February 7, 2024 • 65 mins
Can a machine harbor the soul of an artist? This question leads our latest discussion with Harvard professor and acclaimed artist Matt Saunders.
It's not a simple question to answer. Creativity is a complex concept and we have to dig into the many ways that word could be defined. Who better to ask about creativity than someone who is both an accomplished artist himself and an instructor at one of the most distinguished art schools in the country?
Matt is a Professor of Art, Film, and Visual Studies at Harvard University. His art has been featured in exhibitions around the world, and the list of museums with his work in their collections includes the Museum of Modern Art, the Guggenheim, the Whitney Museum of American Art, and many other names that you would likely recognize.
Matt's art is grounded in the material - particularly painting, but in most of his work you can see how he challenges himself to push back against the idea of a static rectangle of paint-on-canvas and create something more dynamic and transient - something more like an experience than an image.
That same resistance to conventionality comes out in our conversation about AI, and it leads us to some brutally honest conclusions about how we view the techno-obsessed world around us.
Matt is a very thoughtful and inspiring person, and this interview is easily one of the more impactful conversations I've had in a long time.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Matt Saunders (00:04):
I share your anxiety, but not maybe your
pessimism.
Sam Gerdt (00:13):
That's a pretty common response.
Welcome everybody to Road WorkAhead, a podcast that explores
the unmapped future of businessand technology.
My name is Sam Gerdt andtoday's interview is oh so very
good.
Is AI creative?
It's not a simple question toanswer.
(00:35):
Creativity is a complex conceptand we have to dig into the many
ways that word could be defined.
Who better to ask aboutcreativity than someone who's
both an accomplished artisthimself and an instructor at one
of the most distinguished artschools in the country?
Matt Saunders is a professor ofart, film and visual studies at
Harvard University.
(00:56):
His art has been featured inexhibitions around the world,
and the list of museums with hiswork in their collections
includes the Museum of ModernArt, the Guggenheim, the Whitney
Museum of American Art and manyother names that you would
likely recognize.
Matt's art is grounded in thematerial, particularly painting,
but in most of his work you cansee how he challenges himself
(01:20):
to push back against the idea ofa static rectangle of paint on
canvas and create something moredynamic and transient,
something more like anexperience than an image.
That same resistance toconventionality comes out in our
conversation about AI, and itleads us to some brutally honest
conclusions about how we viewthe techno obsessed world around
(01:43):
us.
Matt is a very thoughtful andinspiring person, and this
interview is easily one of themore impactful conversations
I've had in a long time.
Matt, thank you for joining me.
I really only have one questionand I'm sure we'll get into it
after that, but my question isartificial intelligence is it
(02:05):
creative?
Matt Saunders (02:07):
Well, I don't even know how to answer that
question until I think I need tounderstand what you mean by
creative.
I would say yes, but I feellike maybe you have this
availance on creativity that isvery perhaps ennobled in the way
that people think about it inrelationship to art, which is
maybe not how I'm answering.
Yes.
Sam Gerdt (02:29):
That's probably true.
Let's go back to the beginningand let's ask a more basic
question what is your definitionof creativity?
Matt Saunders (02:38):
To be honest, it's not a word I think about or
use very much, but I think thatI would define it as putting
things together in ways thatmake leaps, whether that's a
leap out of a rational path, orwhether it's a leap out of a
(03:01):
series of expectations, whetherit's an inversion of the status
quo.
The difference between an actand a creative act for me might
be one that just hops over somekind of fence.
Sam Gerdt (03:14):
Yeah, you said you don't use the word creativity or
you don't think about it awhole lot as an artist.
Is there a reason for that, oris there another idea that comes
into play far more often thanthe idea of creativity?
Matt Saunders (03:29):
I think I'm a little allergic to the kind of
mythologies that we build aroundartists as a different type of
creature, a different type ofactivity.
I feel like there's acontinuity between making
breakfast and making what turnsout to be a milestone painting
in someone's life.
There's an activity of focus,of thinking, of reacting to past
(03:54):
experience, of working withmaterials.
That is part of creating.
Sometimes I think thosecreations are surprising.
Sometimes that's surpriselinger.
Sometimes it's just somethingthat surprises you at first and
you get over.
Often things come back to staywith you or to challenge you or
(04:18):
to make you think differently inthe future.
In my experience and maybe thisis just the type of artist I am
I don't have a great amount offaith or predictive capacity in
the act of making.
I don't always know when I'mbeing creative.
I feel very strongly thatthere's a difference between the
(04:43):
act of making something, thesort of generative act of
putting things together, andthen the moment of interpreting
it or seeing it or understandingit, which is always based in
time and place and changes overyour life.
When I say that I think AI iscreative, yeah, it makes things,
it's creative, it's creatingthings.
Now, whether that is a virtue,whether creativity with a
(05:05):
capital C is some specialcategory that gives you special
status.
I don't know that.
I even would give that as ablanket to myself or to other
people.
I think it's something thatyou're touched by from time to
time.
Sam Gerdt (05:18):
When I ask the question, you're an artist,
you're an accomplished artist.
Why are you an artist?
Matt Saunders (05:26):
I like things, I like making things, I like
thinking about images.
I'm a visual person and not somuch of a musical person or even
a word-based person.
For me, it's a way to work withmy hands and my mind in a
different way than I'm calledupon to do in other types of
(05:47):
professions.
I think I get great pleasureout of that time thinking in
other ways.
I just love art.
There's always, I think,something I feel kinship with,
with my imagination of thesentience of an AI who's been
trained on data sets, is there'salways an aspect of emulation.
(06:09):
There's an aspect of beinginspired by things that you've
seen other people do, thingsyou've seen in the museum, and
that itchy impulse you get torush back to the studio
afterwards.
I was reflecting on that when Iwas trying to think about AI
before this conversation, that Iguess I do have my own data
sets.
I've been trained to.
Sam Gerdt (06:28):
I think for me, this is a question that I've been
wrestling with for a long time.
That is, what about the factthat we are essentially
computers with our own data sets?
Why am I so quick to say that Ican be creative, but that an AI
can't?
And I think what it comes downto is capacities, and whether or
(06:50):
not an AI can ever have thecapacity that a human can for
synthesis, and I came across aquote by a Dr Blumenthal who, I
think, sums it up very well.
He says there are nopsychological qualities.
In physics, for example, thereis no red, no green or blue.
In the physical world, redness,greenness and blueness are a
(07:14):
phenomena that are created bythe cortex of the experiencing
individual.
A musical quality, the flavorof the wine or the familiarity
of a face, is a rapid creativesynthesis that cannot, in
principle, be explained as amere sum of elemental physical
features, and so, consideringthat quote, I have to conclude
(07:36):
that a computer which can onlyever experience the elemental
physical features of the world,will never be capable of this
creative synthesis.
That's being described In myown thinking on this topic, and
this is something that I think,from the beginning, has rubbed
(07:57):
me the wrong way, in the waythat we are being conditioned to
talk about things likeartificial intelligence.
There were a lot of terms thatstarted being used a lot more so
in the general, you know,public conversation, we started
hearing more and more aboutcreativity.
We started hearing more andmore about generative AI and we
(08:25):
started being I feel, we startedbeing conditioned to believe
that artificial intelligence wasor would become something
threatening to human creativity,and I immediately didn't like
that.
I didn't think it was accurateand I didn't think that it was
(08:45):
healthy for us to even engage inthat kind of thinking, and so,
from the beginning, I've beentrying to figure out why that is
.
You know, I have this gutreaction to it.
But why do I have that gutreaction to it?
And I, you know, I started mycareer.
I went to art school, I studiedgraphic design, so I have a
(09:05):
little bit of a background inart not quite fine art, but I've
participated in the art worldfor a significant period of my
life and what I see is a bigdifference between something
that is creative and somethingthat is generative.
I feel like creative is adescriptor of content that, like
(09:27):
what you're saying, there's athere's a leap there, it
represents a progress that thatwas maybe unexpected, maybe much
needed, but undiscovered.
But there's something new in it, there's something novel in it,
and I also feel like creativityhas to be rare.
(09:52):
I feel like there has to be ararity to it, because how can
you call something that's commoncreative, how can you call
something that lacks that?
It lacks that novelty, thatleap that I think is needed.
And with AI, I think what we'reseeing that's difficult is
(10:12):
we're seeing a blurring of thelines between the artist and the
tool, and we're also seeingthis blurring of the lines
because of how fast things aregoing.
We're seeing AI generatedcontent that is truly beautiful
and technical and it has all ofthe characteristics that we're
(10:33):
accustomed to seeing, andsomething that really truly is
creative.
But the problem is that wehaven't yet caught up to the
fact that, because AI can do itnow, now it's common.
So when mid-journey first cameout and we saw the images from
this stable diffusion is anotherexample the images that it was
(10:57):
generating were striking.
They were beautiful and theywere very stylistic and they
looked like art, and you couldargue about the goodness or the
badness of that art, but theylooked very artistic.
I saw that and I would say itlooks creative but it's not.
And I would say, in a year'stime when 30 million people are
(11:20):
subscribed to this tool and theyeach are generating 10 images a
day, and you have 100 million,200 million of these same kinds
of images flooding into theinternet every day, I have a
feeling you're going to look atit and you're going to feel very
differently about it than youdo right now, seeing one or two
(11:41):
of them here and there, becausethat commonality is going to
make it feel uncreative to you.
Matt Saunders (11:50):
I think I'm with you with what you've said, but
there's a gap that opens up thatI was thinking about when the
creativity happens, and whethercreativity is a singular act.
We think about AI generatingsomething quickly and we ask was
that creative?
I feel very strongly that allart, all aesthetics, all value
is distributed value.
(12:13):
Art is a language and it onlyis good if it means something
and it has the capacity to last.
The magic of art is that thingscan last for generations and
work can be meaningless fordecades, centuries, and then
suddenly mean a lot to a lot ofpeople.
Maybe that's why I strugglewith the word creativity,
(12:34):
because it seems to arrive andjust be, and I actually don't
think that work's power existsin the world in that way.
It might be very powerful inits moment of creation, through
its novelty.
It's a word that you used a lot, that I kind of chafed at as
the ultimate value, but I have amuch longer view of things.
(12:55):
I think what really makes senseto me and what you're saying is
the idea of flooding the market.
Something I really like about AIand the AI images we're seeing
is that it puts pressure, Ithink, on our assumptions about
value in art, words like beauty,words like artistic.
