August 14, 2012 • 45 mins
When it comes making art, most of us operate under a set of rules. So what happens when you feed machines code for churning out art? Plenty, as it turns out. And if creativity is the seat of the soul, what does that say about our uniqueness?
Learn more about your ad-choices at https://www.iheartpodcastnetwork.com
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:03):
Welcome to Stuff to Blow Your Mind from how Stuff
Works dot com. Hey, welcome to Stuff to Blow your Mind.
My name is Robert Lamb and I'm Julie Douglas. We're
both humans. Needed neither of us are machines. We depend
on a number of machines but to make this podcast happen.
(00:24):
But we are both for the most part, human beings.
So far we are. Yes, we have not been replaced
by machines. And to a certain extent, this show is
our art, right, I mean that's what humans do. Humans
make art. It's it's kind of an expression of our humanity,
our desires, our uh scary thoughts. Yeah, yeah, it's this. Yeah,
(00:47):
it's kind of this output, this uh, this expression of
self that uh, it is kind of like the bat
signal of human culture, letting everyone know that we've got
it going on, that we are. Uh. And when I
say everyone, I guess we're trying to tell the animals this.
But um, But but what I'm saying is that we
(01:07):
we have for a long time just considered art as
this thing that humans do, and it's uniquely human and
it's not the sort of thing that a machine could
really manage. Yeah, And we know, we have talked about
it a little bit. In the animal world, occasionally you
can teach an elephant to paint. Yeah, you can get
an eight to get in there and get its its
(01:29):
hands covered in paint and start to having around certain
birds that may make their their nest really like much
more beautiful in order to attract a meat. We've seen
and ultimately you can take things out of that are
out of context and non art, put them in within
an art context and they become art. So, I mean,
(01:51):
it's all in the eyes of the hold the beholder.
But we're the beholder, right, So we're the We're the
ones who get to decide what is art. We're the
ones who get to make it um or at least
that is the way it has been. But robots, robots
are changing this. Robots are are are challenging our assumptions
um concerning what art is, who can make it and
(02:11):
uh and what makes it good? And actually what creativity
is at the heart of the matter here um Studio
three sixty dot com has some really great articles a
suite of articles on creativity and artificial intelligence, and that
is in part what prompted us to start thinking about this.
Today and podcasting on it. Yeah, it's a great show.
If you've never checked out Studio, give it a shot. Um. So, yeah,
(02:35):
we're talking about creativity, um and robots. Now, we did
a whole episode on creativity where it comes from in
the human mind. So creativity in and of itself is
a rather complex idea. There are several different definitions for creativity,
which we explored in that that episode and some depth.
Some of these definitions are more easily appliable to robotic
(02:59):
card this than others. For instance, there's the cognitive view
of creativity. Uh, and this is in this one. A
computer could definitely do it because it's a process by
which the sensory input uh is transformed, reduced, elaborated, stored,
recovered and used. So it's just a it's just an
input output scenario here. So it sounds like memory retrieval
(03:20):
as well, right, Yeah, it's and uh and you can
see this kind of quote unquote creativity at work in
simple You know that you'll find programs online where you
input tax and we use a cut up system to
mesh it around and change it. Uh. You know, random
if you were to do like a random filter on
a on a photograph. I mean, it's a very simple
(03:41):
version of what we're talking about. But essentially, data is
input in, data is changed, and data is output it. Um.
Then there's like the psychoanalytic Freudian view, which argued that
creativity is an occurrence of the subconscious. This not so
easily appliable to machines. Well, we'll get into this in
a little bit, but there is a aspect of this
(04:04):
that is happening in artificial intelligence. But yeah, there's this
idea that art occurs creativity occurs sometimes from these different
states of consciousness, and particularly with unconsciousness. In fact, Sovageot
Dolly supposedly would sit with a spoon in his fingertips,
holding it over a tin cup or plate as he
(04:24):
drifted off to sleep, and then the spoon would fall
from his grasp and wake him up again. And presumably
he did this so that he could spend longer on
the fringes of consciousness unconsciousness and retrieve new ideas. Um. Yeah,
And there have been other artists that have should have
have done that with a silver spoon before, but in
different ways. For some reason, I thought you're gonna say
(04:46):
a chainsaw. No, no, yeah, But but Okay. Other quick
definitions of creativity. There's the behaviorist view, which says that
creativity is a combination of previously no knowledge joined together
sponsor enviously um and uh, which gives the impression of
a new idea or inspiration. So this is kind of
an idea collider idea of creativity and uh. And that
(05:11):
matches up really easily with our ideas of machine intelligence,
that a machine without any kind of real gusto could
get in there and say, okay, um um, zombies are good,
and pride and prejudice is good. One plus one equals too.
Now I'm using creativity. I mean, that's a very granted
(05:35):
that idea was was was generated by a human, but
it's very easily too easy to imagine a machine coming
up with that as well, like looking at, say, you know,
if you had a computer that was analyzing like search
terminology online online, seeing what people were searching for and
just picking the top two categories and crafting a novel.
So it's very much like decision tree decision making. And
(05:57):
so the question arises, can there be a purely creative
product that comes out of a machine? And the other
question is, you know, is a sacred ground. Is it
really unique to humans? And that's what we're going to
really get to explore today. Yeah, it's an interesting look
at just what computers are capable of now and uh
(06:20):
in an idea of where they're going. But also it
really forces us to reevaluate what's going on in creativity
and what's going on in art and and you know,
it just really forces you to re examine how we
view the whole scenario. So, um, we should probably start
in the written word robotic fiction. Yeah, and apparently this
(06:43):
has been around for a while, right for something called
Mark one Baby in a nine British computer scientists Christopher's
stra Key program the Mark one baby computer to generate
love poetry from a database of romantic verbs and nouns. Yeah,
back in the fifties, Back in the fifties. And here's
here's a sample. I'm not gonna read in robotic Okay,
(07:06):
I'll try. I mean, you know how well I do
with other accents. Well, the robot accent is pretty pretty easy,
all right. Honey, love, my anxious fondness impatiently yearns for
your craving enthusiasm. You are my breathless infatuation, my darling, affection,
my ardent, liking my covetous fancy yours anxiously and you
(07:32):
see there you go, that was good. That was okay.
I mean, I'm sure the computer is listening. We'll probably
right now. Yeah, but yeah, here's here's an early example
of this. But you know, fast forward to do today
and we have something called Brutus, which is a machine fabulous. Yeah.
(07:53):
Brutus uses mathematical equations um that are based on the
basics of plot, of setting, of dialogue and uh and
really this if any of any of you that have
ever taken classes on on the structure of a novel
or how to write a short story, you know that
their their formulas out there, and you can really and
(08:13):
you can also look at it like that there's so
many different different guys out there and and and women
that are they have like these models of how to
plan out a novel, how to plan out a short story,
and they're you know, they're like triangle approaches and and
uh you know, multiple story arc approaches, and there's like
a snowflake um approach to writing a novel and it's uh,
(08:37):
so we're really big on formulating it and really guiding
it out you know, as if to make it to
where just anybody could follow these instructions and come up
with a novel. So so it it follows it's really
easy to imagine then a computer taking some of this
this data and using that that that information is an
equation to generate a novel. It knows the the structure
(08:58):
of plot, the the way a story tends to to roll.
Uh and uh, and it can can generate that. Yeah
and uh, I mean it is really cool actually, the
fact that it can generate these these stories. Now there
are limitations. I mean, we're talking about five words stories.
And programmer Selma brings your He said that the program
can't be considered creative unless the machine creates something that's
(09:21):
completely demystifies the programmer. So in other words, if it's
not demystifying um, and it's putting out the expected outcome,
then it's just cloning our intelligence. We'll see. This reminds
me a lot of the scenario of fan fiction um
in uh with with human writers. This is a pretty
(09:42):
simple concept that everyone's encountered. Take takes. Say, someone like
HP Lovecraft wrote a you know, a ton of short
stories and fragments and the occasional slightly longer work, uh,
with a very particular brand of horror. And you have
you have people that read and I mean think I've
probably read most of his his work, but you have
people that read most of his stories, not all of
(10:04):
his stories, And then it's pretty easy at that point
to sort of regurgitate something that is at least love
craft ish in that style. Yeah, you figure out what
he tends to do in a story, what format he
tends to follow, and then you can make something that's
love craft ish. And and certainly there have been a
lot of authors who have have enjoyed some success with
(10:25):
lovecraft Is stuff. But but where you really see something
phenomenal is when someone is clearly inspired by someone like
Lovecraft or or or um Or or various other authors
that have that certain feel and look of their their
own material, where they can take what that author has
done and then spin it around, do something different with
it in a way that makes it their own. Well see,
(10:48):
and this comes the originality really comes into question in
these sort of cases, because you could argue that there
is nothing nothing unique has ever really been created, right,
that it's just intertant. We're all taking ideas colliding them
and and trying to put just enough spin on them
that we don't get sued. It's just iteration of iteration
(11:08):
of iteration, unless you're DaVinci, right, yeah, just coming up
with all these completely original ideas for new machines. But
that's because he stole a time travel machine. I knew it.
