[Music] Oscar sharp I’m Ross Goodwin it starts
Oscar sharp I’m Ross Goodwin it starts
Oscar sharp I’m Ross Goodwin it starts quite a while back we haven’t even met
quite a while back we haven’t even met
quite a while back we haven’t even met yet and I’m in London and I’m trying to
yet and I’m in London and I’m trying to
yet and I’m in London and I’m trying to be a director of films and theater and
be a director of films and theater and
be a director of films and theater and I’m trying to understand how you can
I’m trying to understand how you can
I’m trying to understand how you can make something that is classical and
make something that is classical and
make something that is classical and original and personal how can it be sort
original and personal how can it be sort
original and personal how can it be sort of embodying of the form how can it
of embodying of the form how can it
of embodying of the form how can it innovate on the form and how can it
innovate on the form and how can it
innovate on the form and how can it still say something that’s quite a lot
still say something that’s quite a lot
still say something that’s quite a lot but a tall order and it’s a it’s the the
but a tall order and it’s a it’s the the
but a tall order and it’s a it’s the the classic dilemma for a creative person
classic dilemma for a creative person
classic dilemma for a creative person I’m sitting for this reason in a room
I’m sitting for this reason in a room
I’m sitting for this reason in a room and watching an actor do this so it’s
and watching an actor do this so it’s
and watching an actor do this so it’s quite an odd thing to watch happening
quite an odd thing to watch happening
quite an odd thing to watch happening obviously um the reason I’m watching it
obviously um the reason I’m watching it
obviously um the reason I’m watching it is I’m in this workshop and I don’t know
is I’m in this workshop and I don’t know
is I’m in this workshop and I don’t know why I’m watching and the person running
why I’m watching and the person running
why I’m watching and the person running the workshop says so what did you just
the workshop says so what did you just
the workshop says so what did you just see happen what was that like oh well
see happen what was that like oh well
see happen what was that like oh well obviously this is a person who’s a bit
obviously this is a person who’s a bit
obviously this is a person who’s a bit worried about the shoe on the floor
worried about the shoe on the floor
worried about the shoe on the floor maybe a bit disgusted by it maybe they
maybe a bit disgusted by it maybe they
maybe a bit disgusted by it maybe they don’t really like brokes or Oxford’s or
don’t really like brokes or Oxford’s or
don’t really like brokes or Oxford’s or whatever these anyway it turned out that
whatever these anyway it turned out that
whatever these anyway it turned out that the things that actor were doing was a
the things that actor were doing was a
the things that actor were doing was a random list of actions they’ve been
random list of actions they’ve been
random list of actions they’ve been assigned those actions randomly and
assigned those actions randomly and
assigned those actions randomly and they’ve been pulled out of I thought god
they’ve been pulled out of I thought god
they’ve been pulled out of I thought god that’s interesting is a kind of a story
that’s interesting is a kind of a story
that’s interesting is a kind of a story seemed to be there everyone in the room
seemed to be there everyone in the room
seemed to be there everyone in the room agreed that they’ve seen some of those
agreed that they’ve seen some of those
agreed that they’ve seen some of those story I’m like did that come from the
story I’m like did that come from the
story I’m like did that come from the actor did that come from me I’m not sure
actor did that come from me I’m not sure
actor did that come from me I’m not sure but how far can we push this can we go
but how far can we push this can we go
but how far can we push this can we go further we can we can we make these
further we can we can we make these
further we can we can we make these lists of actions longer so I decided to
lists of actions longer so I decided to
lists of actions longer so I decided to build a machine that made longer lists
build a machine that made longer lists
build a machine that made longer lists of actions and the machines looked like
of actions and the machines looked like
of actions and the machines looked like this it was a pair of dice because I
this it was a pair of dice because I
this it was a pair of dice because I actually can’t build machines so I
actually can’t build machines so I
actually can’t build machines so I rolled the dice over and over again made
rolled the dice over and over again made
rolled the dice over and over again made longer and longer lists and and it still
longer and longer lists and and it still
longer and longer lists and and it still kind of worked and I studied as they got
kind of worked and I studied as they got
kind of worked and I studied as they got longer they can either broke down a bit
longer they can either broke down a bit
longer they can either broke down a bit so I started applying rules of
so I started applying rules of
so I started applying rules of story-like
story-like
story-like complication resolution and crisis and
complication resolution and crisis and
complication resolution and crisis and conflict in it it still kind of held up
conflict in it it still kind of held up
conflict in it it still kind of held up until I tried to apply it to dialogue
until I tried to apply it to dialogue
until I tried to apply it to dialogue here is a piece of dialogue made with
here is a piece of dialogue made with
here is a piece of dialogue made with dice
dice
dice copper explained ill-fated truck neat
copper explained ill-fated truck neat
copper explained ill-fated truck neat unites bronze educated tenuous hum
unites bronze educated tenuous hum
unites bronze educated tenuous hum decisive notice it wasn’t very good I
decisive notice it wasn’t very good I
decisive notice it wasn’t very good I had a problem I couldn’t make the speech
had a problem I couldn’t make the speech
had a problem I couldn’t make the speech work I really needed someone who knew
work I really needed someone who knew
work I really needed someone who knew about speech meanwhile I was working as
about speech meanwhile I was working as
about speech meanwhile I was working as a political speech writer I worked on
a political speech writer I worked on
a political speech writer I worked on the Obama campaign 2008 then at the
the Obama campaign 2008 then at the
the Obama campaign 2008 then at the White House for a year then a Treasury
White House for a year then a Treasury
White House for a year then a Treasury for two years and then I went out on my
for two years and then I went out on my
for two years and then I went out on my own to be a freelance political ghost
own to be a freelance political ghost
own to be a freelance political ghost you know during that time I was also
you know during that time I was also
you know during that time I was also learning to code I was teaching myself
learning to code I was teaching myself
learning to code I was teaching myself and I immediately gravitated toward this
and I immediately gravitated toward this
and I immediately gravitated toward this broad and expanding world of natural
broad and expanding world of natural
broad and expanding world of natural language processing and natural language
language processing and natural language
language processing and natural language generation which are they the terms for
generation which are they the terms for