I might define differently froma lot of people, and that's not
that I'm right or people arewrong.
(13:16):
We live in different contextsand we look at different things,
but I think that there's a kindof unreflected assumption of
some values, like the value ofwhat we think of as beautiful.
If the world is full of acertain type of soft focus,
light infused, I don't know I'mtrying to imagine what an AI
beautiful image would beluminosity then will that
(13:38):
continue to be beautiful?
I think you're asking thatquestion and I agree with that,
because I do think that one ofthe things that AI is doing now
for me is generating a lot ofimages that I think are pretty
drecky.
I think that they are a littlebit received ideas of what art
might look like.
(13:59):
I find most of them completelyuninteresting, but I also see
their appeal and I'm wrong tothink they're uninteresting
because so many people thinkthat they're interesting.
But I appreciate that they'reso easy and that it's not going
to replace artists, but artistsare going to react to how the
visual world has been changed bythis type of image and the
(14:20):
access and ability for people tocreate the AI to create all
these images and that thecreative responses will actually
be the rejections of it or thepeople who find ways to actively
step to the left of it andoffer some kind of alternative.
An alternative will only begood in the moment when the
(14:44):
baseline is the flood of AIimages, and then it will evolve.
Sam Gerdt (14:50):
There was the article that you responded to.
The question that you respondedto in an article several months
ago about this very question isAI art really art?
One of the comments that youmade in that article that I
appreciated was that artistswill be the one to help us
(15:10):
imagine the new world that AI ishelping us to create.
The article I'm referencinghere was published by the
Harvard Gazette in August of2023.
It's called Is Art Generated byArtificial Intelligence Real
Art?
Here's exactly what Matt saidin the article.
To the question of whether AIcan be a threat or a
collaborator, I might respondthat every new technology upends
(15:34):
conventions and delivers notonly new possibilities, but a
new kind of materialintelligence.
I'm sure that many artists willbe intrigued by the agency of
AI and seek ways to grapple orcollaborate with it.
Many already are, and we shouldbe grateful to be challenged
and knocked out of our habitsand assumptions.
Most of the things that worryme about this fall into the
(15:56):
realm of the social and ethical.
I hope there are great artiststo help us imagine around and
work with this new reality.
I think what you were saying isthat the masses are going to use
AI for a particular means andthey're going to move the world
in a particular direction.
(16:16):
And those who are on theoutskirts of that crowd, those
who are more imaginative, thosewho are more creative, are going
to help to pull that mass ofpeople in a particular direction
or help us to understand it innew ways.
(16:37):
Because in some sense, the vastmajority of people are just
following the crowd.
And there are those on theoutskirts who look around and
say you know, they're seeingfurther and they're looking at
places that others aren't, andthey're helping that crowd to
imagine, to see what they see.
I really appreciated that, butI think I appreciated it because
(17:00):
it reinforced my own feelings,and that is that it's those
people who stand on theoutskirts of the masses who are
looking elsewhere.
Those are the people who areexpressing themselves creatively
, teaching others creatively,informing all of our
imaginations and all of ourthinking, and those end up also,
(17:22):
I think, being the ones inother areas of life that are the
innovators, that are theproblem solvers who end up
taking us places we neverthought we would go.
If we do arrive at a placewhere artificial intelligence
truly isn't an entity among us,something of a peer, it will be
because of those people, thoseinnovators who did truly
(17:46):
creative work.
It will be the result of that.
Matt Saunders (17:49):
I mean, the categories we're talking in are
just too big to get that muchtraction, because I think you
know the role of an artist who Iwould admire.
The kind of artist that I'minterested in in my very narrow
set of interests does manythings.
One is making images or ideasor objects that participate in a
(18:12):
current conversation.
I'm not that taken with themyth of the undiscovered genius
and the attic that I think thatthere is a current conversation
in the art world and there's amillion art worlds and that you
are making things that are partof that.
Of course, the people who I'mreally excited about and the
(18:33):
people who, in retrospect, seemlike the most important in the
generations are the people whobring a particularly critical or
transformative or a differentperspective to things, and I
think that you know that tendsto have an aspect of criticism.
So the artists who are workingwith AI right now who I'm most
(18:54):
interested in are the ones whoare kind of pushing its limits,
people who aren't even known yet.
I have a former assistant who'strying to force the AI to break
its own rules, you know, who'sconstantly working with the
prompt and treating it as a kindof dialogue instead of focusing
on the image or what could becreated.
(19:16):
He's focusing on thechallenging creative activity of
interacting with another makerand pushing boundaries, kind of
getting it trying to force theAI to fight with him, and to me
that's a kind of artistic, acreative approach to using the
AI in line with conceptual artand kind of like generations of
(19:41):
artists who've thought aboutarts relationship as pop to more
popular culture, saying not toreject it but to find ways to
cut under it or to lay thingsbare, or to add friction, add
turbulence to what could be asmooth kind of marketplace.
Sam Gerdt (19:58):
Does that necessarily lead you to the conclusion that
AI itself could never be anartist?
Matt Saunders (20:05):
I'm not smart enough to know that, because I
assume that AI can internalizethose strategies.
The AI we're seeing now, Idon't think is being fed David
Hammons or other artists who'vereally worked against the grain,
and people are using it to makeimages.
Sam Gerdt (20:22):
Well, that's what I'm not entirely clear about is,
you know, from a technicalstandpoint, ai can't respond
with anything that it hasn'tbeen trained on.
That is, its responses arenecessarily a diffusion of the
data that it already has.
And if the data that it alreadyhas is limited to what has
(20:44):
already been made by people,then can AI add anything to the
conversation?
They've done experiments withthis.
Where the question is well, canAI generate its own training
data?
Because that right now is thecommodity that's in short supply
is training data.
So the question is well, if youcould take an LLM and from the
(21:07):
training data that it alreadyhas, if it could generate new
training data, maybe we couldtrain a smarter LLM.
And the end result is itdevolves into nothingness.
So when we use AI-generatedcontent as training data for
future AIs, we call this contentsynthetic data.
(21:31):
Synthetic data is not generatedby people, it is generated by AI
.
The usefulness of syntheticdata is going to vary from case
to case and model to model.
So, for example, it makes a lotof sense to use synthetic data
for training a self-driving AI,and we see a lot of support for
this kind of synthetic databeing used in this use case.
(21:53):
But when it comes to generativemodels like the ones we're
talking about here LLMs anddiffusion models that are being
used to create written andvisual content the use of
synthetic data to further trainthese models does have a
degenerative effect.
A July 2023 study entitledSelf-Consuming Generative Models
(22:13):
Go Mad went into great detailabout this phenomenon.
In the paper, the authors saidOur primary conclusion across
all scenarios is that withoutenough fresh, real data in each
generation of an autophagousloop, future generative models
are doomed to have their qualityor diversity progressively
(22:34):
decrease.
Now, this is just one example,but the point that I'm trying to
make is that the generativemodels we have today are not
capable of synthesizing data ina way that could be considered
additive.
So that's where I'm wonderingcan an AI be an artist if it's
incapable of contributingsomething novel to the
(22:55):
conversation, if it's incapableof developing the conversation?
Matt Saunders (22:59):
My understanding of the diffusion is that it's
partially about adding static,and I think anytime you add, you
make a space for thecounterproductive, the
irrational, the random, and thenyou hold that together with the
form that you were given andjudge whether it's working.
I think that's kind of what anartist does.
(23:20):
I feel like that is generativein a way that transcends its
dataset or could be.
Maybe it hasn't gotten thereyet.
I mean, of course I also amteaching, and when you're
describing AI not being able toinvent its own training set, I
see people learn as artists inthe classroom is very much.
They have impulses, they haveprompts for themselves, but
(23:42):
things are always returning towhat they think they know how to
do what they think art lookslike artists who they're trying
to emulate.
It's maybe not that far afield,it may just be kind of infant
right now.
The thing that you know the MoMApiece, the Anadolp installation
(24:03):
that was in the lobby of MoMAfor many months.
That was rethinking thecollection.
I watched it a number of timesand I was very interested in
what was happening, but to someextent what was profoundly
boring to me about it was theformat that it was given, the
screen, which is the same screenthat's used for other art
(24:23):
installations in that space.
It was given a kind of genericsound setup.
I feel like the sound thatbecame attached with the images
felt like middle of the roadvideo art sound to me.
A lot of what I didn't likeabout the piece was that it was
in a box that was predeterminedto make it look like art and it
(24:45):
made it look like video art andso I felt I had no sense of how
creative the AI inside was ableto be, because it was kind of
confined as a screensaver and Idon't know if there's something
to impact in that.
But I do feel like the abilityfor the intelligence to generate
(25:08):
or express things outside.
I don't know how you getoutside the box that the AI is
in, how things come into theworld, because right now things
do feel very confined by a lotof conventions of screen space,
even just how pixels are coded,and so that's kind of an open
question to me.
(25:29):
If I think about AI as aconceptual artist and I guess
here we get into the rise of themachines and things where you
do make me angry, that's I'msorry I'm kind of going on and
on now, but there was agenerational clash when I was in
graduate school in the early2000s around the computer and
people who were using computersas a digital tool, and one
(25:51):
professor in particular.
But a number of the facultymembers had this critique that
was you can't use Photoshopbecause the tools in Photoshop
are programmed by somebody elseand so you don't have the same.
There was some idea that,working within a format, you
couldn't be creative becausethat format was already limiting
you.
(26:11):
And then of course it didn'taccount for the way that our
generation was so native inPhotoshop that people
immediately found ways to workwith it, and I think I arrived
at that through my thinkingabout the AI's functional
limitations.
But that's kind of what I'moptimistic about about AI is I
think that it's just going to bepart of the bigger idiom, it's
(26:37):
going to be part of our visualculture, and people artists
always respond to visual culture, but it's also going to be a
tool that people are.
It'll be a tool that people areused to interacting with, and
we'll find ways to have ideas oftheir own in relation to it.
Do you feel?
Sam Gerdt (26:52):
like the format, the medium.
Do you feel that it matters atall or do you take a stance?
You're a teacher.
Do you take a stance with yourstudents about the mediums that
they choose to engage with ornot?
Obviously, you're payingattention to their expressions.
You're paying attention to whatit is that they're trying to
(27:14):
communicate and, as students,they're exploring and they're
not necessarily dead set on oneparticular direction.