I knew you. You finally said it. It's true. Um,
so yeah, I mean this idea that that this program
is spitting out stuff that's not too original. Well, okay,
(11:30):
I mean you really get into shades of gray of like, well,
what at what point does it become in the realm
of uniqueeness? And and um, just you know, just regurgitating
a bunch of rules. Yeah, and are you which model
of human creativity you're holding it up to. You're holding
up in the Sistine Chapel, You're holding it up to
Dante's Inferno. You're holding it up to pride and prejudice
with zombies, which is the main problem of subjectivity right
(11:53):
when it comes to art. Um. Okay, So I did
want to say that that Brutus created this this short
little or here, and he paints a picture of a
guy named Dave Sdriver. He's a PhD candidate and he's
ambivalent about his current situation and the program pus dashes
of details in there, just as you said, like, there's
(12:13):
very formulaic in some ways. Um, you know, just to
create a setting. Brutus says, uh, here's the Ivory clock
on the campus and eager students. You start to get
a real sense of place. And the pacing also gives
it a sense of drama and foreboding. And I wanted
to play just a really short clip of this. This
is from Studio three sixty and this is a moment
(12:34):
in which Dave Striver is being waylaid by the dissertation
committee for his PhD. Fiery and uncontrollable rose up Dave.
Professor Hart products softly, God, they were pitiful, pitiful, palette
and puny, Dave. Did you hear the question? Later Striver
(12:56):
sat alone in his apartment. What in God's name had
he done? And it should be noted that that was
not the robots voice. That was that was human reading. Yeah,
that was a real dude reading. But I mean you
kind of do get a sense of that, right, That's
not too bad? Yeah, I mean that's not I mean
for a computer to sort of split that out, and
for for to make sense in the first place is
(13:16):
kind of amazing, But for it to give you a
sense of of peering into another person's mind, that is amazing,
because does a machine know what a human mind is? Well?
And certainly learning about it? Because ultimately mean the machine
is dealing with language is kind of like a shadow
of the human mind. It's um you know, it's dealing
(13:37):
with coded data, semiotic web of of meanings and symbols.
And if you align it properly, you create that you
can create a text that has either a little or
a lot of meaning. Well, it's just to what extent
can the machine understand the human mind? I think? And
to that end, it turns out that Patrick Winston, he's
(13:58):
a principal, principle and stigator at m i t S
Computer Science and Artificial Intelligence Lab. He is teaching Macbeth
to computers and he says that storytelling makes it possible
to teach, to learn, and to be creative, and so
he wants these machines to understand humans. And what better
way than to start off with Shakespeare. Yeah, and as
(14:20):
far as teaching, learning and creating GOUM I mean human
beings go pretty long way with just faking any or
all three of those categories. So really, I mean, as
far as understanding the human mind, yeah, I don't know,
that's really not even as important necessarily for a machine
as just the simblance of it faking it, because if you,
(14:40):
you know, create, if you can create something that we
end up investing thought into, then then that is the
insight into the human mind. Right, maybe I'm getting a
little no, Well, I think what you're saying is that
the simulation at some point becomes the thing. Right, So
even if it doesn't truly understand it, if it's executing
(15:03):
the code for it. I mean, because we've all encountered art,
be it um you know, be it, be it a
work of written art or visual art or music where
we take that moment we think if this person brilliant
or are they just pulling one over on me? You know. Well,
now machines have the same opportunity exactly. Uh well, you know, okay,
(15:24):
this this guy who's teaching them Macbeth, he actually has
as a point here beyond just human um mind uh comprehension. Here,
it's is actually to try to help creatively problem solve,
which would make sense of storytelling. And he says that
what we're hoping to do is build systems one day
that are as important to a political analyst as a
(15:44):
spreadsheet is to a financial analyst. We have a system
now that draws parallels between various kinds of historical events,
the Arab Israeli War and the tet offensive, for example.
Finding those parallels could help keep people out of trouble. Okay,
so now we're getting into a larger area than of
robotic design, not just not just robotic writing, but just
(16:07):
robotic creative thought. Well, and it reminded me of the
Living Earth simulator that we've talked about before, which is
this idea that you could take every bit of data
in the world, throw it in there and try to
predict what's going to happen next, whether it's a weather
pattern pattern or um uh, you know, politics, um or
(16:28):
even economic ramifications throughout the world. It's interesting. I'm currently
reading the Hyperion series by Dan Simmons, and in that
novel you have a you have artificial intelligences that have
sort of succeeded, succeeded from from human civilization and that
(16:50):
still advise on various levels and are still involved but
also rather separate. But they have it to the point
where they know exactly how, um, everything is going to
transpire for years and years in the future, with the
exception of one sort of X factor that's pivotal to
the novel. But but but yeah, that back to the
(17:11):
Living Earth simulator, just the idea that that if you
put it up, you put the information in there, to
what extent can a machine navigate all the variables and
come up with an accurate prediction of how things will
play out. Well, we've talked about the limitations here with
a Living Earth simulator simulator because it doesn't take into
account or this version of it as of yet hasn't
(17:34):
been able to take into account what we call these
black swans or these events that seem to come out
of nowhere. But when in fact you look at it
like you know that there's going to be some sort
of entropy at some point, it just feels like it
came out of left field because we didn't expect it. Right. Well,
I mean, it's like human creativity. You know, Um, we'll
have we'll have one thing in mind, then something else
(17:55):
comes along and it changes the form of what you're
trying to create you know, um, I mean, like I
remember I started writing a short story before my dad died,
and then after my dad died, that changed the shape
of that story. Like that was a black swan event.
I mean, not that you know, people die, that's kind
of the It kind of happens all the time. So
not in the sense that it's that it's that unforeseeable
(18:17):
an event, um, but but it occurs and it changes
the direction of the creativity. So so what you're saying
here is that we have to have a creative program
that is also um that we throw a few black
swans at every now and then, Well that could also
create its own black swans. And we should probably take
a break here and we get back, though. We're going
(18:38):
to talk about whether or not there is a machine
that can break this kind of code and possibly even
create new unique thoughts. All right, all right, and we're
back and uh yeah, we're going to talk about a
little something here called the creativity machine. Yeah, okay, so
(19:01):
this is not a you know, something that we read
about in a science fiction novel or something something like
the near not the near Earth simulat not not like
the the Earth Simulator that is that is that may
exist in the future that we're planning. This is something
that he's out there. I actually thought about the Santa
Claus machine when I was reading about this, and we
talked about this in another episode. That of course is
(19:22):
something that that could exist in the future, but doesn't
exist exist now, Okay, so let's get to this creativity machine.
It is pretty much like the ultimate problem solving machine.
There is a guy named Stephen Fuller and there's a
great documentary called In Its Image that covers this. Since
the seventies, he has been fussing around with artificial neural
(19:42):
networks and he has conducted a lot of experiments with us.
Usually people have been dealing with artificial neural networks just
to kind of see the way that the brain processes information.
But he he has a different plan um with his
creativity Shine by using the human brain and trying to
(20:06):
not only replicate the way that neurons the neural network
is working, but also how originality arises in humans. He's
trying to do the same thing in in computers, and
you could argue that he has done it. Yeah, the
created creativity machine learns on its own and uh, and
it and it deals with black swans mathematical variables that
(20:29):
are thrown in to shake it up. Yeah, I mean, okay,
it's got these human generated scripts. UM. It is an
artificial neural network, but it has two distinct functions. One
part contains data and one part runs the data with
tests and gives it feedback. So that's where it learns
from itself. UM. And then occasionally Solar will actually disrupt
these functions by running mathematical noise into the system. That's
(20:53):
the black swan, which creates a false memory of confabulation.
And this is where the machine gets confused and it
begins to drive the artificial neural network to think in
new ways and produce new ideas. And here's an example
of this. UM Follower was actually commissioned in by Oral B,
the toothbrush maker, to create some new designs for the toothbrush,
(21:17):
or their toothbrush as I should say. And he said
that when they injected noise into the first network, it
imagined not the memories of toothbrushes it had previously seen.
It produced confabulations, false memories that were hybrid hybridizations of
what it had seen before. So when he when he
(21:38):
threw that box swan in there all of a sudden,
it didn't. It did create these original new thoughts that
were based on those memories, so it retrieved its memories,
but they were false memories, which is really sort of
the same thing that happens when we have our own
aha moments and creativity. Right, we have these breakthroughs based
on the experience the memories, but you know, come comes
(22:00):
out and a bit differently. Um, he says, you know,
I didn't do anything crazy here. I basically pushed a
return button and sat back and watched two thousand designs emerge,
and then he gave the customer or it will be
two designs, okay, So that sounds like okay, or it'll
be toothbrush whatever. This guy, as I has actually used
(22:22):
this machine to create everything from like snack food to
military devices like robotic swarms. Yeah, he's that. It's been
used to create such as harder than diamonds, one point
five million new English words, it's trained robot cockroaches, and
it's um, it's composed just a whole lot of music, yeah,
(22:43):
which is really cool, and we'll listen to a clip
of that music in a moment. Um. But this is
a really cool thing to uh. Fohler actually filed the
first patent for the autonomous generator generation of useful Information,
and then his second patent was for his self training
neural network object. Patent too was invented by patent number one.