generation which are they the terms for those fields that deal with natural
those fields that deal with natural
those fields that deal with natural language as opposed to computer language
language as opposed to computer language
language as opposed to computer language and I got really obsessed with trying to
and I got really obsessed with trying to
and I got really obsessed with trying to find ways to make my writing process
find ways to make my writing process
find ways to make my writing process he’s more efficient using computation so
he’s more efficient using computation so
he’s more efficient using computation so I developed techniques like this where
I developed techniques like this where
I developed techniques like this where I’d write in a spreadsheet rather than
I’d write in a spreadsheet rather than
I’d write in a spreadsheet rather than in a Word document and if I had to write
in a Word document and if I had to write
in a Word document and if I had to write a batch of letters let’s say I’d write
a batch of letters let’s say I’d write
a batch of letters let’s say I’d write all the first paragraphs together down
all the first paragraphs together down
all the first paragraphs together down the first row then all the second
the first row then all the second
the first row then all the second paragraphs together then all the third
paragraphs together then all the third
paragraphs together then all the third paragraphs together with one letter down
paragraphs together with one letter down
paragraphs together with one letter down each column and a paragraph in each cell
each column and a paragraph in each cell
each column and a paragraph in each cell and then every other day I’d randomly
and then every other day I’d randomly
and then every other day I’d randomly rearrange the paragraphs edit them a
rearrange the paragraphs edit them a
rearrange the paragraphs edit them a little bit and handed it as a new batch
little bit and handed it as a new batch
little bit and handed it as a new batch of letters so doing that I thought I
of letters so doing that I thought I
of letters so doing that I thought I could take an eight hour a day job and
could take an eight hour a day job and
could take an eight hour a day job and turn it into a two hour a day job but I
turn it into a two hour a day job but I
turn it into a two hour a day job but I I was still looking for something more I
I was still looking for something more I
I was still looking for something more I wanted to explore this type of thing in
wanted to explore this type of thing in
wanted to explore this type of thing in a more creative context so I decided to
a more creative context so I decided to
a more creative context so I decided to attend the interactive
attend the interactive
attend the interactive telecommunications program at New York
telecommunications program at New York
telecommunications program at New York University which is sort of like art
University which is sort of like art
University which is sort of like art school for engineers or engineering
school for engineers or engineering
school for engineers or engineering school for artists and I was seeking a
school for artists and I was seeking a
school for artists and I was seeking a more creative context for this work and
more creative context for this work and
more creative context for this work and which floor was it on Ross there’s on
which floor was it on Ross there’s on
which floor was it on Ross there’s on the fourth floor right fourth floor and
the fourth floor right fourth floor and
the fourth floor right fourth floor and meanwhile I was on the tenth floor you
meanwhile I was on the tenth floor you
meanwhile I was on the tenth floor you see I in my pursuit of making something
see I in my pursuit of making something
see I in my pursuit of making something classical and original I’d I tried doing
classical and original I’d I tried doing
classical and original I’d I tried doing a story as a short film that was a it
a story as a short film that was a it
a story as a short film that was a it was it used it was about cancer but it
was it used it was about cancer but it
was it used it was about cancer but it was levitation was levitation and cancer
was levitation was levitation and cancer
was levitation was levitation and cancer and it was personal as well because my
and it was personal as well because my
and it was personal as well because my mom had cancer so I done something in
mom had cancer so I done something in
mom had cancer so I done something in this realm but not using automation and
this realm but not using automation and
this realm but not using automation and that done well it got nominated for a
that done well it got nominated for a
that done well it got nominated for a BAFTA and it got me a Fulbright
BAFTA and it got me a Fulbright
BAFTA and it got me a Fulbright scholarship and brought me to NYU to the
scholarship and brought me to NYU to the
scholarship and brought me to NYU to the tenth floor but I would sometimes get
tenth floor but I would sometimes get
tenth floor but I would sometimes get off early on the fourth floor into this
off early on the fourth floor into this
off early on the fourth floor into this Willy Wonka Wonderland of tech geniuses
Willy Wonka Wonderland of tech geniuses
Willy Wonka Wonderland of tech geniuses where you find people like Ross building
where you find people like Ross building
where you find people like Ross building well he didn’t do this but robot
well he didn’t do this but robot
well he didn’t do this but robot jellyfish I remember I remember a wall
jellyfish I remember I remember a wall
jellyfish I remember I remember a wall of robot dildos that would follow you
of robot dildos that would follow you
of robot dildos that would follow you around the room I’m not joking that was
around the room I’m not joking that was
around the room I’m not joking that was real it’s very strange place but very
real it’s very strange place but very
real it’s very strange place but very exciting place and I thought wow maybe I
exciting place and I thought wow maybe I
exciting place and I thought wow maybe I can find someone in here who can help me
can find someone in here who can help me
can find someone in here who can help me with these dreams that I used to have
with these dreams that I used to have
with these dreams that I used to have and eventually mice to finagle my way
and eventually mice to finagle my way
and eventually mice to finagle my way into a class right the class was called
into a class right the class was called
into a class right the class was called surveillance documentary and we were
surveillance documentary and we were
surveillance documentary and we were learning how to program
learning how to program
learning how to program pan tilt zoom surveillance cameras for
pan tilt zoom surveillance cameras for
pan tilt zoom surveillance cameras for creative applications so of course they
creative applications so of course they
creative applications so of course they combined the filmmakers with the
combined the filmmakers with the
combined the filmmakers with the technologies right cameras that makes
technologies right cameras that makes
technologies right cameras that makes sense but as I walk in the thing that’s
sense but as I walk in the thing that’s
sense but as I walk in the thing that’s exciting to me isn’t the camera it’s
exciting to me isn’t the camera it’s
exciting to me isn’t the camera it’s what I can see on Ross’s laptop screen
what I can see on Ross’s laptop screen
what I can see on Ross’s laptop screen because it’s a poem that’s writing
because it’s a poem that’s writing
because it’s a poem that’s writing itself that’s right it was a Markov