How do you inform thatdirection?
Matt Saunders (27:28):
You think of all the things you could say to
provoke somebody and all thethings you could offer them as
potential other tools and kindof follow your gut about what
would be most generative.
Sometimes people feel like theyreally have their grip on a
certain way of working and youwant to encourage them to
develop their skills or to buildon their strengths to explore
(27:49):
other ideas.
And sometimes you feel likesomeone's ideas are being
confined by their defaultassumptions about how they're
going to generate them and youtry to give them a different
tool to work with.
Sam Gerdt (28:01):
I think that's a great example of what I see
lacking in any kind ofartificial intelligence for
quite a while to come, andpossibly even forever, it not
being possible at all.
Matt Saunders (28:16):
People sometimes ask me how do you teach art?
And I actually think that thatI don't know exactly how you
teach art, but I know that themodel of art education is a
dialogic model.
It's one of you making yourwork but then being confronted
with other opinions and othersuggestions, and it's always a
kind of crowdsourcing, andcertainly when people leave art
(28:38):
school, that's the big challengeis how to generate all that for
yourself.
So I always feel like what I'mdoing in the classroom is
modeling all the ways that onecan turn over a problem and all
the ways that one doubles downon something, and all the
techniques that I should havelearned to keep myself moving
and to generate ideas in thestudio.
(28:58):
Not that the people I'm talkingto will do the same thing, but
hopefully that they can have myvoice in their head and they can
build that for themselves.
Sam Gerdt (29:06):
I'm just not sure how AI challenges itself Right.
I'm not sure how it stretchesitself and tries to force that
model, breaking behaviorindependently.
I see artists and I seetechnologists and even just
normal people who are playingwith the technology.
I see them doing that andcoming up with really
(29:30):
interesting use cases anddifferent results, but I see the
vast majority of people justusing it as it's presented to
them, in the ways that aresuggested to them, and then what
happens is the fork in the road, the divide gets wider.
So those who are challenging itand really really pushing it
are the ones who are generatingtruly unique, novel, creative
(29:55):
outputs, and the ones who arejust using it are the ones who
are contributing to that mass ofunimpressive you know, hotel
art, thomas Kinkade type art.
I think the same idea applies toall uses of AI.
People who simply use thesystem, the systems that they're
(30:16):
given, they won't be standoutsin the areas of productivity or
effectiveness.
The people who create newsystems or subvert systems,
those will be the ones who excel.
So, as we look in recenthistory, I think the creation of
the internet is a goodanalogous example.
The internet is an incrediblypowerful tool and it has done
(30:37):
amazing things to change theworld, but the way that it
affects individuals is the sameacross the board.
I'm no different than you inthe fact that we have the
internet and it helps us to beproductive and effective, but if
you compare me to someone who,say, knows how to subvert the
internet, knows how to hack,knows how to program, knows how
to build new things, newarchitectures, new
(30:58):
infrastructures, that person hasa significant advantage over me
.
You can go even further back inhistory and see examples of
this as well.
In the 17th century, a fewmathematicians were
experimenting with square rootsof negative numbers.
They were breaking all therules of math and most
intelligent people thought theywere nuts.
Rene Descartes was the one whocoined the term imaginary
(31:21):
numbers, and he totally meant itas a slam.
Fast forward to now, and theworld as we know it couldn't
exist without imaginary numbers.
Most applied mathematics thatwe use in electrical and civil
engineering only works withimaginary numbers.
So the computer that you'reworking on right now or the car
that you're driving in, thistechnology that we're using
right now, could not exist ifthose rule-breaking
(31:43):
mathematicians had not dared toexperiment with something that
most people thought couldn't orshouldn't exist Before that
Leonardo da Vinci firsttheorized and then proved how
the circulatory system worked.
The popular belief of the daywas that the liver made the
blood and pushed it out to theperiphery, where it was absorbed
by the various tissues of thebody.
(32:03):
To discover the truth, da Vinciperformed dozens of human
dissections, many of which wereillegal.
He performed surgeries andexperiments on live pigs in
order to see the blood flowingin the body.
He studied the fluid dynamicsof water in rivers.
He constructed his ownartificial hearts out of glass
or wax.
Da Vinci's drawings anddiscoveries were a full 450
(32:27):
years ahead of their time.
So the question that I keepasking myself is if AI gives us
a system within which to operateand that system has rules, and
even if hekm memCH dies, what isthe means by which we improve
upon that system?
In all of human history, itseems, the means of improvement
was to break the rules, and isan AI equipped to do that?
(32:48):
I think what I feel, especiallyeverything that you've said,
talking about how you know youhave this gut instinct and you
have to, you know, constantly bechallenging, and all of these,
there's all of these littlepieces of what you're saying,
and I'm like, yeah, I just don't.
I don't see how an AI can everbreak out of the math of what it
(33:09):
is and make those leaps.
Hmm, but you know, maybe timewill tell.
I just there just doesn't seemyet to be a mechanism.
That part still seems to bescience fiction.
Matt Saunders (33:25):
I'm just nodding because I'm you've thought about
this a lot more than I have,and I buy that I will.
That sounds right to me ifthat's where we are.
I'm just not sure.
Sam Gerdt (33:38):
I don't know.
First, you see it Like a lot ofwhat happens is so far behind
closed doors.
All I see is all I see is whatthe world has like, what we're
given, and I see you know whatmillions of people are doing
with it generally.
And then I hear people like youwho have decades of experience
(33:59):
creating and teaching others tocreate and to participate in
these thoughtful conversationsthrough various means, including
art and all of the mediumswithin art, and I just I don't
see any of that happening.
When people take the tool thatthey're given and just apply it,
(34:21):
it's the ones who are like I'mgoing to take this and I'm going
to break it who are actuallylike finding anything creative
inside of it.
But it's them.
It's the act of, it's the actof breaking it, it's the means
by which they break it, it's thechoice to break it and then to
(34:42):
iterate on that process over andover and over again.
That seems to be generating anyamount of real innovation or
value, whether it's artistic ortechnological.
They're all agree with you.
Matt Saunders (34:52):
I mean, personally, I have not seen any
AI generated images.
That I was.
That I remember, yeah, but butI get, and it's crazy that I'm
now advocating for AI in somesense.
I don't want to predict thatthat's not going to develop.
I mean, I know, before thecurrent model, there was more of
(35:12):
an adversarial model.
There was the idea of havingtwo different programs, one
generating and one judging, andyou know, I think that there's
certainly ways that that thiscan get more and more
complicated.
This, those creators who startto work with AI maybe maybe AI
as a solo entity doesn't becomecreative for a long time, but as
(35:33):
a half of a kind of partnership, I could see it really being a
being a productive tool.
I do, you know, when you weretalking, I was thinking to the
artists who have centered AI asa topic in the kind of art world
that I'm tuned into, who Ifound most exciting and
provocative, have not beenworking with the tools that
we're talking about.
(35:54):
I think that this shift in thelast couple of years towards mid
journey or dally has created adifferent, has created a certain
type of beast that's based onmaking pictures quickly.
Yeah, isn't exactly where itdoesn't cover all of AI.
So I'm thinking of people likeIan Chang or Pierre Huig, both
of whom have thought more aboutthe idea of sentience in life
(36:17):
and kind of making artworks thatquestion whether something is
live or whether life can becontained within the space of a
video in Ian's work, or Pierrehas a a big piece in Kistifos,
norway, that is a forest thatwas scanned and then built as a
(36:37):
built as a digital environmentwith an AI God that's mutating
and creating variants and sothere's this overlay of this
kind of pristine wilderness andthen this mirror of it that
exists, guided by differentintelligence, and he's like
running these things and seeinghow they start to.
He's taking things that havebeen generated in the AI and
(36:59):
actually making them in reallife and putting them into the
forest, and it becomes this kindof a multiverse idea.
But those, to me, are not well,they're not an interesting.
I think both of these pieceshave made me think a lot about
our future, with potential ofliving minds and machines, much
(37:20):
more than any of the any of thepictures.
Sam Gerdt (37:24):
You talked about two, two AIs challenging one another
.
One of the things that thatthey do in in training AIs like
mid journey or Dolly or some ofthe the more common ones, is
they'll use human feedback toinform the the rightness, the
correctness of the response tothe prompt.
(37:45):
So that kind of human feedbackwas it turned out to be
incredibly important in in thetraining process in order to get
the results that that we get,and you still see this today.
You can, you can actuallyprovide your own feedback in in
all of these tools, whether it'san image generation tool or a
text generation tool like chat,gbt, and that feedback is
(38:09):
invaluable in improving theproduct.
The issue there that might beworth talking about is that you
have so many people providingthat feedback.
The baseline is quite standard,it's not, it's not deviating,
and so when you have artists whoare in some sense challenging
themselves or challenging otherartists, you I feel like there's
(38:32):
this distancing, thisdistancing from the norm, and I
feel like, with the kind ofhuman feedback that we provide
to an AI, all it's doing isconsolidating into a single,
almost a singularity of thenormal.
I actually kind of think thatthat's more of from a, from an
intellectual standpoint that'sprobably more of a of an
existential threat than even,like you know, whether or not
(38:56):
they can.
You know, cause a nuclearwinter is.
Is this idea that that ourhuman knowledge is going to be
reduced?
It's not going to be.
It's not going to be varied anddiverse, it's going to be
standardized, globalized Is isit not already?
(39:16):
I mean.
I think, it's more so than maybea hundred years ago.
I mean, technology, theinternet certainly has connected
us in ways that nobody everimagined, and so, yeah, human
knowledge and experience isdefinitely more globalized and
more standardized now than itwas, and there are good aspects
of that and there are badaspects of that.
(39:37):
But what happens if, if you goto the extreme with it and
there's only one?
There's only one answer.
It's not that it's right orwrong or that you know, it's
just that we've, we've trainedand we've trained and we've
trained, and now there's no moreroom for any more feedback and
and there's in some sense youwould have to have you start
(39:57):
from scratch, train somethingnew, but by that point you've
conditioned the people, and sothe people are thinking less
differently than they were ahundred years ago, and so now
when the people train the newthing, they'll be less capable
of training a diverse thing.
Matt Saunders (40:14):
It's very dystopian.
I think I have a better, a moreconfident view of human nature
in some regards.
I'd say not in regards to peace, but I have a lot of filmmaking
friends.