That's crazy, that's amazing the machine created the patent. I
(23:06):
mean this is this really does kind of spell out
a very interesting future with this technology. Um. He says
that this is the way that we're going to cure
cancer and the way that we'll do multi trilling and
dimensional optimizations to bring some peace of mind for the
world's economy. Okay, so think back to the Living Earth
simulator and it's limitations. If you were to marry the
(23:28):
two technologies, is there the possibility that this machine could
become like, you know, basically a crystal ball for us?
It could be like oral b comes to the the
the AI s then and says, hey, we want to
to roll out a new toothbrush, but we also want
to know how this toothbrush will change the world. So
the computers begin to say, all right, here are ten
designs you can choose from, and here is a layout
(23:51):
of how each of these designs will affect your financial
well being as well as the world at large. Right,
could you could overlay all that data, right, you could say,
if you decide to produce it in this area of
the world, this is the impact it will have on
that economy and on on those people. You know, if
you did it here, it would have this sort of impact,
so on and so forth. You could look at it's uh,
(24:12):
it's carbon footprint. I mean, you could look at all
sorts of slices of data on top and ultimately that's
what the the Living Art Simulator it comes down to
as well. It's it's not really so much an idea
that will we will create a virtual reality Earth too.
It's it's the idea of, well, we have in this
simulation as much data as we want to throw at it.
(24:33):
So you would say, all right, I'm interested in making toothbrushes,
so I want to know how how this device might
affect um you know, this particular hygiene market, how it
will play out you know, legally, how it'll play out
with profits, how it will play out with you know,
with the with the perception of the brand. You know,
whatever the variables happen to be, as long as you
have the data to back it up. Yeah, of course,
(24:54):
you know you would have to mesh these two technologies
and that that right there is the fiction part of this.
But those two technologies do exist stand alone, and so
it is it's very exciting to sort of extrapolate what
that might look like in five years, ten years, twenty years. Um.
I think that this this is the great opportunity to
(25:15):
hear a little clip of music by this machine. It's
actually called the Dingularity and uh, this is from the
album Song of New Neurons. And again this is this
is from Studio three sixty. All right, but we're talking
(25:42):
about are here, so let's let's really get down to
the to the paintings, now, to the to the to
the canvas itself. UM, this is nothing new either. There
any number of robots out there that have been designed
to put a little paint on a canvas. Um. I
mean you can and on a very basic level, we
have robots out there painting cars every day. The works
(26:05):
pretty basic, it tends to me it's it's it's very
often one color, but it is a robot painting something.
So it's very easy from there to extrapolate, Um, robots
painting various other things. Actually we've been able to We've
trained elephants to paint. We've you know, you you get
an ape uh hands dirty, it will paint something on
(26:27):
a canvas. Um. So so yeah, we have machines that
are capable of this as well. UM. I mentioned a
few of them here that are um that are fairly interesting.
For instance, there is uh the van Go bot, as
in Vincent van Go and this, uh, this is a
(26:48):
particular bot boast eight teen brushes, a paint mixer, and
three D spatial awareness, so it can combine artistic influences
to create fresh takes on a given subject. So the
very ables here being a subject and what styles you
want to see. So you can sort of think of
this in terms of almost like like an Instagram kind
of a thing. Imagine if if a robot were in charge.
(27:11):
All right, the photo is of a dog, and say
I wanted to combine filter one and two, and it
would combine those in a way that made sense. So
you might say, all right, I have I have a
picture of a bowl of fruit. I would like that
in a way that's cubist and uh and also surrealist. Well,
and then there is what seems to be a more
(27:34):
nuanced program called Erin, Right, and this is where, um,
you get more of the expression. I guess you could say,
even in the eyes of the people that it's depicting
in its paintings. Yeah. Erin has been going since about
nineteen seventy three, or the project itself, and uh, it's
the ranch out of Harold Cohen. Uh, Aaron drawls and
(27:57):
paint stylized still lives and portraits of human years out
of its programmed quote unquote imagination. Um, no, no images
or additional human input is necessary. So it sees something
and then it creates an artistic interpretation of what it sees,
which would point to original thought. Right, it's kind of interesting, yeah, okay,
(28:20):
and then the thing again, it's like once it's created,
like it is, it creates this thing on cannabis. Then
that then interacts with the human mind. And we've talked
before about all the stuff that goes on in the
human mind when you look at a piece of art.
So you know, even if there's if it's not thinking
about subtext or whatever, it can't it's not actually quote
unquote thinking. Um, if the piece ends up summoning that
(28:44):
thought in the viewer, then I mean it's art. Yeah,
I don't know. I mean I've seen some of the examples,
and you can go to Studio three sixt dot com
and see the samples of them. Um, that's a good
place to start at least. But some of them really
do elicit some feelings, I have to say, but there
(29:04):
is a flatness to it that you can kind of
tell that it's computer generated. Um. But that being said,
there's uh, there's that I can't remember the first name
of the artist, but it's Fred's grandson or something. Um.
There were some scenes of people that reminded me of
Fred's paintings that have a certain mood to them, a
(29:24):
certain quality, uh, in the way that it's spatially arranged.
And I thought that was interesting because I thought, well,
there seems to be some sort of psychology going on
in this painting. Yeah. But but then again that that
reminds me of of any number of cases of you'll
see like an interview in with an artist or an
actor or musician and they'll they'll be like, oh, you
know when you were when you did this particular thing,
(29:44):
when you decided to paint this blue instead of gray,
or you know, what was going on in your mind,
and they'll have all of this, this thought invested in this,
in this one little detail, and then when the artist
just says, oh, yeah, well we ran out of gray
paint that day or lit or it will be something like,
oh yeah, well it's different live than it wasn't an
album because we kind of forgot how the music how
(30:07):
we actually played at the first time. You know, there
ends up being some sort of mechanical explanation for something
that you thought was going to be really just tied
into the creative process. Well so, so much of that
then is projection by the viewer, right, Okay, So, just
in case you were worried that there there wasn't some
sort of computer around here to try to take the
(30:27):
highly subjective um job of making, uh this the stance
that some art is good and someone is bad. Don't worry.
There's something called DARCY. It's the Digital Artist Communicating Intent Machine,
and it's essentially kind of like an art historian. They
are teaching this computer to basically build a DA database
(30:50):
of artwork in order to carry out this subjective task
of saying this is good and this is bad, and
basically dissociates images with an adjective, and it can learn.
It can build up this database and begin to learn
what is considered good and what is considered bad. And
then you can actually throw another image in front of
it and it will rate it, which is interesting. Computer
(31:12):
scientists from Brigham Young University actually have been working on this,
and we missed it when it came to Atlanta. It
came to the High Museum of Art and judge the
works of Man it did. It dashed a lot of
people's hopes to be artists. Apparently a couple of kids too,
got some some thumbs down from Darcy. Um. All right,
(31:33):
so let's talk about film. Could could a computer be
an all tuo of sorts in this medium? Well, certainly
we're now we're dealing with well, of course we're dealing
with this with fiction, but um, with film, with digital
film editing, it's it's all digital. All that information is
just data in the computer. And and again this is
(31:53):
another area where there are certain rules, there are certain
types of cuts, there are certain uh there's a certain
uh law that you abide by when you're editing a film.
So it's perfectly believable that you could hand over the
raw footage and the basic flow of the piece. And
and you could expect a machine to to give you
some results, whether it would be the results on par
(32:16):
with the work of a great film editor. Well that's
the discussion. Well, yeah, that's that's the question mark here. Um.
The film is called White on White Algorithmic Noar, and
it's by Eves Suessman, and it's kind of build as
a paranoid sci fi thriller. And uh, Eve Sussman says,
it's about an American engineer Holts. He gets a job
(32:38):
in a mysterious form place called City A and starts
finding out things are wrong. The water supply seems to
be drugged with lithium, time seems to be slowing down,
people are running out of language, and the people who
hired him may not be who he thinks they are.
So everything is a little bit beyond his control in
this this situation very calas kind of scenario. Yes, yes,
(32:59):
it's very tough us and uh, it actually is very
interesting to watch it. And here's another reason why it's interesting,
because every time you watch it, it is a different
scenario because of this computer program which is essentially re
editing the film. Um, every time you view it, she
Eve has been worked with programmer Jeff Garnot and he
designed the computer system, so there's no one version of
(33:21):
white on White. This is interesting because this ties into
some of the the ideas we've discussed about the future
of music, like the idea that in the future, um,
we won't so much as listen, we won't listen to
a piece of music that it was composed by an
author or an artist. Rather, but we would listen to
a piece of music composed by an artist that then
taylor's itself to our particular taste and needs or an environment,
(33:45):
which is just like the the robotic piece of music.