itself that’s right it was a Markov
itself that’s right it was a Markov chain poem I’m at the time I’ve been
chain poem I’m at the time I’ve been
chain poem I’m at the time I’ve been experimenting with a simple algorithm
experimenting with a simple algorithm
experimenting with a simple algorithm called a Markov chain
called a Markov chain
called a Markov chain a very old developed in 1906 it’s
a very old developed in 1906 it’s
a very old developed in 1906 it’s basically it what’s called a state
basically it what’s called a state
basically it what’s called a state machine so given the current state it
machine so given the current state it
machine so given the current state it tries to predict the next state using
tries to predict the next state using
tries to predict the next state using only the information contained in the
only the information contained in the
only the information contained in the current state so how that translates to
current state so how that translates to
current state so how that translates to words is given a sequence of words let’s
words is given a sequence of words let’s
words is given a sequence of words let’s say three words it tries to predict
say three words it tries to predict
say three words it tries to predict which word would come next given the
which word would come next given the
which word would come next given the occurrence of those three words and all
occurrence of those three words and all
occurrence of those three words and all the words that come after those three
the words that come after those three
the words that come after those three words in a corpus or a body of text so
words in a corpus or a body of text so
words in a corpus or a body of text so using that you can predict the next word
using that you can predict the next word
using that you can predict the next word over and over again to generate text so
over and over again to generate text so
over and over again to generate text so obviously I got kind of excited and he
obviously I got kind of excited and he
obviously I got kind of excited and he tried to explain to me and exactly that
tried to explain to me and exactly that
tried to explain to me and exactly that way and maybe I kind of lost track about
way and maybe I kind of lost track about
way and maybe I kind of lost track about halfway through but I was shaking him by
halfway through but I was shaking him by
halfway through but I was shaking him by the lapels by this point going Ross can
the lapels by this point going Ross can
the lapels by this point going Ross can we use this method to write dialogue and
we use this method to write dialogue and
we use this method to write dialogue and I said yes why not so I tried Markov
I said yes why not so I tried Markov
I said yes why not so I tried Markov chains on some movie dialogue and I got
chains on some movie dialogue and I got
chains on some movie dialogue and I got results that were somewhat like this
results that were somewhat like this
results that were somewhat like this cloverleaf will own toontown quite
cloverleaf will own toontown quite
cloverleaf will own toontown quite legally so your cloverleaf no 1
legally so your cloverleaf no 1
legally so your cloverleaf no 1 hippopotamus is my favorite genius to
hippopotamus is my favorite genius to
hippopotamus is my favorite genius to figure out what’s going on here Tina
figure out what’s going on here Tina
figure out what’s going on here Tina Tina you I’ll kill you like
Tina you I’ll kill you like
Tina you I’ll kill you like a moth to a flame so it wasn’t quite
a moth to a flame so it wasn’t quite
a moth to a flame so it wasn’t quite ready for primetime I think you’ll agree
ready for primetime I think you’ll agree
ready for primetime I think you’ll agree round of applause to Ross though amazing
round of applause to Ross though amazing
round of applause to Ross though amazing performance right so we can’t quite make
performance right so we can’t quite make
performance right so we can’t quite make a movie with that maybe we can go back
a movie with that maybe we can go back
a movie with that maybe we can go back to the action stuff and I was talking to
to the action stuff and I was talking to
to the action stuff and I was talking to Ross about the rules of screenplay I
Ross about the rules of screenplay I
Ross about the rules of screenplay I said in in action in a screenplay it’s
said in in action in a screenplay it’s
said in in action in a screenplay it’s not like a novel you can’t describe what
not like a novel you can’t describe what
not like a novel you can’t describe what people are thinking you can only say
people are thinking you can only say
people are thinking you can only say what you see can we use these cameras in
what you see can we use these cameras in
what you see can we use these cameras in some way with saying what you see right
some way with saying what you see right
some way with saying what you see right so I sort of thought about that I
so I sort of thought about that I
so I sort of thought about that I thought about that really long and hard
thought about that really long and hard
thought about that really long and hard and I thought you know how to create a
and I thought you know how to create a
and I thought you know how to create a realistically descriptive action
realistically descriptive action
realistically descriptive action sequence because it’s not trivial with
sequence because it’s not trivial with
sequence because it’s not trivial with the computer computer-generated text
the computer computer-generated text
the computer computer-generated text especially with Markoff tends to venture
especially with Markoff tends to venture
especially with Markoff tends to venture into the realms of the Unreal rather
into the realms of the Unreal rather
into the realms of the Unreal rather quickly it doesn’t want to stay grounded
quickly it doesn’t want to stay grounded
quickly it doesn’t want to stay grounded to a realistically descriptive set of
to a realistically descriptive set of
to a realistically descriptive set of words that might apply to a particular
words that might apply to a particular
words that might apply to a particular image or a particular scene so I thought
image or a particular scene so I thought
image or a particular scene so I thought about that you know and I and I realized
about that you know and I and I realized
about that you know and I and I realized that there are these algorithms that are
that there are these algorithms that are
that there are these algorithms that are coming
coming
coming to use called convolutional neural
to use called convolutional neural
to use called convolutional neural networks that can take an image and
networks that can take an image and
networks that can take an image and extract a set of words the set of nouns
extract a set of words the set of nouns
extract a set of words the set of nouns representing concepts or items found in
representing concepts or items found in
representing concepts or items found in that image by the machine so the machine
that image by the machine so the machine
that image by the machine so the machine that can quite literally describe what
that can quite literally describe what
that can quite literally describe what it’s seeing and I thought you know maybe
it’s seeing and I thought you know maybe
it’s seeing and I thought you know maybe a set of nouns from an image would be a
a set of nouns from an image would be a
a set of nouns from an image would be a good place to start and so I built this
good place to start and so I built this
good place to start and so I built this camera that wrote poetry from images it
camera that wrote poetry from images it