I mean also because I teach ina department that has filmmaking
and of course that's constantlybeen the conversation about the
(40:35):
Hollywood models and what'sallowed and how, how things get
made, and it's a question ofmarketplaces, it's a question of
economies as well.
I think there's always, alwaysgoing to be a lot of people who
want to do the opposite, and soit's more question of.
To me, it's not a question ofwhether that creativity will
emerge or exist in humanimagination.
(40:55):
It's like are we going to havea society where it can be made?
So I don't know.
I spoke earlier about you know Isaid it puts pressure on
language.
That's my strongest reaction alot of times, looking to AI I
mean even some of the articlesyou sent over for me to read,
you know is one where it's askedto make a Renaissance painting,
and I'm looking at this imageand thinking there's zero in
(41:18):
this image that has to do withthe painting.
You know that this is like it'san idea of painting that's
completely traveled through acultural flattening into a style
that is recognizable to thecomputer's painting, and it's
that word is is totally undoneand even the monk you know
there's Kermit the frog is monk.
It is zero relationship, Iwould say, to the things that
(41:42):
were made by the historicalperson, edvard Munch, but has
100% relationship to the posterof the screen.
And so, really, you know, myexperience of AI is both
complete irritation combinedwith the kind of provocation,
the kind of sense of how, howideas and language and styles,
how visual ideas move into verybroad understanding and get kind
of misrepresented or gettransformed, and so forces me to
(42:05):
grapple with thinking aboutwhat is Munch from my
perspective, having just gone tolook at a ton of his work in
the last couple of weeks, orwhat is a painting?
And I bristle, of course, Ithink, oh you idiots, that's not
a painting at all.
That is the flattest idea youcan imagine it.
But then it's up to me torespond to that.
(42:25):
You know, we're going to livein this world and and you know,
and if it forces me to beoutraged and feel like people
are misunderstanding, maybe thatpivots my work to engaging with
those feelings more, versusjust assuming that we're all
communicating and that's, for me, the silver lining is that it,
like things that have happenedin all kinds of pop culture.
(42:46):
I think it has flattened, but italso creates a space for a
different type of of weirdness.
On top of the form.
It just becomes a differentpoint of reference or kind of
touched on it Does it?
Does it piss me off when peopleare sharing images that I think
are absolutely banal and sayinghow amazing they are.
(43:07):
Do I think?
Don't be an idiot, come to themuseum with me.
Yes, of course I think that,but I don't know what else we
can do.
It's how culture is all.
Sam Gerdt (43:17):
You're highlighting the exact same feelings that I
have in different areas.
You're highlighting how youlook at an AI-generated image
and there's an immediaterecognition of what it is and
you're highlighting the factthat, in that recognition,
you've already made a decisionabout it and you're already
moving on, and there's a sensein which a person can make
(43:43):
themselves even less distinctthan if they did nothing by
generating the AI image.
It's like you've justclassified yourself in some
sense in my mind, because Irecognize immediately what
you've done.
There's nothing creative aboutit.
(44:04):
It falls into the exact samecategory as every other
AI-generated image I've everseen.
And then I'm so quickly movingon.
I'm ready to move to the nextthing, and I'm not necessarily
saying that that's exactly whatyou just said, but I have felt
that way.
Certainly, and going back towhat I said earlier, that you
(44:25):
used words that I think betterdescribe how I was feeling
cultural, flattening this ideaof removing the diversity and
variation that exists withincultures and giving us all just
a baseline understanding of therightness or wrongness of
something, the goodness orbadness of something.
(44:46):
If we look at life 100 years agoversus life now, I feel like
the disconnected nature ofhumanity, the localization of it
resulted in wildly distinctopinions, worldviews, art, all
(45:08):
kinds of distinction.
And the more that you exposeyourself to all of the rest of
the world and everyone elseparticipates in that as well
that distinction starts toflatten, it starts to diminish.
So I guess you could phrasethat as a question what are some
of the differences that you seebetween an artist who is a more
(45:32):
modern artist, who is exposinghimself to all of the world, all
of culture, who's absorbing allof that, versus, maybe, an
artist from centuries past whowas very local, his experiences
were very limited?
Do you see those differenceswhen you look at their work?
(45:52):
Do you feel those differences,or do you interpret it all from
your own perspective?
Matt Saunders (45:58):
Of course to see those differences and of course
I interpret it.
I feel like we can't undo theinterconnected nature of the
world and I think that we'reprojecting, maybe backwards, a
little bit of difference onthings that feel from a very
different time.
You know, if you think aboutwhatever, like Baraceli in
Florence is living in a Catholicmonoculture with near
(46:22):
parameters, never left his sidegoing to Rome once.
He never left Florence.
There's a flattening to that.
Perhaps Now we see it inrelationship to other things
that are happening.
So we don't see it as aflattening.
But I share your anxiety, butnot maybe your pessimism.
Sam Gerdt (46:45):
That's a pretty common response.
Matt Saunders (46:47):
And yeah, I, I am not.
I am someone who has A mixedrelationship to kind of the
biggest kinds of pop culture.
I'm a little out of it and I'veworked in my own work kind of
against it.
I've tended to be moreinterested in unearthing stories
and kind of digging into thearchives and countercultures.
(47:07):
But that's more who I am.
That's not a claim thateveryone should be like that.
Growing up in the sort of late20th century, the art history we
were taught about painting wasalways about the crisis of
photography, the idea thatphotography comes along and this
is, of course, a very flatdescription of this idea but the
photography replaces therepresentational function of
painting and that opens up acentury of abstraction and
(47:32):
innovation and crisis.
It's more complicated than that.
But if we if enough people arefeeling a crisis of flattening
out of AI, then there's a reallyinteresting crisis that we can
work with.
I don't think that we'retrapped.
We're trapped to live in aworld where it's going to be
full of these images.
Yes, that's going to happen nomatter what, but I don't think
(47:53):
that as artists we're not ableto push back against that.
If that starts to feelsuffocating, I found myself
thinking really quickly aboutnumber of artists who emulate
popular forms, like artists whotry to become pop stars, or the
large number of artists who getsuccessful and then try to make
a feature film and there is thiskind of artistic creative
(48:14):
impact to infiltrate or to actthe culture.
So I wonder, if an artist actsas AI, how that, what that space
might be at some point, likeimpersonating Dolly, hacking
someone's system and never mindCut that part of the video.
Sam Gerdt (48:32):
I know Matt felt like he was derailing the
conversation with thisadmittedly silly idea, but here
I am in the edit a few weekslater and I cannot help but
Google this idea of humansimpersonating AI.
So, without getting toosidetracked, here are two
interesting examples that Ifound.
Last summer, several creatorson TikTok started trending for
(48:55):
their NPC live streams.
They would stand in front ofthe camera with their hands held
awkwardly in front of them andsway back and forth like an AI
video game character.
Viewers could pay money to sendvirtual gifts to these
streamers and the streamerswould react to the gifts in the
same canned way that you'dexpect from an AI avatar.
It is absolutely bizarre.
You can go on TikTok and findthese people still doing this,
(49:21):
but apparently it's wildlypopular and the whole thing
reminds me of the living statuestreet performers in New York
City.
The second example I found isfrom 2018 and it's a startup
strategy and it's called pseudoAI.
There were several casesreported where startup companies
would present themselves toinvestors or customers as AI
powered companies, but theprocesses that were supposed to
(49:46):
have been AI powered wereactually offshore contractor
powered.
The idea was to reduce or delaythe initial development costs
of implementing AI poweredservices, so, by taking
advantage of cheaper offshorelabor.
Startups could more efficientlyprototype their processes.
Once they scaled, they couldhand things over to the AI
systems.
(50:08):
What I found interesting aboutthis example is the people that
they interviewed for thearticles that I found who were
doing this work were expressinghow much they hated it, how
mundane it was, how much theycould not wait for a bot to take
their job.
So in answer to Matt's questionof what would it look like, I
kind of think imitating AI,imitating the computer, looks
(50:31):
boring, it looks bland, it looksunappealing, and I think that
says something about theconversation we're having here.
The experience of being humanseems to me to be far better
than the experience of being acomputer when you look at the
(50:56):
students coming in today the artstudents coming in now and you
compare them to maybe when youwere first starting as a student
.
What are some of the biggestgenerational differences that
you see when answering the samequestions?
Like you know, what is?
(51:16):
Creativity, what is art?
Matt Saunders (51:17):
That's a really good question and there are
definite differences.
There are differences about theassumption of what art does and
where it lives.
Certainly, I have one reallygreat student in particular
whose highest goal is to be onInstagram and those are
fundamentally different ideasabout what you're actually
(51:39):
producing and what's circulating.
A sense of what rolemateriality plays in a work is
quite different.
I think the thing that I quotethe most is something that a lot
of artists quote I'm not evensure if it's possible, but it's
always attributed to Ed Ruschaand it's this idea that some
work you look at and you say wow, huh, but that he loves work
(52:02):
that you say huh, wow, you know.
That made so much sense to mewhen I was told that, because I
believe that work should bedifficult and should open up and
should be inscrutable andshould be complicated to read,
and I think that thatfundamental sense of values has
shifted a bit.
I think that the legibility,efficiency, speed of consumption
(52:22):
, delivery of a message, what Iwould call the kind of more
illustrative quality of work,which was a very pejorative term
when I was in school, is now apositive for a lot of people how
quickly we consume these images, how quickly we read the thing
and judge it.
Sam Gerdt (52:40):
I hadn't thought too much about this shift in favor
of the illustrative until Mattmentioned it, but it's something
that I also recognize.
It's an evolution that makessense when you consider how our
world has changed over the past30 years.
We live in an attention economy, in a digital world.
Now.
The demand for things likegraphic design and digital
illustration has skyrocketedsince the mid-90s, and that
(53:03):
demand tugs on everything.
The way we look at art ischanging too.
The pace at which we consumeand demand new content is
mind-numbing.
Compared to previous decades,we look at art on four-inch
screens instead of a canvas.
In having a conversation aboutcreativity, it's hard to ignore
(53:23):
the fact that our collectivedefinition of the word itself
and the standard by which wemeasure it has changed radically
, and I would argue that it'snot changing for the better, but
for the worse.
Matt Saunders (53:38):
And there are ideas that come up around the AI
question, around rightness,around function, around
efficiency that I think, forsome of my generation, raised
some hackles.
In relation to the creativity,I fundamentally don't think that
art is problem solving.