It's called Emily Howell ye or that's the program. David
Cope designed it, and it composes classical music based on
influences from Bach and lose Art, And some people say
that you can't tell the difference at times. So that's
interesting that you say that, because if you were, if
(34:06):
you knew that you were interested in that and then
that's what it was inspired from, then it would no
longer be actually the works of Bach and Mozart. I
think of some of the films that have like they're
famous for multiple cuts. Uh, the one that comes to mind.
Most obviously and most fitting for this conversation is Blade Runner,
where some cuts lean more towards h Harrison Ford's character
(34:27):
being a replicant and Natroid an artificial person versus him
being human. So what have you had in the future?
What if a cut of Blade Runner Taylor's itself to
your particular values. If you're the kind of person who
is more satisfied with with with Harrison Ford being a robot,
then that is the version you end up seeing. Likewise,
(34:47):
if you're if you're the kind of person who thinks
Han Solo should should shoot first or you should shoot second,
then then a copy of Star Wars Taylor's itself to
what kind of a viewer you are? Well? Then, oh,
and then to say nothing of of curse words, nudity, violence.
You could have the film automatically adjust itself to the
(35:11):
age of the viewer or even the sex of the
viewer versus Maleston. If you they're sort of demographical variables
in play, it's no longer tied though, to that particular
person's perspective, and is no longer the artwork of that person.
Correct Well, you would have to have some very we
like for instance, the language thing is pretty easy. Like
(35:31):
there's there are artists out there where be like, oh no,
you can't, you can't have scarface say um, dang it,
you know you you this this character, this character's cursing
is is important to the work, and if you take
the cursing out there, it's no longer that work. But
there are other my little car garages, Yeah, you can
really leave the cockroaches. Yeah, but but I feel like
(35:54):
there are other films out there where the artists would
probably say, no, I don't think it harms my artistic
vision if they say fudge instead of the F word. Oh,
I don't know. Budge is a very different thing than
the other thing. Well maybe, but you get my point.
I think there depends on the level to which the
work adjust itself. Okay, well, and I suppose it depends
(36:15):
on the author of the work too, and how she
or he intends it to to interact with it. Take Brazil,
for instance, Terry Gillis film, there was a there was
a like a happy ending cut of the film that
the studio did, which is horrible. Um, but it has
a very happy ending, whereas the actual cut of the
(36:35):
film as a really down ending, like even the sort
of even the compromise, the cut that they ended up
putting together is really kind of a downer. If you
had a version of Brazil tailored itself to your viewing
and would sometimes choose the happy ending, that that would
definitely compromise the piece. Again, I think it all goes
(36:57):
back to what the what the person is trying to
say with particular piece of artwork, which goes back to robots.
Do they have a particular perspective when it comes to that,
you know, can can they really have a perspective? You
feed them lots of data about the environment, if they
can learn from themselves, uh and create original thought, there's
(37:17):
the possibility there, right, um, But let me let me
before we get too far into that. UM, I do
want to play this clip of Emily Howell. This is
the music that we're talking about that is inspired by
Bach and Mozart, just to give you guys a sense
of what that sounds like. Now. I'm not the I'm
(37:47):
not the biggest kind of sewer of classical music, but
but I certainly I don't think i'd be able to
tell you if that was Bach, Mozart or a computer. No,
And I think that The issue that a lot of
who are having with this this iteration of music at least,
is that there's not a standout piece yet. And and
(38:08):
even David Cope, who designed it, says like there's there's
not anything too terribly creative or original about this music.
It's just a variation on a theme that someone else
has done before. So eventually though, there presumably there is
going to be some sort of breakthrough, some sort of
piece of music created by a machine, possibly even the
creativity machine, that people say, oh wow, this this is
(38:29):
really struck a chord in me as a human. It
reached this the center of emotion in me that I
didn't think that a robot could inhabit my perspective in
order to elicit that feeling. So then the question becomes, well,
who is the artist here? Then? Is that the programmer
or the computer? And at what point do we consider
(38:50):
human creat creativity be the same as machine creativity. And
here's the other question I have, because we've talked about
human cyborgs in the way that we are always uh
manipulating ourselves with technology, is this just another way to
manipulate our abilities? So is this part of the human
experience to some degree. Yeah, Like when I when a
(39:11):
human creates a robot that then creates art, is it
really is that human just creating via proxy? Is it? Is?
It kind of like saying, like, like looking at the
difference between me singing and me creating music with a trumpet. Um,
you know, I'm with the trumpet. I'm using a piece
of technology that amplifies it amplifies and changes my output
(39:32):
into a different type of output, but it's still dependent
upon my original output. Yeah, I know. It's like it's
it's an extension of creativity. Right. Just because you use
something to to express yourself, albeit you know, a machine,
it doesn't mean that it's not somehow an expression of
the human self because we're the ones who coded it. Um.
(39:52):
So I think it boils down to this idea of consciousness.
I really do. And I think that the creativity machine
and its ability to mimic these little blips and consciousness
and I'm not gonna call them unconsciousness because that's probably
giving a little bit too much credit. But I think
until we can really crack this nut of what consciousness is,
(40:13):
we can't really answer the question about whether or not
robots uh could emulate it. Or or create truly unique art. Yeah,
because I mean when you get into the motivations, like
something like Dante's Inferno, you know, the the artist wants
to create this piece, and he wants to create a
piece that that that says something about about God, that
(40:39):
says something about faith, that says something about the world
that he lives in, that takes jabs at people that
he doesn't like. You know, I mean all these all
these things that go on and in so many works
of art, you know, where it's there's a there's a
little bit of themselves in the work as well. Um
to what extent can a computer ever pull that off? Well,
(41:00):
So some would argue that man humans have been creating
art for at least thirty five thousand years, right, like
from cave drawings onto up to like creating a program
like Aaron that can create art, and that machines have
only been creating art for about forty years, so that
you know, we just need to give him a chance. Yeah,
(41:23):
because yeah, when we were forty years in we probably
we weren't we well, we weren't really all that impressive either, right, So,
and and you know, because they can carry things out
at much more rapid speeds than we do. Presumably they
could become pretty sophisticated artists. I don't know or you know,
is this predicated on this idea of a soul and God?
This is for another podcast, but is there? I mean
(41:44):
for the next podcast, because we're gonna that's true, we
are going to Before we do that, we have to
close this one out. So indeed, so there you have it. Um,
some thoughts on the computer as an artist, the art
as a creation of machine versus a creation by and
and then as they work that is ultimately viewed by
human or in the case of Darcy, by a computer. So, UM,
(42:07):
we would love to hear from from you guys to
you know, what what do you think about that? What
do you what? What is art? Do you think a
computer can really create? Uh? Music, literature, UM or visual
art in a meaningful way? Uh? And then how much
of it just does does it depend on the view
or the reader or the listener as to whether it's
art or not? Let us know. We'd love to hear
(42:29):
what you think. UM. I have just uh already one
quick listener mail if we want to call the robot
over here. It's always a little awkward calling the robot
over here after we've been talking about robots. Yeah, and
all right, now this one is related to movies. Uh.
Our listener Sean writes then about horror and talking about
(42:50):
old versus new horror. Uh, Sean says, really quick. My
observation on new horror versus old horror in the mental
versus the visual. The old schoolers like Hitchcock and Uh
and King left what happened next up to the mind.
Now with modern technology they show you the gore and
leave nothing to the imagination. I know there are except exceptions,
(43:11):
but I feel new horror is no more scary than
it is disturbing. I also find myself more frequently feeling
that new horror glorifies individuals with psych psychotic tendencies, especially
when the character is based on a real individual. Great
job every week, Matt Love Sean, Um, Yeah, I mean, uh,
I've certainly seen that argument that uh that that you know,
(43:33):
now horror is more about just really throwing up disturbing images.
But but but even then, it's like I feel like
when when when horror movies show disturbing images, it's like
that you disynthesize to that so quickly that it's there
has to be if it even if a disturbing image
is going to be effective, it needs to have an
(43:54):
underlying disturbing idea. And and sometimes I think that the
disturbing image actually in the way of the of the
the idea. And to tie this back into what we're
talking about today, I sometimes feel like when you'll you'll
look at horror that it seems to work or almost works,
and it'll you'll because you think that there is a
(44:16):
disturbing idea at the heart of it. And in some
cases it just seems to turn out that the the
the creators of this particular piece are just they're just
dealing in the visuals. They're just dealing in the disturbing
images side of it, and any um in any disturbing
idea beneath it is just the viewers perception. So you
(44:37):
get into ideas of you know, something can actually seem
to have more ideas and more more depth to it
based on what the view or the listener, the reader
is bringing to this. So again projection. So so yeah,
let us know what you think about computers creating art.
You can find us online. You can find us on
Facebook where we are stuff to blow your mind and
(44:58):
you can find us on Twitter where handle is below
the Mind, and you can also drop us a line
at blow the Mind at discovery dot com
Welcome to Stuff to Blow Your Mind from how Stuff
Works dot com. Hey, welcome to Stuff to Blow your Mind.