camera that wrote poetry from images it was called word camera and I made it web
was called word camera and I made it web
was called word camera and I made it web app in a series of physical devices like
app in a series of physical devices like
app in a series of physical devices like I just wonder them I can just take
I just wonder them I can just take
I just wonder them I can just take Oscars picture and it’ll it’ll print
Oscars picture and it’ll it’ll print
Oscars picture and it’ll it’ll print something based on what it sees so in
something based on what it sees so in
something based on what it sees so in this case similarly the man has a long
this case similarly the man has a long
this case similarly the man has a long hair otherwise it has a Bigfoot than
hair otherwise it has a Bigfoot than
hair otherwise it has a Bigfoot than woman immediately it made journey to
woman immediately it made journey to
woman immediately it made journey to Europe earlier it may fear commitment so
Europe earlier it may fear commitment so
Europe earlier it may fear commitment so that’s a good one of those so so I had
that’s a good one of those so so I had
that’s a good one of those so so I had asked Ross for a camera which I thought
asked Ross for a camera which I thought
asked Ross for a camera which I thought was maybe gonna write scripts based on
was maybe gonna write scripts based on
was maybe gonna write scripts based on what we’d shot and instead he made a
what we’d shot and instead he made a
what we’d shot and instead he made a camera that insulted me accurately in
camera that insulted me accurately in
camera that insulted me accurately in many cases and and and so then we get
many cases and and and so then we get
many cases and and and so then we get four people following along in story
four people following along in story
four people following along in story structure land a twist a complication I
structure land a twist a complication I
structure land a twist a complication I got a phone call from this man did you
got a phone call from this man did you
got a phone call from this man did you choose that that’s not okay okay um so I
choose that that’s not okay okay um so I
choose that that’s not okay okay um so I get phone calls from Tokyo Nagoya the
get phone calls from Tokyo Nagoya the
get phone calls from Tokyo Nagoya the form of Spider Man he had seen my film
form of Spider Man he had seen my film
form of Spider Man he had seen my film at the levitating woman he really liked
at the levitating woman he really liked
at the levitating woman he really liked it he’s a producer as well and he said
it he’s a producer as well and he said
it he’s a producer as well and he said hey you want to come to Hollywood and
hey you want to come to Hollywood and
hey you want to come to Hollywood and write your first screenplay I said oh
write your first screenplay I said oh
write your first screenplay I said oh that sounds fun we haven’t got the
that sounds fun we haven’t got the
that sounds fun we haven’t got the Machine working so I might as well do
Machine working so I might as well do
Machine working so I might as well do something so if I go to Hollywood to try
something so if I go to Hollywood to try
something so if I go to Hollywood to try to do that to try to do something
to do that to try to do something
to do that to try to do something original and and classical and personal
original and and classical and personal
original and and classical and personal meanwhile I was still playing with my
meanwhile I was still playing with my
meanwhile I was still playing with my word camera and I was looking for ways
word camera and I was looking for ways
word camera and I was looking for ways to improve the output and to have more
to improve the output and to have more
to improve the output and to have more control over the output because prior to
control over the output because prior to
control over the output because prior to this I hadn’t been training my own
this I hadn’t been training my own
this I hadn’t been training my own neural nets I’d been using those other
neural nets I’d been using those other
neural nets I’d been using those other people had trained through api’s and web
people had trained through api’s and web
people had trained through api’s and web services and I wanted to start training
services and I wanted to start training
services and I wanted to start training my own just to get more of a handle on
my own just to get more of a handle on
my own just to get more of a handle on the technology and an understanding of
the technology and an understanding of
the technology and an understanding of what was really happening as well as
what was really happening as well as
what was really happening as well as just that level of control over the
just that level of control over the
just that level of control over the output so I discovered some code on
output so I discovered some code on
output so I discovered some code on github by a gentleman named Andre
github by a gentleman named Andre
github by a gentleman named Andre carpathia at Stanford called char RNN
carpathia at Stanford called char RNN
carpathia at Stanford called char RNN and knurl talked to you respectively
and knurl talked to you respectively
and knurl talked to you respectively which were a long short term memory
which were a long short term memory
which were a long short term memory recurrent neural net
recurrent neural net
recurrent neural net which is a neural net that’s a lot like
which is a neural net that’s a lot like
which is a neural net that’s a lot like a Markov chain and that it’s predicting
a Markov chain and that it’s predicting
a Markov chain and that it’s predicting in this case the next letter over and
in this case the next letter over and
in this case the next letter over and over again to generate text and then
over again to generate text and then
over again to generate text and then neural talk to is an image captioning
neural talk to is an image captioning
neural talk to is an image captioning system using a convolutional neural net
system using a convolutional neural net
system using a convolutional neural net so I put those two things together to
so I put those two things together to
so I put those two things together to make a new version of word camera and
make a new version of word camera and
make a new version of word camera and the results were really astonishing I
the results were really astonishing I
the results were really astonishing I was really amazed with with the tape of
was really amazed with with the tape of
was really amazed with with the tape of poetry it was producing compared to what
poetry it was producing compared to what
poetry it was producing compared to what extra grand a moment ago and so I wanted
extra grand a moment ago and so I wanted
extra grand a moment ago and so I wanted some voiceover of somebody reading one
some voiceover of somebody reading one
some voiceover of somebody reading one of these poems for a documentation video
of these poems for a documentation video
of these poems for a documentation video I was making and Oscar has a great voice
I was making and Oscar has a great voice
I was making and Oscar has a great voice so I sent him this poem generated by the
so I sent him this poem generated by the
so I sent him this poem generated by the word camera from a picture of a clock
word camera from a picture of a clock
word camera from a picture of a clock well yes so by the way it’s Sisson
well yes so by the way it’s Sisson
well yes so by the way it’s Sisson English accent I’m cheating but he sends
English accent I’m cheating but he sends
English accent I’m cheating but he sends me this thing it’s 2 p.m. I’m lying in
me this thing it’s 2 p.m. I’m lying in