I feel like it's makingpropositions which may or may
(53:59):
not mean anything to anyone, andif they mean something to
someone, then they become agreat proposition.
But I don't feel like I'm everI mean so consciously on some
level.
You, of course, are, but Igenerally don't feel like I'm
judging my work in terms of doesit do what it should do?
It's more what did it do and isthat something worth keeping?
(54:19):
Is it a hope that it arrives atsome provocative, unexpected
state?
Maybe that's the jump over thefence that I started out by
talking about as creativity.
If it's not what I expected itto be and it seems better, then
that's the work that I cherishand put out into the world and
(54:41):
that's kind of my approach tomaking.
But I think that I take veryseriously that I don't.
That doesn't seem like animmediately shared value.
Sam Gerdt (54:52):
As an artist yourself , your work tends to be a
mixture of material and digital.
Yeah, mixture of physical anddigital.
You definitely do a lot ofexperimenting with both.
Do you have a?
Matt Saunders (55:05):
favorite?
Oh, 100%.
The digital is secondary andcomes in almost in terms of its
flexibility in showing the work.
Most of the digital things I doare digital videos which are
made out of real.
The work is always made withreal materials and then scanned,
(55:26):
and I'm very conscious thatbetween the scan and the
experience of the installationor the projection in the gallery
there's a lot of things you cando in the digital space.
That's a kind of digitalmateriality and I do play with
that and experiment with that.
But everything starts on thepage.
It starts with, in my case, abrush of some kind, because I
(55:47):
hate pencils.
Sam Gerdt (55:47):
But yeah, again it's.
Will an artificial intelligenceever express anything like that
?
I mean even just like I alwaysstart this way because I hate
that.
It's like, well, wait a minute,what do you mean?
You're gonna have to explainthat to me.
(56:09):
It's that, it's that diversity.
I think that that makes it so,so appealing, anything so
appealing.
And as soon as you, as soon asyou remove that element, I'm in
marketing now, and so, like,what is marketing except to pull
(56:30):
on an emotional string?
I just don't.
I don't see how we are doingany better, accomplishing that
by being overly dependent on onthese tools, when all that they
do seems to be reductive, andmaybe that's like you said, I
tend to be a pessimist withthese things.
Even you talk about thegenerational differences, your
(56:53):
students.
I've talked to other professorsbefore and they've said exactly
the same thing.
I have, I have, students whosegoal, whose aspiration is to be,
is to be, on social media.
I bristle at that because itseems to me to be foolish in so
many ways.
But the flip side of that is,I'm also belonging to a
(57:16):
generation who just doesn't getit in the same way, and that's
why I asked about, you know, thegeneration that you're teaching
now it seems like they're.
They're living more of adigital life, and so the the
conversations that they'rehaving are going to be digital,
and the the ways that they'rethinking and the issues that are
(57:38):
affecting them are going to bedigital.
I, I'm, I'm with you.
I much prefer the material, butI also don't have any interest
in living in that digital world,in that digital space there's
such a diversity of spaces inwhich to live.
Matt Saunders (57:55):
Yes, it's would seem horrendous to me to be to
be big on social media, but Iextend that also to the idea of
genres and and something that Ialways thought about visual art
is that it's behind a lot ofother types of art making, like
within, you know, within music.
(58:16):
We really understand thediversity of different types of
music and what genres differentthings fit into, and it's
absolutely commonplace to say Ilike this and I don't like this,
but we do seem to often havethis kind of fatalitarian
impulse to define all the art ofa time and I've never seen it
that way.
There's so much art that Irespect and I respect people's
(58:40):
desire to want to do that, but Ihave absolutely no interest in
making that kind of work.
For me at the moment, ai isvery much in that category.
But the world changes and new,you know, new genres come and
new places to live come and I amcurious about next generations
(59:03):
and what options they have andand I'm trying to understand it,
which is maybe why I'm soopen-minded right now you talked
about the flatness ofBotticelli, living in Florence's
whole life, hardly ever leavingexcept for the one time, and
how.
Sam Gerdt (59:16):
That is a flat existence in its own way.
And how, the the differencesyou know.
From our perspective, we see.
We see that existence with allof the.
You know thousands and millionsof other existences and we can
appreciate each of them fortheir diversity.
Looking at all of them togetherand how, now, with the, with a
(59:39):
more global view, global mindset, there there seems to be a
global flattening in some sense.
It seems like there's a wholelot of value in understanding
where you live and where youbelong and what your particular.
As individuals we can't livethese incredibly diverse lives.
We are going to have our ownflatness to us, as you know,
(01:00:02):
beings living in a particularplace at a particular time, with
particular beliefs andexperiences.
There's flatness in that.
But understanding where youbelong and then staying there
and developing it and developingthat conversation with other
unique individuals, whilemaintaining your integrity, that
seems to be to me the goal ofany artist today to know who you
(01:00:27):
are and and what, what yourpurpose is and and what your
belief is, and then taking thatand interfacing with other
people who are doing the samething.
The threat from AI that I thinkI've talked about could be
restated as the antithesis ofthat idea.
People know less of who theyare and where they belong.
(01:00:51):
They know less of theirindividuality, their uniqueness,
and they see themselves asbeing just one tiny piece of
this big, more fists blob ofpeople and culture and just a
mishmash of everything.
Yeah so, yeah, so I wow.
Matt Saunders (01:01:14):
It's such an interesting thought yeah, so
dark, I mean it's so dark andwith you on that.
I think it's a well, this is no.
Sam Gerdt (01:01:29):
I was just gonna say this has been one of the
weirdest conversations about AIthat I've ever had.
Matt Saunders (01:01:36):
I feel like I leave with more questions than I
came in with and I keep comingback to the, to the film idea
that came up and and the sort ofthe history of of that medium.
Since it's relatively short, youcan look at it and think about
the way that studiosconsolidated the ability to make
(01:01:57):
films and you have an era thatgoes through all the studio
codes and then there's a momentwhen the tools change and indie
cinema becomes, you know,accessible and then that gets
consolidated and you know we'reliving in a moment of kind of
consolidation around televisionand film, but also in opening up
(01:02:19):
of all the tools, including AI,that can make movies.
I have we haven't talked aboutthis, but I do have I've
encountered film students atHarvard who are using AI to edit
their clips and even generatethings, and you know, I think
there's some curiosity aboutthat, but that's part of where I
have hope about this is.
I feel like right now, all our,all our anxiety is up and all
(01:02:43):
our eyes are focused on one ortwo big platforms, and it seems
very flattening.
But I do.
You know, I have to assume thatthings will come along that
will change those dynamics andcreate radically different
versions of AI creators andmaybe a kind of underground and
(01:03:05):
off, off cycle, and I could betotally wrong.
But there's the basic questionyou started with about
creativity, but for me thatimmediately gets into questions
of the marketplace and the toolsof production, the means of
production for that, and yeah,and I think that we're at the
real infancy of accessibilityand and use of AI.
Sam Gerdt (01:03:29):
So hopefully goes and hopefully people mess with it
in the right ways yeah, well,for my part, I, I put my, I put
my money on the artists and theinnovators and the technologists
who are doing that breaking,who are doing that challenging
(01:03:50):
and who aren't, who areresisting the urge to just join.
You know, go go with that flowof of where these tools might
take us, especially in theirinfancy, because, over and over
and again, again, I feel likethat gets validated in
everything that I see the oneswho are actually adding value to
the world or the ones who are,who are trying to break AI who
(01:04:16):
are?
trying to recreate it intosomething different than what it
is, not the ones who are usingit, right?
Yeah, yeah, well, that is aninteresting discussion.
I love that, matt, I thank youfor.
Thank you for talking to me.
(01:04:37):
It was, it was a real pleasure.
You are an incrediblyinteresting person to talk to.
I love your ideas and maybemaybe in time we reconnect and
see how you know how we feelabout what we've said, how we've
changed, how we've been changedand, and at some point I'm sure
(01:05:01):
we're both going to have toengage with AI a little bit more
, break some, break down some ofthose walls that that we have,
those those resistances thatexist, and and I I would love to
see, like, what happens withthat yeah, yeah, no, thank you
very much.
Matt Saunders (01:05:20):
It was really I'm .
It was surprising that youreached out and I've really
enjoyed the conversation.
I share your anxiety, but not maybe your
pessimism.
Sam Gerdt (00:13):
That's a pretty common response.
Welcome everybody to Road WorkAhead, a podcast that explores
the unmapped future of businessand technology.
My name is Sam Gerdt andtoday's interview is oh so very
good.
Is AI creative?
It's not a simple question toanswer.
(00:35):
Creativity is a complex conceptand we have to dig into the many
ways that word could be defined.
Who better to ask aboutcreativity than someone who's
both an accomplished artisthimself and an instructor at one
of the most distinguished artschools in the country?
Matt Saunders is a professor ofart, film and visual studies at
Harvard University.
(00:56):
His art has been featured inexhibitions around the world,
and the list of museums with hiswork in their collections
includes the Museum of ModernArt, the Guggenheim, the Whitney
Museum of American Art and manyother names that you would
likely recognize.
Matt's art is grounded in thematerial, particularly painting,
but in most of his work you cansee how he challenges himself
(01:20):
to push back against the idea ofa static rectangle of paint on
canvas and create something moredynamic and transient,
something more like anexperience than an image.
That same resistance toconventionality comes out in our
conversation about AI, and itleads us to some brutally honest
conclusions about how we viewthe techno obsessed world around
(01:43):
us.
Matt is a very thoughtful andinspiring person, and this
interview is easily one of themore impactful conversations
I've had in a long time.
Matt, thank you for joining me.
I really only have one questionand I'm sure we'll get into it
after that, but my question isartificial intelligence is it
(02:05):
creative?
Matt Saunders (02:07):
Well, I don't even know how to answer that
question until I think I need tounderstand what you mean by
creative.
I would say yes, but I feellike maybe you have this
availance on creativity that isvery perhaps ennobled in the way
that people think about it inrelationship to art, which is
maybe not how I'm answering.
Yes.
Sam Gerdt (02:29):
That's probably true.
Let's go back to the beginningand let's ask a more basic
question what is your definitionof creativity?
Matt Saunders (02:38):
To be honest, it's not a word I think about or
use very much, but I think thatI would define it as putting
things together in ways thatmake leaps, whether that's a
leap out of a rational path, orwhether it's a leap out of a
(03:01):
series of expectations, whetherit's an inversion of the status
quo.