My name is Robert Lamb and I'm Julie Douglas. We're
both humans. Needed neither of us are machines. We depend
on a number of machines but to make this podcast happen.
(00:24):
But we are both for the most part, human beings.
So far we are. Yes, we have not been replaced
by machines. And to a certain extent, this show is
our art, right, I mean that's what humans do. Humans
make art. It's it's kind of an expression of our humanity,
our desires, our uh scary thoughts. Yeah, yeah, it's this. Yeah,
(00:47):
it's kind of this output, this uh, this expression of
self that uh, it is kind of like the bat
signal of human culture, letting everyone know that we've got
it going on, that we are. Uh. And when I
say everyone, I guess we're trying to tell the animals this.
But um, But but what I'm saying is that we
(01:07):
we have for a long time just considered art as
this thing that humans do, and it's uniquely human and
it's not the sort of thing that a machine could
really manage. Yeah, And we know, we have talked about
it a little bit. In the animal world, occasionally you
can teach an elephant to paint. Yeah, you can get
an eight to get in there and get its its
(01:29):
hands covered in paint and start to having around certain
birds that may make their their nest really like much
more beautiful in order to attract a meat. We've seen
and ultimately you can take things out of that are
out of context and non art, put them in within
an art context and they become art. So, I mean,
(01:51):
it's all in the eyes of the hold the beholder.
But we're the beholder, right, So we're the We're the
ones who get to decide what is art. We're the
ones who get to make it um or at least
that is the way it has been. But robots, robots
are changing this. Robots are are are challenging our assumptions
um concerning what art is, who can make it and
(02:11):
uh and what makes it good? And actually what creativity
is at the heart of the matter here um Studio
three sixty dot com has some really great articles a
suite of articles on creativity and artificial intelligence, and that
is in part what prompted us to start thinking about this.
Today and podcasting on it. Yeah, it's a great show.
If you've never checked out Studio, give it a shot. Um. So, yeah,
(02:35):
we're talking about creativity, um and robots. Now, we did
a whole episode on creativity where it comes from in
the human mind. So creativity in and of itself is
a rather complex idea. There are several different definitions for creativity,
which we explored in that that episode and some depth.
Some of these definitions are more easily appliable to robotic
(02:59):
card this than others. For instance, there's the cognitive view
of creativity. Uh, and this is in this one. A
computer could definitely do it because it's a process by
which the sensory input uh is transformed, reduced, elaborated, stored,
recovered and used. So it's just a it's just an
input output scenario here. So it sounds like memory retrieval
(03:20):
as well, right, Yeah, it's and uh and you can
see this kind of quote unquote creativity at work in
simple You know that you'll find programs online where you
input tax and we use a cut up system to
mesh it around and change it. Uh. You know, random
if you were to do like a random filter on
a on a photograph. I mean, it's a very simple
(03:41):
version of what we're talking about. But essentially, data is
input in, data is changed, and data is output it. Um.
Then there's like the psychoanalytic Freudian view, which argued that
creativity is an occurrence of the subconscious. This not so
easily appliable to machines. Well, we'll get into this in
a little bit, but there is a aspect of this
(04:04):
that is happening in artificial intelligence. But yeah, there's this
idea that art occurs creativity occurs sometimes from these different
states of consciousness, and particularly with unconsciousness. In fact, Sovageot
Dolly supposedly would sit with a spoon in his fingertips,
holding it over a tin cup or plate as he
(04:24):
drifted off to sleep, and then the spoon would fall
from his grasp and wake him up again. And presumably
he did this so that he could spend longer on
the fringes of consciousness unconsciousness and retrieve new ideas. Um. Yeah,
And there have been other artists that have should have
have done that with a silver spoon before, but in
different ways. For some reason, I thought you're gonna say
(04:46):
a chainsaw. No, no, yeah, But but Okay. Other quick
definitions of creativity. There's the behaviorist view, which says that
creativity is a combination of previously no knowledge joined together
sponsor enviously um and uh, which gives the impression of
a new idea or inspiration. So this is kind of
an idea collider idea of creativity and uh. And that
(05:11):
matches up really easily with our ideas of machine intelligence,
that a machine without any kind of real gusto could
get in there and say, okay, um um, zombies are good,
and pride and prejudice is good. One plus one equals too.
Now I'm using creativity. I mean, that's a very granted
(05:35):
that idea was was was generated by a human, but
it's very easily too easy to imagine a machine coming
up with that as well, like looking at, say, you know,
if you had a computer that was analyzing like search
terminology online online, seeing what people were searching for and
just picking the top two categories and crafting a novel.
So it's very much like decision tree decision making. And
(05:57):
so the question arises, can there be a purely creative
product that comes out of a machine? And the other
question is, you know, is a sacred ground. Is it
really unique to humans? And that's what we're going to
really get to explore today. Yeah, it's an interesting look
at just what computers are capable of now and uh
(06:20):
in an idea of where they're going. But also it
really forces us to reevaluate what's going on in creativity
and what's going on in art and and you know,
it just really forces you to re examine how we
view the whole scenario. So, um, we should probably start
in the written word robotic fiction. Yeah, and apparently this
(06:43):
has been around for a while, right for something called
Mark one Baby in a nine British computer scientists Christopher's
stra Key program the Mark one baby computer to generate
love poetry from a database of romantic verbs and nouns. Yeah,
back in the fifties, Back in the fifties. And here's
here's a sample. I'm not gonna read in robotic Okay,
(07:06):
I'll try. I mean, you know how well I do
with other accents. Well, the robot accent is pretty pretty easy,
all right. Honey, love, my anxious fondness impatiently yearns for
your craving enthusiasm. You are my breathless infatuation, my darling, affection,
my ardent, liking my covetous fancy yours anxiously and you
(07:32):
see there you go, that was good. That was okay.
I mean, I'm sure the computer is listening. We'll probably
right now. Yeah, but yeah, here's here's an early example
of this. But you know, fast forward to do today
and we have something called Brutus, which is a machine fabulous. Yeah.
(07:53):
Brutus uses mathematical equations um that are based on the
basics of plot, of setting, of dialogue and uh and
really this if any of any of you that have
ever taken classes on on the structure of a novel
or how to write a short story, you know that
their their formulas out there, and you can really and
(08:13):
you can also look at it like that there's so
many different different guys out there and and and women
that are they have like these models of how to
plan out a novel, how to plan out a short story,
and they're you know, they're like triangle approaches and and
uh you know, multiple story arc approaches, and there's like
a snowflake um approach to writing a novel and it's uh,
(08:37):
so we're really big on formulating it and really guiding
it out you know, as if to make it to
where just anybody could follow these instructions and come up
with a novel. So so it it follows it's really
easy to imagine then a computer taking some of this
this data and using that that that information is an
equation to generate a novel. It knows the the structure
(08:58):
of plot, the the way a story tends to to roll.
Uh and uh, and it can can generate that. Yeah
and uh, I mean it is really cool actually, the
fact that it can generate these these stories. Now there
are limitations. I mean, we're talking about five words stories.
And programmer Selma brings your He said that the program
can't be considered creative unless the machine creates something that's
(09:21):
completely demystifies the programmer. So in other words, if it's
not demystifying um, and it's putting out the expected outcome,
then it's just cloning our intelligence. We'll see. This reminds
me a lot of the scenario of fan fiction um
in uh with with human writers. This is a pretty
(09:42):
simple concept that everyone's encountered. Take takes. Say, someone like
HP Lovecraft wrote a you know, a ton of short
stories and fragments and the occasional slightly longer work, uh,
with a very particular brand of horror. And you have
you have people that read and I mean think I've
probably read most of his his work, but you have
people that read most of his stories, not all of
(10:04):
his stories, And then it's pretty easy at that point
to sort of regurgitate something that is at least love
craft ish in that style. Yeah, you figure out what
he tends to do in a story, what format he
tends to follow, and then you can make something that's
love craft ish. And and certainly there have been a
lot of authors who have have enjoyed some success with
(10:25):
lovecraft Is stuff. But but where you really see something
phenomenal is when someone is clearly inspired by someone like
Lovecraft or or or um Or or various other authors
that have that certain feel and look of their their
own material, where they can take what that author has
done and then spin it around, do something different with
it in a way that makes it their own. Well see,
(10:48):
and this comes the originality really comes into question in
these sort of cases, because you could argue that there
is nothing nothing unique has ever really been created, right,
that it's just intertant. We're all taking ideas colliding them
and and trying to put just enough spin on them
that we don't get sued. It's just iteration of iteration
(11:08):
of iteration, unless you're DaVinci, right, yeah, just coming up
with all these completely original ideas for new machines. But
that's because he stole a time travel machine. I knew it.