me this thing it’s 2 p.m. I’m lying in bed in Hollywood and I’m depressed I
bed in Hollywood and I’m depressed I
bed in Hollywood and I’m depressed I have writer’s block I’ve been trying to
have writer’s block I’ve been trying to
have writer’s block I’ve been trying to write the screenplay that is supposedly
write the screenplay that is supposedly
write the screenplay that is supposedly going to be original in classical
going to be original in classical
going to be original in classical impersonal and I am stuck and Toby
impersonal and I am stuck and Toby
impersonal and I am stuck and Toby occasionally phones me up and goes what
occasionally phones me up and goes what
occasionally phones me up and goes what are you doing and I go promise but I get
are you doing and I go promise but I get
are you doing and I go promise but I get this email and I open it up and turn on
this email and I open it up and turn on
this email and I open it up and turn on my microphone because Ross is a friend
my microphone because Ross is a friend
my microphone because Ross is a friend and record the following a close-up of a
and record the following a close-up of a
and record the following a close-up of a clock on a wall of four o’clock in the
clock on a wall of four o’clock in the
clock on a wall of four o’clock in the morning I am NOT so strange and will not
morning I am NOT so strange and will not
morning I am NOT so strange and will not delay the room is blown away from the
delay the room is blown away from the
delay the room is blown away from the door and the stones are beginning to
door and the stones are beginning to
door and the stones are beginning to shine the silence is hardly final
shine the silence is hardly final
shine the silence is hardly final somewhere in the street I can see the
somewhere in the street I can see the
somewhere in the street I can see the trees begin ladies and gentlemen I
trees begin ladies and gentlemen I
trees begin ladies and gentlemen I experienced emotions it was a very
experienced emotions it was a very
experienced emotions it was a very strange moment for me the thing is all
strange moment for me the thing is all
strange moment for me the thing is all of this text that we generated before it
of this text that we generated before it
of this text that we generated before it was so weird in a way that when you read
was so weird in a way that when you read
was so weird in a way that when you read it your main emotion was that’s weird
it your main emotion was that’s weird
it your main emotion was that’s weird but the thing about an actor is when you
but the thing about an actor is when you
but the thing about an actor is when you give them some stuff that just has
give them some stuff that just has
give them some stuff that just has enough connection in it that they can
enough connection in it that they can
enough connection in it that they can feel something you know that you can
feel something you know that you can
feel something you know that you can start making drama so I immediately
start making drama so I immediately
start making drama so I immediately found out Ross and I go grabbing those
found out Ross and I go grabbing those
found out Ross and I go grabbing those lapels over the airwaves
lapels over the airwaves
lapels over the airwaves I don’t know how you did this Ross but
I don’t know how you did this Ross but
I don’t know how you did this Ross but can we use it to make dialogue and I
can we use it to make dialogue and I
can we use it to make dialogue and I said yes so I immediately went and
said yes so I immediately went and
said yes so I immediately went and downloaded the Cornell movie scripts
downloaded the Cornell movie scripts
downloaded the Cornell movie scripts corpus which is a collection of about
corpus which is a collection of about
corpus which is a collection of about 5000 screenplays that I ripped out all
5000 screenplays that I ripped out all
5000 screenplays that I ripped out all the dialogue is at that point he just
the dialogue is at that point he just
the dialogue is at that point he just asked for dialogue so I wanted to try
asked for dialogue so I wanted to try
asked for dialogue so I wanted to try dialogue and trained on that and this is
dialogue and trained on that and this is
dialogue and trained on that and this is the an example the first thing he sent
the an example the first thing he sent
the an example the first thing he sent me I was so stupid I had to go to the
me I was so stupid I had to go to the
me I was so stupid I had to go to the party I was sorry and I was wondering if
party I was sorry and I was wondering if
party I was sorry and I was wondering if you were gonna make me cry I wondered I
you were gonna make me cry I wondered I
you were gonna make me cry I wondered I wanted to be a good man I was thinking
wanted to be a good man I was thinking
wanted to be a good man I was thinking of coming back to the
of coming back to the
of coming back to the and I was missing something I couldn’t
and I was missing something I couldn’t
and I was missing something I couldn’t get the baby so this was better than the
get the baby so this was better than the
get the baby so this was better than the stuff from rolling the dice it’s still
stuff from rolling the dice it’s still
stuff from rolling the dice it’s still quite weird but it was better and I
quite weird but it was better and I
quite weird but it was better and I thought you know we can do this and at
thought you know we can do this and at
thought you know we can do this and at the same moment I found out we were we
the same moment I found out we were we
the same moment I found out we were we were one month from the sci-fi 48 hour
were one month from the sci-fi 48 hour
were one month from the sci-fi 48 hour film challenge run by sci-fi London
film challenge run by sci-fi London
film challenge run by sci-fi London which I do every single year is a really
which I do every single year is a really
which I do every single year is a really good way of getting back into filmmaking
good way of getting back into filmmaking
good way of getting back into filmmaking keeping yourself going and the way it
keeping yourself going and the way it
keeping yourself going and the way it works is you get 48 hours they give you
works is you get 48 hours they give you
works is you get 48 hours they give you a title a proper line of dialogue and 48
a title a proper line of dialogue and 48
a title a proper line of dialogue and 48 hours later you ever written shot cut
hours later you ever written shot cut
hours later you ever written shot cut and uploaded a movie it’s really fun and
and uploaded a movie it’s really fun and
and uploaded a movie it’s really fun and I thought well if we’re gonna use this
I thought well if we’re gonna use this
I thought well if we’re gonna use this dialogue maybe it would be so much
dialogue maybe it would be so much
dialogue maybe it would be so much better if it was the whole script but I
better if it was the whole script but I
better if it was the whole script but I remember when we’re doing Markoff you
remember when we’re doing Markoff you
remember when we’re doing Markoff you Markov chains couldn’t deal with
Markov chains couldn’t deal with
Markov chains couldn’t deal with formatting so Ross I guess we can’t we
formatting so Ross I guess we can’t we
formatting so Ross I guess we can’t we can’t do the whole script can we well
can’t do the whole script can we well
can’t do the whole script can we well fortunately time L STM’s can handle
fortunately time L STM’s can handle
fortunately time L STM’s can handle formatting because this one was written
formatting because this one was written