The difference between an actand a creative act for me might
be one that just hops over somekind of fence.
Sam Gerdt (03:14):
Yeah, you said you don't use the word creativity or
you don't think about it awhole lot as an artist.
Is there a reason for that, oris there another idea that comes
into play far more often thanthe idea of creativity?
Matt Saunders (03:29):
I think I'm a little allergic to the kind of
mythologies that we build aroundartists as a different type of
creature, a different type ofactivity.
I feel like there's acontinuity between making
breakfast and making what turnsout to be a milestone painting
in someone's life.
There's an activity of focus,of thinking, of reacting to past
(03:54):
experience, of working withmaterials.
That is part of creating.
Sometimes I think thosecreations are surprising.
Sometimes that's surpriselinger.
Sometimes it's just somethingthat surprises you at first and
you get over.
Often things come back to staywith you or to challenge you or
(04:18):
to make you think differently inthe future.
In my experience and maybe thisis just the type of artist I am
I don't have a great amount offaith or predictive capacity in
the act of making.
I don't always know when I'mbeing creative.
I feel very strongly thatthere's a difference between the
(04:43):
act of making something, thesort of generative act of
putting things together, andthen the moment of interpreting
it or seeing it or understandingit, which is always based in
time and place and changes overyour life.
When I say that I think AI iscreative, yeah, it makes things,
it's creative, it's creatingthings.
Now, whether that is a virtue,whether creativity with a
(05:05):
capital C is some specialcategory that gives you special
status.
I don't know that.
I even would give that as ablanket to myself or to other
people.
I think it's something thatyou're touched by from time to
time.
Sam Gerdt (05:18):
When I ask the question, you're an artist,
you're an accomplished artist.
Why are you an artist?
Matt Saunders (05:26):
I like things, I like making things, I like
thinking about images.
I'm a visual person and not somuch of a musical person or even
a word-based person.
For me, it's a way to work withmy hands and my mind in a
different way than I'm calledupon to do in other types of
(05:47):
professions.
I think I get great pleasureout of that time thinking in
other ways.
I just love art.
There's always, I think,something I feel kinship with,
with my imagination of thesentience of an AI who's been
trained on data sets, is there'salways an aspect of emulation.
(06:09):
There's an aspect of beinginspired by things that you've
seen other people do, thingsyou've seen in the museum, and
that itchy impulse you get torush back to the studio
afterwards.
I was reflecting on that when Iwas trying to think about AI
before this conversation, that Iguess I do have my own data
sets.
I've been trained to.
Sam Gerdt (06:28):
I think for me, this is a question that I've been
wrestling with for a long time.
That is, what about the factthat we are essentially
computers with our own data sets?
Why am I so quick to say that Ican be creative, but that an AI
can't?
And I think what it comes downto is capacities, and whether or
(06:50):
not an AI can ever have thecapacity that a human can for
synthesis, and I came across aquote by a Dr Blumenthal who, I
think, sums it up very well.
He says there are nopsychological qualities.
In physics, for example, thereis no red, no green or blue.
In the physical world, redness,greenness and blueness are a
(07:14):
phenomena that are created bythe cortex of the experiencing
individual.
A musical quality, the flavorof the wine or the familiarity
of a face, is a rapid creativesynthesis that cannot, in
principle, be explained as amere sum of elemental physical
features, and so, consideringthat quote, I have to conclude
(07:36):
that a computer which can onlyever experience the elemental
physical features of the world,will never be capable of this
creative synthesis.
That's being described In myown thinking on this topic, and
this is something that I think,from the beginning, has rubbed
(07:57):
me the wrong way, in the waythat we are being conditioned to
talk about things likeartificial intelligence.
There were a lot of terms thatstarted being used a lot more so
in the general, you know,public conversation, we started
hearing more and more aboutcreativity.
We started hearing more andmore about generative AI and we
(08:25):
started being I feel, we startedbeing conditioned to believe
that artificial intelligence wasor would become something
threatening to human creativity,and I immediately didn't like
that.
I didn't think it was accurateand I didn't think that it was
(08:45):
healthy for us to even engage inthat kind of thinking, and so,
from the beginning, I've beentrying to figure out why that is
.
You know, I have this gutreaction to it.
But why do I have that gutreaction to it?
And I, you know, I started mycareer.
I went to art school, I studiedgraphic design, so I have a
(09:05):
little bit of a background inart not quite fine art, but I've
participated in the art worldfor a significant period of my
life and what I see is a bigdifference between something
that is creative and somethingthat is generative.
I feel like creative is adescriptor of content that, like
(09:27):
what you're saying, there's athere's a leap there, it
represents a progress that thatwas maybe unexpected, maybe much
needed, but undiscovered.
But there's something new in it, there's something novel in it,
and I also feel like creativityhas to be rare.
(09:52):
I feel like there has to be ararity to it, because how can
you call something that's commoncreative, how can you call
something that lacks that?
It lacks that novelty, thatleap that I think is needed.
And with AI, I think what we'reseeing that's difficult is
(10:12):
we're seeing a blurring of thelines between the artist and the
tool, and we're also seeingthis blurring of the lines
because of how fast things aregoing.
We're seeing AI generatedcontent that is truly beautiful
and technical and it has all ofthe characteristics that we're
(10:33):
accustomed to seeing, andsomething that really truly is
creative.
But the problem is that wehaven't yet caught up to the
fact that, because AI can do itnow, now it's common.
So when mid-journey first cameout and we saw the images from
this stable diffusion is anotherexample the images that it was
(10:57):
generating were striking.
They were beautiful and theywere very stylistic and they
looked like art, and you couldargue about the goodness or the
badness of that art, but theylooked very artistic.
I saw that and I would say itlooks creative but it's not.
And I would say, in a year'stime when 30 million people are
(11:20):
subscribed to this tool and theyeach are generating 10 images a
day, and you have 100 million,200 million of these same kinds
of images flooding into theinternet every day, I have a
feeling you're going to look atit and you're going to feel very
differently about it than youdo right now, seeing one or two
(11:41):
of them here and there, becausethat commonality is going to
make it feel uncreative to you.
Matt Saunders (11:50):
I think I'm with you with what you've said, but
there's a gap that opens up thatI was thinking about when the
creativity happens, and whethercreativity is a singular act.
We think about AI generatingsomething quickly and we ask was
that creative?
I feel very strongly that allart, all aesthetics, all value
is distributed value.
(12:13):
Art is a language and it onlyis good if it means something
and it has the capacity to last.
The magic of art is that thingscan last for generations and
work can be meaningless fordecades, centuries, and then
suddenly mean a lot to a lot ofpeople.
Maybe that's why I strugglewith the word creativity,
(12:34):
because it seems to arrive andjust be, and I actually don't
think that work's power existsin the world in that way.
It might be very powerful inits moment of creation, through
its novelty.
It's a word that you used a lot, that I kind of chafed at as
the ultimate value, but I have amuch longer view of things.
(12:55):
I think what really makes senseto me and what you're saying is
the idea of flooding the market.
Something I really like about AIand the AI images we're seeing
is that it puts pressure, Ithink, on our assumptions about
value in art, words like beauty,words like artistic.
I might define differently froma lot of people, and that's not
that I'm right or people arewrong.
(13:16):
We live in different contextsand we look at different things,
but I think that there's a kindof unreflected assumption of
some values, like the value ofwhat we think of as beautiful.
If the world is full of acertain type of soft focus,
light infused, I don't know I'mtrying to imagine what an AI
beautiful image would beluminosity then will that
(13:38):
continue to be beautiful?
I think you're asking thatquestion and I agree with that,
because I do think that one ofthe things that AI is doing now
for me is generating a lot ofimages that I think are pretty
drecky.
I think that they are a littlebit received ideas of what art
might look like.
(13:59):
I find most of them completelyuninteresting, but I also see
their appeal and I'm wrong tothink they're uninteresting
because so many people thinkthat they're interesting.
But I appreciate that they'reso easy and that it's not going
to replace artists, but artistsare going to react to how the
visual world has been changed bythis type of image and the
(14:20):
access and ability for people tocreate the AI to create all
these images and that thecreative responses will actually
be the rejections of it or thepeople who find ways to actively
step to the left of it andoffer some kind of alternative.
An alternative will only begood in the moment when the
(14:44):
baseline is the flood of AIimages, and then it will evolve.
Sam Gerdt (14:50):
There was the article that you responded to.
The question that you respondedto in an article several months
ago about this very question isAI art really art?
One of the comments that youmade in that article that I
appreciated was that artistswill be the one to help us
(15:10):
imagine the new world that AI ishelping us to create.
The article I'm referencinghere was published by the
Harvard Gazette in August of2023.
It's called Is Art Generated byArtificial Intelligence Real
Art?
Here's exactly what Matt saidin the article.
To the question of whether AIcan be a threat or a
collaborator, I might respondthat every new technology upends
(15:34):
conventions and delivers notonly new possibilities, but a
new kind of materialintelligence.
I'm sure that many artists willbe intrigued by the agency of
AI and seek ways to grapple orcollaborate with it.
Many already are, and we shouldbe grateful to be challenged
and knocked out of our habitsand assumptions.
Most of the things that worryme about this fall into the
(15:56):
realm of the social and ethical.
I hope there are great artiststo help us imagine around and
work with this new reality.
I think what you were saying isthat the masses are going to use
AI for a particular means andthey're going to move the world
in a particular direction.
(16:16):
And those who are on theoutskirts of that crowd, those
who are more imaginative, thosewho are more creative, are going
to help to pull that mass ofpeople in a particular direction
or help us to understand it innew ways.
(16:37):
Because in some sense, the vastmajority of people are just
following the crowd.
And there are those on theoutskirts who look around and
say you know, they're seeingfurther and they're looking at
places that others aren't, andthey're helping that crowd to
imagine, to see what they see.
I really appreciated that, butI think I appreciated it because
(17:00):
it reinforced my own feelings,and that is that it's those
people who stand on theoutskirts of the masses who are
looking elsewhere.
Those are the people who areexpressing themselves creatively
, teaching others creatively,informing all of our
imaginations and all of ourthinking, and those end up also,
(17:22):
I think, being the ones inother areas of life that are the
innovators, that are theproblem solvers who end up
taking us places we neverthought we would go.