I knew you. You finally said it. It's true. Um,
so yeah, I mean this idea that that this program
is spitting out stuff that's not too original. Well, okay,
(11:30):
I mean you really get into shades of gray of like, well,
what at what point does it become in the realm
of uniqueeness? And and um, just you know, just regurgitating
a bunch of rules. Yeah, and are you which model
of human creativity you're holding it up to. You're holding
up in the Sistine Chapel, You're holding it up to
Dante's Inferno. You're holding it up to pride and prejudice
with zombies, which is the main problem of subjectivity right
(11:53):
when it comes to art. Um. Okay, So I did
want to say that that Brutus created this this short
little or here, and he paints a picture of a
guy named Dave Sdriver. He's a PhD candidate and he's
ambivalent about his current situation and the program pus dashes
of details in there, just as you said, like, there's
(12:13):
very formulaic in some ways. Um, you know, just to
create a setting. Brutus says, uh, here's the Ivory clock
on the campus and eager students. You start to get
a real sense of place. And the pacing also gives
it a sense of drama and foreboding. And I wanted
to play just a really short clip of this. This
is from Studio three sixty and this is a moment
(12:34):
in which Dave Striver is being waylaid by the dissertation
committee for his PhD. Fiery and uncontrollable rose up Dave.
Professor Hart products softly, God, they were pitiful, pitiful, palette
and puny, Dave. Did you hear the question? Later Striver
(12:56):
sat alone in his apartment. What in God's name had
he done? And it should be noted that that was
not the robots voice. That was that was human reading. Yeah,
that was a real dude reading. But I mean you
kind of do get a sense of that, right, That's
not too bad? Yeah, I mean that's not I mean
for a computer to sort of split that out, and
for for to make sense in the first place is
(13:16):
kind of amazing, But for it to give you a
sense of of peering into another person's mind, that is amazing,
because does a machine know what a human mind is? Well?
And certainly learning about it? Because ultimately mean the machine
is dealing with language is kind of like a shadow
of the human mind. It's um you know, it's dealing
(13:37):
with coded data, semiotic web of of meanings and symbols.
And if you align it properly, you create that you
can create a text that has either a little or
a lot of meaning. Well, it's just to what extent
can the machine understand the human mind? I think? And
to that end, it turns out that Patrick Winston, he's
(13:58):
a principal, principle and stigator at m i t S
Computer Science and Artificial Intelligence Lab. He is teaching Macbeth
to computers and he says that storytelling makes it possible
to teach, to learn, and to be creative, and so
he wants these machines to understand humans. And what better
way than to start off with Shakespeare. Yeah, and as
(14:20):
far as teaching, learning and creating GOUM I mean human
beings go pretty long way with just faking any or
all three of those categories. So really, I mean, as
far as understanding the human mind, yeah, I don't know,
that's really not even as important necessarily for a machine
as just the simblance of it faking it, because if you,
(14:40):
you know, create, if you can create something that we
end up investing thought into, then then that is the
insight into the human mind. Right, maybe I'm getting a
little no, Well, I think what you're saying is that
the simulation at some point becomes the thing. Right, So
even if it doesn't truly understand it, if it's executing
(15:03):
the code for it. I mean, because we've all encountered art,
be it um you know, be it, be it a
work of written art or visual art or music where
we take that moment we think if this person brilliant
or are they just pulling one over on me? You know. Well,
now machines have the same opportunity exactly. Uh well, you know, okay,
(15:24):
this this guy who's teaching them Macbeth, he actually has
as a point here beyond just human um mind uh comprehension. Here,
it's is actually to try to help creatively problem solve,
which would make sense of storytelling. And he says that
what we're hoping to do is build systems one day
that are as important to a political analyst as a
(15:44):
spreadsheet is to a financial analyst. We have a system
now that draws parallels between various kinds of historical events,
the Arab Israeli War and the tet offensive, for example.
Finding those parallels could help keep people out of trouble. Okay,
so now we're getting into a larger area than of
robotic design, not just not just robotic writing, but just
(16:07):
robotic creative thought. Well, and it reminded me of the
Living Earth simulator that we've talked about before, which is
this idea that you could take every bit of data
in the world, throw it in there and try to
predict what's going to happen next, whether it's a weather
pattern pattern or um uh, you know, politics, um or
(16:28):
even economic ramifications throughout the world. It's interesting. I'm currently
reading the Hyperion series by Dan Simmons, and in that
novel you have a you have artificial intelligences that have
sort of succeeded, succeeded from from human civilization and that
(16:50):
still advise on various levels and are still involved but
also rather separate. But they have it to the point
where they know exactly how, um, everything is going to
transpire for years and years in the future, with the
exception of one sort of X factor that's pivotal to
the novel. But but but yeah, that back to the
(17:11):
Living Earth simulator, just the idea that that if you
put it up, you put the information in there, to
what extent can a machine navigate all the variables and
come up with an accurate prediction of how things will
play out. Well, we've talked about the limitations here with
a Living Earth simulator simulator because it doesn't take into
account or this version of it as of yet hasn't
(17:34):
been able to take into account what we call these
black swans or these events that seem to come out
of nowhere. But when in fact you look at it
like you know that there's going to be some sort
of entropy at some point, it just feels like it
came out of left field because we didn't expect it. Right. Well,
I mean, it's like human creativity. You know, Um, we'll
have we'll have one thing in mind, then something else
(17:55):
comes along and it changes the form of what you're
trying to create you know, um, I mean, like I
remember I started writing a short story before my dad died,
and then after my dad died, that changed the shape
of that story. Like that was a black swan event.
I mean, not that you know, people die, that's kind
of the It kind of happens all the time. So
not in the sense that it's that it's that unforeseeable
(18:17):
an event, um, but but it occurs and it changes
the direction of the creativity. So so what you're saying
here is that we have to have a creative program
that is also um that we throw a few black
swans at every now and then, Well that could also
create its own black swans. And we should probably take
a break here and we get back, though. We're going
(18:38):
to talk about whether or not there is a machine
that can break this kind of code and possibly even
create new unique thoughts. All right, all right, and we're
back and uh yeah, we're going to talk about a
little something here called the creativity machine. Yeah, okay, so
(19:01):
this is not a you know, something that we read
about in a science fiction novel or something something like
the near not the near Earth simulat not not like
the the Earth Simulator that is that is that may
exist in the future that we're planning. This is something
that he's out there. I actually thought about the Santa
Claus machine when I was reading about this, and we
talked about this in another episode. That of course is
(19:22):
something that that could exist in the future, but doesn't
exist exist now, Okay, so let's get to this creativity machine.
It is pretty much like the ultimate problem solving machine.
There is a guy named Stephen Fuller and there's a
great documentary called In Its Image that covers this. Since
the seventies, he has been fussing around with artificial neural
(19:42):
networks and he has conducted a lot of experiments with us.
Usually people have been dealing with artificial neural networks just
to kind of see the way that the brain processes information.
But he he has a different plan um with his
creativity Shine by using the human brain and trying to
(20:06):
not only replicate the way that neurons the neural network
is working, but also how originality arises in humans. He's
trying to do the same thing in in computers, and
you could argue that he has done it. Yeah, the
created creativity machine learns on its own and uh, and
it and it deals with black swans mathematical variables that
(20:29):
are thrown in to shake it up. Yeah, I mean, okay,
it's got these human generated scripts. UM. It is an
artificial neural network, but it has two distinct functions. One
part contains data and one part runs the data with
tests and gives it feedback. So that's where it learns
from itself. UM. And then occasionally Solar will actually disrupt
these functions by running mathematical noise into the system. That's
(20:53):
the black swan, which creates a false memory of confabulation.
And this is where the machine gets confused and it
begins to drive the artificial neural network to think in
new ways and produce new ideas. And here's an example
of this. UM Follower was actually commissioned in by Oral B,
the toothbrush maker, to create some new designs for the toothbrush,
(21:17):
or their toothbrush as I should say. And he said
that when they injected noise into the first network, it
imagined not the memories of toothbrushes it had previously seen.
It produced confabulations, false memories that were hybrid hybridizations of
what it had seen before. So when he when he
(21:38):
threw that box swan in there all of a sudden,
it didn't. It did create these original new thoughts that
were based on those memories, so it retrieved its memories,
but they were false memories, which is really sort of
the same thing that happens when we have our own
aha moments and creativity. Right, we have these breakthroughs based
on the experience the memories, but you know, come comes
(22:00):
out and a bit differently. Um, he says, you know,
I didn't do anything crazy here. I basically pushed a
return button and sat back and watched two thousand designs emerge,
and then he gave the customer or it will be
two designs, okay, So that sounds like okay, or it'll
be toothbrush whatever. This guy, as I has actually used
(22:22):
this machine to create everything from like snack food to
military devices like robotic swarms. Yeah, he's that. It's been
used to create such as harder than diamonds, one point
five million new English words, it's trained robot cockroaches, and
it's um, it's composed just a whole lot of music, yeah,
(22:43):
which is really cool, and we'll listen to a clip
of that music in a moment. Um. But this is
a really cool thing to uh. Fohler actually filed the
first patent for the autonomous generator generation of useful Information,
and then his second patent was for his self training
neural network object. Patent too was invented by patent number one.