formatting because this one was written at the character level rather than the
at the character level rather than the
at the character level rather than the word level and it was able given a
word level and it was able given a
word level and it was able given a complete screenplay to replicate that
complete screenplay to replicate that
complete screenplay to replicate that formatting fairly well and differentiate
formatting fairly well and differentiate
formatting fairly well and differentiate between action description and dialogue
between action description and dialogue
between action description and dialogue and it was really fascinating you know I
and it was really fascinating you know I
and it was really fascinating you know I didn’t say this before watching these
didn’t say this before watching these
didn’t say this before watching these machines training to write letter by
machines training to write letter by
machines training to write letter by letter at first they’re just spitting
letter at first they’re just spitting
letter at first they’re just spitting out random gibberish random sequences of
out random gibberish random sequences of
out random gibberish random sequences of characters and then eventually they
characters and then eventually they
characters and then eventually they learn what words are and then they learn
learn what words are and then they learn
learn what words are and then they learn how to make sentences out of without of
how to make sentences out of without of
how to make sentences out of without of those words and then paragraphs and it’s
those words and then paragraphs and it’s
those words and then paragraphs and it’s kind of incredible to watch yeah I’ve
kind of incredible to watch yeah I’ve
kind of incredible to watch yeah I’ve taught kids to write screenplays even
taught kids to write screenplays even
taught kids to write screenplays even medium age kids and honestly our little
medium age kids and honestly our little
medium age kids and honestly our little screenwriter was doing it better it
screenwriter was doing it better it
screenwriter was doing it better it learns quite quickly that action is only
learns quite quickly that action is only
learns quite quickly that action is only what you can see that something it takes
what you can see that something it takes
what you can see that something it takes in open-toe me five or six lessons to
in open-toe me five or six lessons to
in open-toe me five or six lessons to get into their heads
get into their heads
get into their heads it made a dialogue that was sort of the
it made a dialogue that was sort of the
it made a dialogue that was sort of the right length the sentences were right
right length the sentences were right
right length the sentences were right they were declaratives and accusative
they were declaratives and accusative
they were declaratives and accusative zan an emotional statements and it was
zan an emotional statements and it was
zan an emotional statements and it was it was kind of learning to write
it was kind of learning to write
it was kind of learning to write screenplays as much as they were weird
screenplays as much as they were weird
screenplays as much as they were weird and there was still one problem which is
and there was still one problem which is
and there was still one problem which is we have 48 hours and it had taken three
we have 48 hours and it had taken three
we have 48 hours and it had taken three weeks to train this thing I was like we
weeks to train this thing I was like we
weeks to train this thing I was like we can’t we don’t have three weeks to write
can’t we don’t have three weeks to write
can’t we don’t have three weeks to write the screenplay well I had been training
the screenplay well I had been training
the screenplay well I had been training on graphics processing units or GPUs and
on graphics processing units or GPUs and
on graphics processing units or GPUs and why use high-performance computing
why use high-performance computing
why use high-performance computing facility and I didn’t have a GPU to
facility and I didn’t have a GPU to
facility and I didn’t have a GPU to generate text with generating text takes
generate text with generating text takes
generate text with generating text takes a lot less time than training but it
a lot less time than training but it
a lot less time than training but it still takes quite a while if you’re on a
still takes quite a while if you’re on a
still takes quite a while if you’re on a CPU and you have a big model you know I
CPU and you have a big model you know I
CPU and you have a big model you know I was looking for a solution and I
was looking for a solution and I
was looking for a solution and I discovered this
discovered this
discovered this Nvidia Jetson system-on-chip which is
Nvidia Jetson system-on-chip which is
Nvidia Jetson system-on-chip which is sort of like a Raspberry Pi if you know
sort of like a Raspberry Pi if you know
sort of like a Raspberry Pi if you know what that is it’s like a little computer
what that is it’s like a little computer
what that is it’s like a little computer with the GPU so using that we were able
with the GPU so using that we were able
with the GPU so using that we were able to generate a screenplay in just a few
to generate a screenplay in just a few
to generate a screenplay in just a few seconds so up there is a little
seconds so up there is a little
seconds so up there is a little screenwriter Jetson as we called him
screenwriter Jetson as we called him
screenwriter Jetson as we called him because
because
because we just named him after that their thing
we just named him after that their thing
we just named him after that their thing so there was a screenwriter in the in
so there was a screenwriter in the in
so there was a screenwriter in the in the morning in three seconds flat he
the morning in three seconds flat he
the morning in three seconds flat he wrote our screenplay we wanted to get a
wrote our screenplay we wanted to get a
wrote our screenplay we wanted to get a wonderful team of people on board a
wonderful team of people on board a
wonderful team of people on board a great group of actors including Thomas
great group of actors including Thomas
great group of actors including Thomas Middleditch in Silicon Valley you may
Middleditch in Silicon Valley you may
Middleditch in Silicon Valley you may recognize him and we had some producers
recognize him and we had some producers
recognize him and we had some producers Allison Friedman Andrew sweat and you
Allison Friedman Andrew sweat and you
Allison Friedman Andrew sweat and you gave us some money and we went to shoot
gave us some money and we went to shoot
gave us some money and we went to shoot the film and that the best part the
the film and that the best part the
the film and that the best part the whole was to read through just
whole was to read through just
whole was to read through just giving these actors this screenplay for
giving these actors this screenplay for
giving these actors this screenplay for the first time like I said they are
the first time like I said they are
the first time like I said they are machines for making meaning and you give
machines for making meaning and you give
machines for making meaning and you give them this thing I didn’t tell them what
them this thing I didn’t tell them what
them this thing I didn’t tell them what the story was about and just like the
the story was about and just like the
the story was about and just like the actor with the shoe as soon as they read
actor with the shoe as soon as they read
actor with the shoe as soon as they read it to each other
it to each other
it to each other they’d not read it and that on their own
