
wonderful very human-centric framework
you've uh given us as a wonderful
basis for for the rest of our our conference
so we very much appreciate you and appreciate you
taking the time uh for sharing your ideas
about the world of law and technology
and uh and yeah
please accept you know my applause as an applause on behalf
the over 600 people who i'm
sure share the enthusiasm about your about your