That's crazy, that's amazing the machine created the patent. I
(23:06):
mean this is this really does kind of spell out
a very interesting future with this technology. Um. He says
that this is the way that we're going to cure
cancer and the way that we'll do multi trilling and
dimensional optimizations to bring some peace of mind for the
world's economy. Okay, so think back to the Living Earth
simulator and it's limitations. If you were to marry the
(23:28):
two technologies, is there the possibility that this machine could
become like, you know, basically a crystal ball for us?
It could be like oral b comes to the the
the AI s then and says, hey, we want to
to roll out a new toothbrush, but we also want
to know how this toothbrush will change the world. So
the computers begin to say, all right, here are ten
designs you can choose from, and here is a layout
(23:51):
of how each of these designs will affect your financial
well being as well as the world at large. Right,
could you could overlay all that data, right, you could say,
if you decide to produce it in this area of
the world, this is the impact it will have on
that economy and on on those people. You know, if
you did it here, it would have this sort of impact,
so on and so forth. You could look at it's uh,
(24:12):
it's carbon footprint. I mean, you could look at all
sorts of slices of data on top and ultimately that's
what the the Living Art Simulator it comes down to
as well. It's it's not really so much an idea
that will we will create a virtual reality Earth too.
It's it's the idea of, well, we have in this
simulation as much data as we want to throw at it.
(24:33):
So you would say, all right, I'm interested in making toothbrushes,
so I want to know how how this device might
affect um you know, this particular hygiene market, how it
will play out you know, legally, how it'll play out
with profits, how it will play out with you know,
with the with the perception of the brand. You know,
whatever the variables happen to be, as long as you
have the data to back it up. Yeah, of course,
(24:54):
you know you would have to mesh these two technologies
and that that right there is the fiction part of this.
But those two technologies do exist stand alone, and so
it is it's very exciting to sort of extrapolate what
that might look like in five years, ten years, twenty years. Um.
I think that this this is the great opportunity to
(25:15):
hear a little clip of music by this machine. It's
actually called the Dingularity and uh, this is from the
album Song of New Neurons. And again this is this
is from Studio three sixty. All right, but we're talking
(25:42):
about are here, so let's let's really get down to
the to the paintings, now, to the to the to
the canvas itself. UM, this is nothing new either. There
any number of robots out there that have been designed
to put a little paint on a canvas. Um. I
mean you can and on a very basic level, we
have robots out there painting cars every day. The works
(26:05):
pretty basic, it tends to me it's it's it's very
often one color, but it is a robot painting something.
So it's very easy from there to extrapolate, Um, robots
painting various other things. Actually we've been able to We've
trained elephants to paint. We've you know, you you get
an ape uh hands dirty, it will paint something on
(26:27):
a canvas. Um. So so yeah, we have machines that
are capable of this as well. UM. I mentioned a
few of them here that are um that are fairly interesting.
For instance, there is uh the van Go bot, as
in Vincent van Go and this, uh, this is a
(26:48):
particular bot boast eight teen brushes, a paint mixer, and
three D spatial awareness, so it can combine artistic influences
to create fresh takes on a given subject. So the
very ables here being a subject and what styles you
want to see. So you can sort of think of
this in terms of almost like like an Instagram kind
of a thing. Imagine if if a robot were in charge.
(27:11):
All right, the photo is of a dog, and say
I wanted to combine filter one and two, and it
would combine those in a way that made sense. So
you might say, all right, I have I have a
picture of a bowl of fruit. I would like that
in a way that's cubist and uh and also surrealist. Well,
and then there is what seems to be a more
(27:34):
nuanced program called Erin, Right, and this is where, um,
you get more of the expression. I guess you could say,
even in the eyes of the people that it's depicting
in its paintings. Yeah. Erin has been going since about
nineteen seventy three, or the project itself, and uh, it's
the ranch out of Harold Cohen. Uh, Aaron drawls and
(27:57):
paint stylized still lives and portraits of human years out
of its programmed quote unquote imagination. Um, no, no images
or additional human input is necessary. So it sees something
and then it creates an artistic interpretation of what it sees,
which would point to original thought. Right, it's kind of interesting, yeah, okay,
(28:20):
and then the thing again, it's like once it's created,
like it is, it creates this thing on cannabis. Then
that then interacts with the human mind. And we've talked
before about all the stuff that goes on in the
human mind when you look at a piece of art.
So you know, even if there's if it's not thinking
about subtext or whatever, it can't it's not actually quote
unquote thinking. Um, if the piece ends up summoning that
(28:44):
thought in the viewer, then I mean it's art. Yeah,
I don't know. I mean I've seen some of the examples,
and you can go to Studio three sixt dot com
and see the samples of them. Um, that's a good
place to start at least. But some of them really
do elicit some feelings, I have to say, but there
(29:04):
is a flatness to it that you can kind of
tell that it's computer generated. Um. But that being said,
there's uh, there's that I can't remember the first name
of the artist, but it's Fred's grandson or something. Um.
There were some scenes of people that reminded me of
Fred's paintings that have a certain mood to them, a
(29:24):
certain quality, uh, in the way that it's spatially arranged.
And I thought that was interesting because I thought, well,
there seems to be some sort of psychology going on
in this painting. Yeah. But but then again that that
reminds me of of any number of cases of you'll
see like an interview in with an artist or an
actor or musician and they'll they'll be like, oh, you
know when you were when you did this particular thing,
(29:44):
when you decided to paint this blue instead of gray,
or you know, what was going on in your mind,
and they'll have all of this, this thought invested in this,
in this one little detail, and then when the artist
just says, oh, yeah, well we ran out of gray
paint that day or lit or it will be something like,
oh yeah, well it's different live than it wasn't an
album because we kind of forgot how the music how
(30:07):
we actually played at the first time. You know, there
ends up being some sort of mechanical explanation for something
that you thought was going to be really just tied
into the creative process. Well so, so much of that
then is projection by the viewer, right, Okay, So, just
in case you were worried that there there wasn't some
sort of computer around here to try to take the
(30:27):
highly subjective um job of making, uh this the stance
that some art is good and someone is bad. Don't worry.
There's something called DARCY. It's the Digital Artist Communicating Intent Machine,
and it's essentially kind of like an art historian. They
are teaching this computer to basically build a DA database
(30:50):
of artwork in order to carry out this subjective task
of saying this is good and this is bad, and
basically dissociates images with an adjective, and it can learn.
It can build up this database and begin to learn
what is considered good and what is considered bad. And
then you can actually throw another image in front of
it and it will rate it, which is interesting. Computer
(31:12):
scientists from Brigham Young University actually have been working on this,
and we missed it when it came to Atlanta. It
came to the High Museum of Art and judge the
works of Man it did. It dashed a lot of
people's hopes to be artists. Apparently a couple of kids too,
got some some thumbs down from Darcy. Um. All right,
(31:33):
so let's talk about film. Could could a computer be
an all tuo of sorts in this medium? Well, certainly
we're now we're dealing with well, of course we're dealing
with this with fiction, but um, with film, with digital
film editing, it's it's all digital. All that information is
just data in the computer. And and again this is
(31:53):
another area where there are certain rules, there are certain
types of cuts, there are certain uh there's a certain
uh law that you abide by when you're editing a film.
So it's perfectly believable that you could hand over the
raw footage and the basic flow of the piece. And
and you could expect a machine to to give you
some results, whether it would be the results on par
(32:16):
with the work of a great film editor. Well that's
the discussion. Well, yeah, that's that's the question mark here. Um.
The film is called White on White Algorithmic Noar, and
it's by Eves Suessman, and it's kind of build as
a paranoid sci fi thriller. And uh, Eve Sussman says,
it's about an American engineer Holts. He gets a job
(32:38):
in a mysterious form place called City A and starts
finding out things are wrong. The water supply seems to
be drugged with lithium, time seems to be slowing down,
people are running out of language, and the people who
hired him may not be who he thinks they are.
So everything is a little bit beyond his control in
this this situation very calas kind of scenario. Yes, yes,
(32:59):
it's very tough us and uh, it actually is very
interesting to watch it. And here's another reason why it's interesting,
because every time you watch it, it is a different
scenario because of this computer program which is essentially re
editing the film. Um, every time you view it, she
Eve has been worked with programmer Jeff Garnot and he
designed the computer system, so there's no one version of
(33:21):
white on White. This is interesting because this ties into
some of the the ideas we've discussed about the future
of music, like the idea that in the future, um,
we won't so much as listen, we won't listen to
a piece of music that it was composed by an
author or an artist. Rather, but we would listen to
a piece of music composed by an artist that then
taylor's itself to our particular taste and needs or an environment,
(33:45):
which is just like the the robotic piece of music.
It's called Emily Howell ye or that's the program. David
Cope designed it, and it composes classical music based on
influences from Bach and lose Art, And some people say
that you can't tell the difference at times. So that's
interesting that you say that, because if you were, if
(34:06):
you knew that you were interested in that and then
that's what it was inspired from, then it would no
longer be actually the works of Bach and Mozart. I
think of some of the films that have like they're
famous for multiple cuts. Uh, the one that comes to mind.