they’d not read it and that on their own
they’d not read it and that on their own this love triangle emerged from nowhere
this love triangle emerged from nowhere
this love triangle emerged from nowhere it just popped into being and and
it just popped into being and and
it just popped into being and and everyone was laughing it was it was a
everyone was laughing it was it was a
everyone was laughing it was it was a beautiful beautiful day we shot the
beautiful beautiful day we shot the
beautiful beautiful day we shot the thing cut it uploaded it and we made the
thing cut it uploaded it and we made the
thing cut it uploaded it and we made the top ten
top ten
top ten one of the jurors said I will give them
one of the jurors said I will give them
one of the jurors said I will give them top marks if they promise never to do it
top marks if they promise never to do it
top marks if they promise never to do it again another now that your said the
again another now that your said the
again another now that your said the only thing that would enjoy this is
only thing that would enjoy this is
only thing that would enjoy this is another computer both of those jurors by
another computer both of those jurors by
another computer both of those jurors by the way were screenwriters which might
the way were screenwriters which might
the way were screenwriters which might tell you something so there was a little
tell you something so there was a little
tell you something so there was a little bit of fear involved actually it was a
bit of fear involved actually it was a
bit of fear involved actually it was a bit more they didn’t we didn’t win but
bit more they didn’t we didn’t win but
bit more they didn’t we didn’t win but they did decide to interview Jetson our
they did decide to interview Jetson our
they did decide to interview Jetson our screenwriter at the award show and this
screenwriter at the award show and this
screenwriter at the award show and this was the last question from that
was the last question from that
was the last question from that interview what’s next for you Jetson and
interview what’s next for you Jetson and
interview what’s next for you Jetson and the answer was here we go
the answer was here we go
the answer was here we go the staff is divided by the Train of the
the staff is divided by the Train of the
the staff is divided by the Train of the burning machine building with sweat no
burning machine building with sweat no
burning machine building with sweat no one will see your face the children
one will see your face the children
one will see your face the children reach into the furnace but the light is
reach into the furnace but the light is
reach into the furnace but the light is still slipping to the floor the world is
still slipping to the floor the world is
still slipping to the floor the world is still embarrassed the party is with your
still embarrassed the party is with your
still embarrassed the party is with your staff my name is Benjamin
so we got a little frightened some of us
so we got a little frightened some of us where we start loading it and that was a
where we start loading it and that was a
where we start loading it and that was a lot of the response actually was fear
lot of the response actually was fear
lot of the response actually was fear there were a lot of people who were kind
there were a lot of people who were kind
there were a lot of people who were kind of afraid they’re going you’re gonna
of afraid they’re going you’re gonna
of afraid they’re going you’re gonna take creativity away from human beings
take creativity away from human beings
take creativity away from human beings you’re gonna ruin the one precious bit
you’re gonna ruin the one precious bit
you’re gonna ruin the one precious bit of work that’s left to us you’re taking
of work that’s left to us you’re taking
of work that’s left to us you’re taking it away and then other people just
it away and then other people just
it away and then other people just thought it was stupid and pointless that
thought it was stupid and pointless that
thought it was stupid and pointless that we were effectively tearing up a
we were effectively tearing up a
we were effectively tearing up a dictionary and throwing it in the air
dictionary and throwing it in the air
dictionary and throwing it in the air and sticking it together getting people
and sticking it together getting people
and sticking it together getting people to read it out we disagree with both of
to read it out we disagree with both of
to read it out we disagree with both of those that’s right personally I don’t
those that’s right personally I don’t
those that’s right personally I don’t think there’s any reason to be afraid of
think there’s any reason to be afraid of
think there’s any reason to be afraid of this I don’t think there’s any reason to
this I don’t think there’s any reason to
this I don’t think there’s any reason to think it’s stupid and pointless either
think it’s stupid and pointless either
think it’s stupid and pointless either enough there’s any reason to be afraid
enough there’s any reason to be afraid
enough there’s any reason to be afraid because the thing that I just got
because the thing that I just got
because the thing that I just got discovered working with computational
discovered working with computational
discovered working with computational systems and creative writing is that
systems and creative writing is that
systems and creative writing is that when we teach machines to write the
when we teach machines to write the
when we teach machines to write the machines don’t replace us any more than
machines don’t replace us any more than
machines don’t replace us any more than a piano replaces a pianist they become
a piano replaces a pianist they become
a piano replaces a pianist they become extensions of us they become our pens in
extensions of us they become our pens in
extensions of us they become our pens in a way and we are able to become more
a way and we are able to become more
a way and we are able to become more than writers we can become writers of
than writers we can become writers of
than writers we can become writers of writers and additionally there’s not
writers and additionally there’s not
writers and additionally there’s not really much of a reason to think it’s
really much of a reason to think it’s
really much of a reason to think it’s totally pointless because even text that
totally pointless because even text that
totally pointless because even text that is coherent at the sentence level can be
is coherent at the sentence level can be
is coherent at the sentence level can be extremely useful in certain cases you
extremely useful in certain cases you
extremely useful in certain cases you can imagine a lyricist writing a song
can imagine a lyricist writing a song
can imagine a lyricist writing a song and they’re stuck they don’t know what
and they’re stuck they don’t know what
and they’re stuck they don’t know what line should come next well here’s a
line should come next well here’s a
line should come next well here’s a machine that given that lyricist
machine that given that lyricist
machine that given that lyricist influences and their references can
influences and their references can
influences and their references can suggest a thousand possible next lines
suggest a thousand possible next lines
suggest a thousand possible next lines in just a few seconds and the lyricist
in just a few seconds and the lyricist
in just a few seconds and the lyricist can browse those maybe find one she
can browse those maybe find one she
can browse those maybe find one she likes edit it a little bit and then move
likes edit it a little bit and then move