If we do arrive at a placewhere artificial intelligence
truly isn't an entity among us,something of a peer, it will be
because of those people, thoseinnovators who did truly
(17:46):
creative work.
It will be the result of that.
Matt Saunders (17:49):
I mean, the categories we're talking in are
just too big to get that muchtraction, because I think you
know the role of an artist who Iwould admire.
The kind of artist that I'minterested in in my very narrow
set of interests does manythings.
One is making images or ideasor objects that participate in a
(18:12):
current conversation.
I'm not that taken with themyth of the undiscovered genius
and the attic that I think thatthere is a current conversation
in the art world and there's amillion art worlds and that you
are making things that are partof that.
Of course, the people who I'mreally excited about and the
(18:33):
people who, in retrospect, seemlike the most important in the
generations are the people whobring a particularly critical or
transformative or a differentperspective to things, and I
think that you know that tendsto have an aspect of criticism.
So the artists who are workingwith AI right now who I'm most
(18:54):
interested in are the ones whoare kind of pushing its limits,
people who aren't even known yet.
I have a former assistant who'strying to force the AI to break
its own rules, you know, who'sconstantly working with the
prompt and treating it as a kindof dialogue instead of focusing
on the image or what could becreated.
(19:16):
He's focusing on thechallenging creative activity of
interacting with another makerand pushing boundaries, kind of
getting it trying to force theAI to fight with him, and to me
that's a kind of artistic, acreative approach to using the
AI in line with conceptual artand kind of like generations of
(19:41):
artists who've thought aboutarts relationship as pop to more
popular culture, saying not toreject it but to find ways to
cut under it or to lay thingsbare, or to add friction, add
turbulence to what could be asmooth kind of marketplace.
Sam Gerdt (19:58):
Does that necessarily lead you to the conclusion that
AI itself could never be anartist?
Matt Saunders (20:05):
I'm not smart enough to know that, because I
assume that AI can internalizethose strategies.
The AI we're seeing now, Idon't think is being fed David
Hammons or other artists who'vereally worked against the grain,
and people are using it to makeimages.
Sam Gerdt (20:22):
Well, that's what I'm not entirely clear about is,
you know, from a technicalstandpoint, ai can't respond
with anything that it hasn'tbeen trained on.
That is, its responses arenecessarily a diffusion of the
data that it already has.
And if the data that it alreadyhas is limited to what has
(20:44):
already been made by people,then can AI add anything to the
conversation?
They've done experiments withthis.
Where the question is well, canAI generate its own training
data?
Because that right now is thecommodity that's in short supply
is training data.
So the question is well, if youcould take an LLM and from the
(21:07):
training data that it alreadyhas, if it could generate new
training data, maybe we couldtrain a smarter LLM.
And the end result is itdevolves into nothingness.
So when we use AI-generatedcontent as training data for
future AIs, we call this contentsynthetic data.
(21:31):
Synthetic data is not generatedby people, it is generated by AI
.
The usefulness of syntheticdata is going to vary from case
to case and model to model.
So, for example, it makes a lotof sense to use synthetic data
for training a self-driving AI,and we see a lot of support for
this kind of synthetic databeing used in this use case.
(21:53):
But when it comes to generativemodels like the ones we're
talking about here LLMs anddiffusion models that are being
used to create written andvisual content the use of
synthetic data to further trainthese models does have a
degenerative effect.
A July 2023 study entitledSelf-Consuming Generative Models
(22:13):
Go Mad went into great detailabout this phenomenon.
In the paper, the authors saidOur primary conclusion across
all scenarios is that withoutenough fresh, real data in each
generation of an autophagousloop, future generative models
are doomed to have their qualityor diversity progressively
(22:34):
decrease.
Now, this is just one example,but the point that I'm trying to
make is that the generativemodels we have today are not
capable of synthesizing data ina way that could be considered
additive.
So that's where I'm wonderingcan an AI be an artist if it's
incapable of contributingsomething novel to the
(22:55):
conversation, if it's incapableof developing the conversation?
Matt Saunders (22:59):
My understanding of the diffusion is that it's
partially about adding static,and I think anytime you add, you
make a space for thecounterproductive, the
irrational, the random, and thenyou hold that together with the
form that you were given andjudge whether it's working.
I think that's kind of what anartist does.
(23:20):
I feel like that is generativein a way that transcends its
dataset or could be.
Maybe it hasn't gotten thereyet.
I mean, of course I also amteaching, and when you're
describing AI not being able toinvent its own training set, I
see people learn as artists inthe classroom is very much.
They have impulses, they haveprompts for themselves, but
(23:42):
things are always returning towhat they think they know how to
do what they think art lookslike artists who they're trying
to emulate.
It's maybe not that far afield,it may just be kind of infant
right now.
The thing that you know the MoMApiece, the Anadolp installation
(24:03):
that was in the lobby of MoMAfor many months.
That was rethinking thecollection.
I watched it a number of timesand I was very interested in
what was happening, but to someextent what was profoundly
boring to me about it was theformat that it was given, the
screen, which is the same screenthat's used for other art
(24:23):
installations in that space.
It was given a kind of genericsound setup.
I feel like the sound thatbecame attached with the images
felt like middle of the roadvideo art sound to me.
A lot of what I didn't likeabout the piece was that it was
in a box that was predeterminedto make it look like art and it
(24:45):
made it look like video art andso I felt I had no sense of how
creative the AI inside was ableto be, because it was kind of
confined as a screensaver and Idon't know if there's something
to impact in that.
But I do feel like the abilityfor the intelligence to generate
(25:08):
or express things outside.
I don't know how you getoutside the box that the AI is
in, how things come into theworld, because right now things
do feel very confined by a lotof conventions of screen space,
even just how pixels are coded,and so that's kind of an open
question to me.
(25:29):
If I think about AI as aconceptual artist and I guess
here we get into the rise of themachines and things where you
do make me angry, that's I'msorry I'm kind of going on and
on now, but there was agenerational clash when I was in
graduate school in the early2000s around the computer and
people who were using computersas a digital tool, and one
(25:51):
professor in particular.
But a number of the facultymembers had this critique that
was you can't use Photoshopbecause the tools in Photoshop
are programmed by somebody elseand so you don't have the same.
There was some idea that,working within a format, you
couldn't be creative becausethat format was already limiting
you.
(26:11):
And then of course it didn'taccount for the way that our
generation was so native inPhotoshop that people
immediately found ways to workwith it, and I think I arrived
at that through my thinkingabout the AI's functional
limitations.
But that's kind of what I'moptimistic about about AI is I
think that it's just going to bepart of the bigger idiom, it's
(26:37):
going to be part of our visualculture, and people artists
always respond to visual culture, but it's also going to be a
tool that people are.
It'll be a tool that people areused to interacting with, and
we'll find ways to have ideas oftheir own in relation to it.
Do you feel?
Sam Gerdt (26:52):
like the format, the medium.
Do you feel that it matters atall or do you take a stance?
You're a teacher.
Do you take a stance with yourstudents about the mediums that
they choose to engage with ornot?
Obviously, you're payingattention to their expressions.
You're paying attention to whatit is that they're trying to
(27:14):
communicate and, as students,they're exploring and they're
not necessarily dead set on oneparticular direction.
How do you inform thatdirection?
Matt Saunders (27:28):
You think of all the things you could say to
provoke somebody and all thethings you could offer them as
potential other tools and kindof follow your gut about what
would be most generative.
Sometimes people feel like theyreally have their grip on a
certain way of working and youwant to encourage them to
develop their skills or to buildon their strengths to explore
(27:49):
other ideas.
And sometimes you feel likesomeone's ideas are being
confined by their defaultassumptions about how they're
going to generate them and youtry to give them a different
tool to work with.
Sam Gerdt (28:01):
I think that's a great example of what I see
lacking in any kind ofartificial intelligence for
quite a while to come, andpossibly even forever, it not
being possible at all.
Matt Saunders (28:16):
People sometimes ask me how do you teach art?
And I actually think that thatI don't know exactly how you
teach art, but I know that themodel of art education is a
dialogic model.
It's one of you making yourwork but then being confronted
with other opinions and othersuggestions, and it's always a
kind of crowdsourcing, andcertainly when people leave art
(28:38):
school, that's the big challengeis how to generate all that for
yourself.
So I always feel like what I'mdoing in the classroom is
modeling all the ways that onecan turn over a problem and all
the ways that one doubles downon something, and all the
techniques that I should havelearned to keep myself moving
and to generate ideas in thestudio.
(28:58):
Not that the people I'm talkingto will do the same thing, but
hopefully that they can have myvoice in their head and they can
build that for themselves.
Sam Gerdt (29:06):
I'm just not sure how AI challenges itself Right.
I'm not sure how it stretchesitself and tries to force that
model, breaking behaviorindependently.
I see artists and I seetechnologists and even just
normal people who are playingwith the technology.
I see them doing that andcoming up with really
(29:30):
interesting use cases anddifferent results, but I see the
vast majority of people justusing it as it's presented to
them, in the ways that aresuggested to them, and then what
happens is the fork in the road, the divide gets wider.
So those who are challenging itand really really pushing it
are the ones who are generatingtruly unique, novel, creative
(29:55):
outputs, and the ones who arejust using it are the ones who
are contributing to that mass ofunimpressive you know, hotel
art, thomas Kinkade type art.
I think the same idea applies toall uses of AI.
People who simply use thesystem, the systems that they're
(30:16):
given, they won't be standoutsin the areas of productivity or
effectiveness.
The people who create newsystems or subvert systems,
those will be the ones who excel.
So, as we look in recenthistory, I think the creation of
the internet is a goodanalogous example.
The internet is an incrediblypowerful tool and it has done
(30:37):
amazing things to change theworld, but the way that it
affects individuals is the sameacross the board.
I'm no different than you inthe fact that we have the
internet and it helps us to beproductive and effective, but if
you compare me to someone who,say, knows how to subvert the
internet, knows how to hack,knows how to program, knows how
to build new things, newarchitectures, new
(30:58):
infrastructures, that person hasa significant advantage over me
.
You can go even further back inhistory and see examples of
this as well.
In the 17th century, a fewmathematicians were
experimenting with square rootsof negative numbers.
They were breaking all therules of math and most
intelligent people thought theywere nuts.
Rene Descartes was the one whocoined the term imaginary
(31:21):
numbers, and he totally meant itas a slam.
Fast forward to now, and theworld as we know it couldn't
exist without imaginary numbers.