Most obviously and most fitting for this conversation is Blade Runner,
where some cuts lean more towards h Harrison Ford's character
(34:27):
being a replicant and Natroid an artificial person versus him
being human. So what have you had in the future?
What if a cut of Blade Runner Taylor's itself to
your particular values. If you're the kind of person who
is more satisfied with with with Harrison Ford being a robot,
then that is the version you end up seeing. Likewise,
(34:47):
if you're if you're the kind of person who thinks
Han Solo should should shoot first or you should shoot second,
then then a copy of Star Wars Taylor's itself to
what kind of a viewer you are? Well? Then, oh,
and then to say nothing of of curse words, nudity, violence.
You could have the film automatically adjust itself to the
(35:11):
age of the viewer or even the sex of the
viewer versus Maleston. If you they're sort of demographical variables
in play, it's no longer tied though, to that particular
person's perspective, and is no longer the artwork of that person.
Correct Well, you would have to have some very we
like for instance, the language thing is pretty easy. Like
(35:31):
there's there are artists out there where be like, oh no,
you can't, you can't have scarface say um, dang it,
you know you you this this character, this character's cursing
is is important to the work, and if you take
the cursing out there, it's no longer that work. But
there are other my little car garages, Yeah, you can
really leave the cockroaches. Yeah, but but I feel like
(35:54):
there are other films out there where the artists would
probably say, no, I don't think it harms my artistic
vision if they say fudge instead of the F word. Oh,
I don't know. Budge is a very different thing than
the other thing. Well maybe, but you get my point.
I think there depends on the level to which the
work adjust itself. Okay, well, and I suppose it depends
(36:15):
on the author of the work too, and how she
or he intends it to to interact with it. Take Brazil,
for instance, Terry Gillis film, there was a there was
a like a happy ending cut of the film that
the studio did, which is horrible. Um, but it has
a very happy ending, whereas the actual cut of the
(36:35):
film as a really down ending, like even the sort
of even the compromise, the cut that they ended up
putting together is really kind of a downer. If you
had a version of Brazil tailored itself to your viewing
and would sometimes choose the happy ending, that that would
definitely compromise the piece. Again, I think it all goes
(36:57):
back to what the what the person is trying to
say with particular piece of artwork, which goes back to robots.
Do they have a particular perspective when it comes to that,
you know, can can they really have a perspective? You
feed them lots of data about the environment, if they
can learn from themselves, uh and create original thought, there's
(37:17):
the possibility there, right, um, But let me let me
before we get too far into that. UM, I do
want to play this clip of Emily Howell. This is
the music that we're talking about that is inspired by
Bach and Mozart, just to give you guys a sense
of what that sounds like. Now. I'm not the I'm
(37:47):
not the biggest kind of sewer of classical music, but
but I certainly I don't think i'd be able to
tell you if that was Bach, Mozart or a computer. No,
And I think that The issue that a lot of
who are having with this this iteration of music at least,
is that there's not a standout piece yet. And and
(38:08):
even David Cope, who designed it, says like there's there's
not anything too terribly creative or original about this music.
It's just a variation on a theme that someone else
has done before. So eventually though, there presumably there is
going to be some sort of breakthrough, some sort of
piece of music created by a machine, possibly even the
creativity machine, that people say, oh wow, this this is
(38:29):
really struck a chord in me as a human. It
reached this the center of emotion in me that I
didn't think that a robot could inhabit my perspective in
order to elicit that feeling. So then the question becomes, well,
who is the artist here? Then? Is that the programmer
or the computer? And at what point do we consider
(38:50):
human creat creativity be the same as machine creativity. And
here's the other question I have, because we've talked about
human cyborgs in the way that we are always uh
manipulating ourselves with technology, is this just another way to
manipulate our abilities? So is this part of the human
experience to some degree. Yeah, Like when I when a
(39:11):
human creates a robot that then creates art, is it
really is that human just creating via proxy? Is it? Is?
It kind of like saying, like, like looking at the
difference between me singing and me creating music with a trumpet. Um,
you know, I'm with the trumpet. I'm using a piece
of technology that amplifies it amplifies and changes my output
(39:32):
into a different type of output, but it's still dependent
upon my original output. Yeah, I know. It's like it's
it's an extension of creativity. Right. Just because you use
something to to express yourself, albeit you know, a machine,
it doesn't mean that it's not somehow an expression of
the human self because we're the ones who coded it. Um.
(39:52):
So I think it boils down to this idea of consciousness.
I really do. And I think that the creativity machine
and its ability to mimic these little blips and consciousness
and I'm not gonna call them unconsciousness because that's probably
giving a little bit too much credit. But I think
until we can really crack this nut of what consciousness is,
(40:13):
we can't really answer the question about whether or not
robots uh could emulate it. Or or create truly unique art. Yeah,
because I mean when you get into the motivations, like
something like Dante's Inferno, you know, the the artist wants
to create this piece, and he wants to create a
piece that that that says something about about God, that
(40:39):
says something about faith, that says something about the world
that he lives in, that takes jabs at people that
he doesn't like. You know, I mean all these all
these things that go on and in so many works
of art, you know, where it's there's a there's a
little bit of themselves in the work as well. Um
to what extent can a computer ever pull that off? Well,
(41:00):
So some would argue that man humans have been creating
art for at least thirty five thousand years, right, like
from cave drawings onto up to like creating a program
like Aaron that can create art, and that machines have
only been creating art for about forty years, so that
you know, we just need to give him a chance. Yeah,
(41:23):
because yeah, when we were forty years in we probably
we weren't we well, we weren't really all that impressive either, right, So,
and and you know, because they can carry things out
at much more rapid speeds than we do. Presumably they
could become pretty sophisticated artists. I don't know or you know,
is this predicated on this idea of a soul and God?
This is for another podcast, but is there? I mean
(41:44):
for the next podcast, because we're gonna that's true, we
are going to Before we do that, we have to
close this one out. So indeed, so there you have it. Um,
some thoughts on the computer as an artist, the art
as a creation of machine versus a creation by and
and then as they work that is ultimately viewed by
human or in the case of Darcy, by a computer. So, UM,
(42:07):
we would love to hear from from you guys to
you know, what what do you think about that? What
do you what? What is art? Do you think a
computer can really create? Uh? Music, literature, UM or visual
art in a meaningful way? Uh? And then how much
of it just does does it depend on the view
or the reader or the listener as to whether it's
art or not? Let us know. We'd love to hear
(42:29):
what you think. UM. I have just uh already one
quick listener mail if we want to call the robot
over here. It's always a little awkward calling the robot
over here after we've been talking about robots. Yeah, and
all right, now this one is related to movies. Uh.
Our listener Sean writes then about horror and talking about
(42:50):
old versus new horror. Uh, Sean says, really quick. My
observation on new horror versus old horror in the mental
versus the visual. The old schoolers like Hitchcock and Uh
and King left what happened next up to the mind.
Now with modern technology they show you the gore and
leave nothing to the imagination. I know there are except exceptions,
(43:11):
but I feel new horror is no more scary than
it is disturbing. I also find myself more frequently feeling
that new horror glorifies individuals with psych psychotic tendencies, especially
when the character is based on a real individual. Great
job every week, Matt Love Sean, Um, Yeah, I mean, uh,
I've certainly seen that argument that uh that that you know,
(43:33):
now horror is more about just really throwing up disturbing images.
But but but even then, it's like I feel like
when when when horror movies show disturbing images, it's like
that you disynthesize to that so quickly that it's there
has to be if it even if a disturbing image
is going to be effective, it needs to have an
(43:54):
underlying disturbing idea. And and sometimes I think that the
disturbing image actually in the way of the of the
the idea. And to tie this back into what we're
talking about today, I sometimes feel like when you'll you'll
look at horror that it seems to work or almost works,
and it'll you'll because you think that there is a
(44:16):
disturbing idea at the heart of it. And in some
cases it just seems to turn out that the the
the creators of this particular piece are just they're just
dealing in the visuals. They're just dealing in the disturbing
images side of it, and any um in any disturbing
idea beneath it is just the viewers perception. So you
(44:37):
get into ideas of you know, something can actually seem
to have more ideas and more more depth to it
based on what the view or the listener, the reader
is bringing to this. So again projection. So so yeah,
let us know what you think about computers creating art.
You can find us online. You can find us on
Facebook where we are stuff to blow your mind and
(44:58):
you can find us on Twitter where handle is below
the Mind, and you can also drop us a line
at blow the Mind at discovery dot com
Stuff To Blow Your Mind News
Hosts And Creators
Robert Lamb
Joe McCormick
Popular Podcasts
United States of Kennedy
United States of Kennedy is a podcast about our cultural fascination with the Kennedy dynasty. Every week, hosts Lyra Smith and George Civeris go into one aspect of the Kennedy story.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com