likes edit it a little bit and then move on there’s known to be blocked up it
on there’s known to be blocked up it
on there’s known to be blocked up it introduces new paradigms for writing and
introduces new paradigms for writing and
introduces new paradigms for writing and possibly new ways of thinking about
possibly new ways of thinking about
possibly new ways of thinking about writing any ways of thinking in general
writing any ways of thinking in general
writing any ways of thinking in general and helping with writer’s block that I
and helping with writer’s block that I
and helping with writer’s block that I was laboring under in in in LA you
was laboring under in in in LA you
was laboring under in in in LA you remember that well do you know what I’m
remember that well do you know what I’m
remember that well do you know what I’m writing something even more difficult at
writing something even more difficult at
writing something even more difficult at the moment I’m adapting the work of a
the moment I’m adapting the work of a
the moment I’m adapting the work of a man called Ben Mezrich who I believe is
man called Ben Mezrich who I believe is
man called Ben Mezrich who I believe is in the room somewhere his previous book
in the room somewhere his previous book
in the room somewhere his previous book was adapted by by Aaron Sorkin who won
was adapted by by Aaron Sorkin who won
was adapted by by Aaron Sorkin who won an Oscar that’s kind of pressure that’s
an Oscar that’s kind of pressure that’s
an Oscar that’s kind of pressure that’s kind of writer’s book inducing and Ross
kind of writer’s book inducing and Ross
kind of writer’s book inducing and Ross trained Benjamin on air all of Aaron
trained Benjamin on air all of Aaron
trained Benjamin on air all of Aaron Sorkin’s work and now I can press a
Sorkin’s work and now I can press a
Sorkin’s work and now I can press a button and get a little Sorkin
button and get a little Sorkin
button and get a little Sorkin suggestion every time I need one helps
suggestion every time I need one helps
suggestion every time I need one helps break through that barrier it’s useful
break through that barrier it’s useful
break through that barrier it’s useful for inspiration Benjamin writes
for inspiration Benjamin writes
for inspiration Benjamin writes thousands and millions even of titles at
thousands and millions even of titles at
thousands and millions even of titles at a single move one of my favorites was
a single move one of my favorites was
a single move one of my favorites was the short films of lappa roost are fake
the short films of lappa roost are fake
the short films of lappa roost are fake that’s clearly about finding a deserted
that’s clearly about finding a deserted
that’s clearly about finding a deserted planet where there’s a box buried of the
planet where there’s a box buried of the
planet where there’s a box buried of the short films about the culture that used
short films about the culture that used
short films about the culture that used to live that great idea for a movie
to live that great idea for a movie
to live that great idea for a movie right I’d love to make that and we tried
right I’d love to make that and we tried
right I’d love to make that and we tried other things we um we trained on all of
other things we um we trained on all of
other things we um we trained on all of the
the
the at the previous work of an actor you may
at the previous work of an actor you may
at the previous work of an actor you may have heard of called David Hasselhoff
have heard of called David Hasselhoff
have heard of called David Hasselhoff and he performed a character that was
and he performed a character that was
and he performed a character that was really just a gift out version of all of
really just a gift out version of all of
really just a gift out version of all of his former selves something he’s sort of
his former selves something he’s sort of
his former selves something he’s sort of been doing for the last 20 years and he
been doing for the last 20 years and he
been doing for the last 20 years and he did he did that beautifully honestly I
did he did that beautifully honestly I
did he did that beautifully honestly I think it’s one the best performances I
think it’s one the best performances I
think it’s one the best performances I really loved it and so did he I think
really loved it and so did he I think
really loved it and so did he I think and that film is called it’s no game if
and that film is called it’s no game if
and that film is called it’s no game if you want to look it up there’s still
you want to look it up there’s still
you want to look it up there’s still other things we want to do we want to
other things we want to do we want to
other things we want to do we want to build other stuff but for me the main
build other stuff but for me the main
build other stuff but for me the main thing I’ve learned is that remember I
thing I’ve learned is that remember I
thing I’ve learned is that remember I wanted to make something that was
wanted to make something that was
wanted to make something that was classical and original and personal well
classical and original and personal well
classical and original and personal well I don’t know these machine systems what
I don’t know these machine systems what
I don’t know these machine systems what they seem to be good at better than any
they seem to be good at better than any
they seem to be good at better than any human being has before is looking at
human being has before is looking at
human being has before is looking at what humans have done and finding away
what humans have done and finding away
what humans have done and finding away with like a fairground mirror to reflect
with like a fairground mirror to reflect
with like a fairground mirror to reflect how all of our paths work to ourselves
how all of our paths work to ourselves
how all of our paths work to ourselves which isn’t that what you do when you’re
which isn’t that what you do when you’re
which isn’t that what you do when you’re trying to learn what filmmaking is or
trying to learn what filmmaking is or
trying to learn what filmmaking is or what theatre is it’s it’s doing that but
what theatre is it’s it’s doing that but
what theatre is it’s it’s doing that but it’s doing it in this more sophisticated
it’s doing it in this more sophisticated
it’s doing it in this more sophisticated way so it shows me what’s happened
way so it shows me what’s happened
way so it shows me what’s happened before and in a way that helps me to be
before and in a way that helps me to be
before and in a way that helps me to be original because I can see the cliches
original because I can see the cliches
original because I can see the cliches and then break them and as as far as
and then break them and as as far as
and then break them and as as far as personal is concerned well every time
personal is concerned well every time
personal is concerned well every time you’re inspired by a shoe or a bit of
you’re inspired by a shoe or a bit of
you’re inspired by a shoe or a bit of random something you’ve seen and it
random something you’ve seen and it
random something you’ve seen and it makes you think of something new I think
makes you think of something new I think
makes you think of something new I think that’s when the personal comes out so
that’s when the personal comes out so
that’s when the personal comes out so anything that can inspire you does that
anything that can inspire you does that
anything that can inspire you does that thank you so much for listening I hope
thank you so much for listening I hope
thank you so much for listening I hope you can thank you
you can thank you
you can thank you [Applause]
”
Be First to Comment