Most applied mathematics thatwe use in electrical and civil
engineering only works withimaginary numbers.
So the computer that you'reworking on right now or the car
that you're driving in, thistechnology that we're using
right now, could not exist ifthose rule-breaking
(31:43):
mathematicians had not dared toexperiment with something that
most people thought couldn't orshouldn't exist Before that
Leonardo da Vinci firsttheorized and then proved how
the circulatory system worked.
The popular belief of the daywas that the liver made the
blood and pushed it out to theperiphery, where it was absorbed
by the various tissues of thebody.
(32:03):
To discover the truth, da Vinciperformed dozens of human
dissections, many of which wereillegal.
He performed surgeries andexperiments on live pigs in
order to see the blood flowingin the body.
He studied the fluid dynamicsof water in rivers.
He constructed his ownartificial hearts out of glass
or wax.
Da Vinci's drawings anddiscoveries were a full 450
(32:27):
years ahead of their time.
So the question that I keepasking myself is if AI gives us
a system within which to operateand that system has rules, and
even if hekm memCH dies, what isthe means by which we improve
upon that system?
In all of human history, itseems, the means of improvement
was to break the rules, and isan AI equipped to do that?
(32:48):
I think what I feel, especiallyeverything that you've said,
talking about how you know youhave this gut instinct and you
have to, you know, constantly bechallenging, and all of these,
there's all of these littlepieces of what you're saying,
and I'm like, yeah, I just don't.
I don't see how an AI can everbreak out of the math of what it
(33:09):
is and make those leaps.
Hmm, but you know, maybe timewill tell.
I just there just doesn't seemyet to be a mechanism.
That part still seems to bescience fiction.
Matt Saunders (33:25):
I'm just nodding because I'm you've thought about
this a lot more than I have,and I buy that I will.
That sounds right to me ifthat's where we are.
I'm just not sure.
Sam Gerdt (33:38):
I don't know.
First, you see it Like a lot ofwhat happens is so far behind
closed doors.
All I see is all I see is whatthe world has like, what we're
given, and I see you know whatmillions of people are doing
with it generally.
And then I hear people like youwho have decades of experience
(33:59):
creating and teaching others tocreate and to participate in
these thoughtful conversationsthrough various means, including
art and all of the mediumswithin art, and I just I don't
see any of that happening.
When people take the tool thatthey're given and just apply it,
(34:21):
it's the ones who are like I'mgoing to take this and I'm going
to break it who are actuallylike finding anything creative
inside of it.
But it's them.
It's the act of, it's the actof breaking it, it's the means
by which they break it, it's thechoice to break it and then to
(34:42):
iterate on that process over andover and over again.
That seems to be generating anyamount of real innovation or
value, whether it's artistic ortechnological.
They're all agree with you.
Matt Saunders (34:52):
I mean, personally, I have not seen any
AI generated images.
That I was.
That I remember, yeah, but butI get, and it's crazy that I'm
now advocating for AI in somesense.
I don't want to predict thatthat's not going to develop.
I mean, I know, before thecurrent model, there was more of
(35:12):
an adversarial model.
There was the idea of havingtwo different programs, one
generating and one judging, andyou know, I think that there's
certainly ways that that thiscan get more and more
complicated.
This, those creators who startto work with AI maybe maybe AI
as a solo entity doesn't becomecreative for a long time, but as
(35:33):
a half of a kind of partnership, I could see it really being a
being a productive tool.
I do, you know, when you weretalking, I was thinking to the
artists who have centered AI asa topic in the kind of art world
that I'm tuned into, who Ifound most exciting and
provocative, have not beenworking with the tools that
we're talking about.
(35:54):
I think that this shift in thelast couple of years towards mid
journey or dally has created adifferent, has created a certain
type of beast that's based onmaking pictures quickly.
Yeah, isn't exactly where itdoesn't cover all of AI.
So I'm thinking of people likeIan Chang or Pierre Huig, both
of whom have thought more aboutthe idea of sentience in life
(36:17):
and kind of making artworks thatquestion whether something is
live or whether life can becontained within the space of a
video in Ian's work, or Pierrehas a a big piece in Kistifos,
norway, that is a forest thatwas scanned and then built as a
(36:37):
built as a digital environmentwith an AI God that's mutating
and creating variants and sothere's this overlay of this
kind of pristine wilderness andthen this mirror of it that
exists, guided by differentintelligence, and he's like
running these things and seeinghow they start to.
He's taking things that havebeen generated in the AI and
(36:59):
actually making them in reallife and putting them into the
forest, and it becomes this kindof a multiverse idea.
But those, to me, are not well,they're not an interesting.
I think both of these pieceshave made me think a lot about
our future, with potential ofliving minds and machines, much
(37:20):
more than any of the any of thepictures.
Sam Gerdt (37:24):
You talked about two, two AIs challenging one another
.
One of the things that thatthey do in in training AIs like
mid journey or Dolly or some ofthe the more common ones, is
they'll use human feedback toinform the the rightness, the
correctness of the response tothe prompt.
(37:45):
So that kind of human feedbackwas it turned out to be
incredibly important in in thetraining process in order to get
the results that that we get,and you still see this today.
You can, you can actuallyprovide your own feedback in in
all of these tools, whether it'san image generation tool or a
text generation tool like chat,gbt, and that feedback is
(38:09):
invaluable in improving theproduct.
The issue there that might beworth talking about is that you
have so many people providingthat feedback.
The baseline is quite standard,it's not, it's not deviating,
and so when you have artists whoare in some sense challenging
themselves or challenging otherartists, you I feel like there's
(38:32):
this distancing, thisdistancing from the norm, and I
feel like, with the kind ofhuman feedback that we provide
to an AI, all it's doing isconsolidating into a single,
almost a singularity of thenormal.
I actually kind of think thatthat's more of from a, from an
intellectual standpoint that'sprobably more of a of an
existential threat than even,like you know, whether or not
(38:56):
they can.
You know, cause a nuclearwinter is.
Is this idea that that ourhuman knowledge is going to be
reduced?
It's not going to be.
It's not going to be varied anddiverse, it's going to be
standardized, globalized Is isit not already?
(39:16):
I mean.
I think, it's more so than maybea hundred years ago.
I mean, technology, theinternet certainly has connected
us in ways that nobody everimagined, and so, yeah, human
knowledge and experience isdefinitely more globalized and
more standardized now than itwas, and there are good aspects
of that and there are badaspects of that.
(39:37):
But what happens if, if you goto the extreme with it and
there's only one?
There's only one answer.
It's not that it's right orwrong or that you know, it's
just that we've, we've trainedand we've trained and we've
trained, and now there's no moreroom for any more feedback and
and there's in some sense youwould have to have you start
(39:57):
from scratch, train somethingnew, but by that point you've
conditioned the people, and sothe people are thinking less
differently than they were ahundred years ago, and so now
when the people train the newthing, they'll be less capable
of training a diverse thing.
Matt Saunders (40:14):
It's very dystopian.
I think I have a better, a moreconfident view of human nature
in some regards.
I'd say not in regards to peace, but I have a lot of filmmaking
friends.
I mean also because I teach ina department that has filmmaking
and of course that's constantlybeen the conversation about the
(40:35):
Hollywood models and what'sallowed and how, how things get
made, and it's a question ofmarketplaces, it's a question of
economies as well.
I think there's always, alwaysgoing to be a lot of people who
want to do the opposite, and soit's more question of.
To me, it's not a question ofwhether that creativity will
emerge or exist in humanimagination.
(40:55):
It's like are we going to havea society where it can be made?
So I don't know.
I spoke earlier about you know Isaid it puts pressure on
language.
That's my strongest reaction alot of times, looking to AI I
mean even some of the articlesyou sent over for me to read,
you know is one where it's askedto make a Renaissance painting,
and I'm looking at this imageand thinking there's zero in
(41:18):
this image that has to do withthe painting.
You know that this is like it'san idea of painting that's
completely traveled through acultural flattening into a style
that is recognizable to thecomputer's painting, and it's
that word is is totally undoneand even the monk you know
there's Kermit the frog is monk.
It is zero relationship, Iwould say, to the things that
(41:42):
were made by the historicalperson, edvard Munch, but has
100% relationship to the posterof the screen.
And so, really, you know, myexperience of AI is both
complete irritation combinedwith the kind of provocation,
the kind of sense of how, howideas and language and styles,
how visual ideas move into verybroad understanding and get kind
of misrepresented or gettransformed, and so forces me to
(42:05):
grapple with thinking aboutwhat is Munch from my
perspective, having just gone tolook at a ton of his work in
the last couple of weeks, orwhat is a painting?
And I bristle, of course, Ithink, oh you idiots, that's not
a painting at all.
That is the flattest idea youcan imagine it.
But then it's up to me torespond to that.
(42:25):
You know, we're going to livein this world and and you know,
and if it forces me to beoutraged and feel like people
are misunderstanding, maybe thatpivots my work to engaging with
those feelings more, versusjust assuming that we're all
communicating and that's, for me, the silver lining is that it,
like things that have happenedin all kinds of pop culture.
(42:46):
I think it has flattened, but italso creates a space for a
different type of of weirdness.
On top of the form.
It just becomes a differentpoint of reference or kind of
touched on it Does it?
Does it piss me off when peopleare sharing images that I think
are absolutely banal and sayinghow amazing they are.
(43:07):
Do I think?
Don't be an idiot, come to themuseum with me.
Yes, of course I think that,but I don't know what else we
can do.
It's how culture is all.
Sam Gerdt (43:17):
You're highlighting the exact same feelings that I
have in different areas.
You're highlighting how youlook at an AI-generated image
and there's an immediaterecognition of what it is and
you're highlighting the factthat, in that recognition,
you've already made a decisionabout it and you're already
moving on, and there's a sensein which a person can make
(43:43):
themselves even less distinctthan if they did nothing by
generating the AI image.
It's like you've justclassified yourself in some
sense in my mind, because Irecognize immediately what
you've done.
There's nothing creative aboutit.
(44:04):
It falls into the exact samecategory as every other
AI-generated image I've everseen.
And then I'm so quickly movingon.
I'm ready to move to the nextthing, and I'm not necessarily
saying that that's exactly whatyou just said, but I have felt
that way.
Certainly, and going back towhat I said earlier, that you
(44:25):
used words that I think betterdescribe how I was feeling
cultural, flattening this ideaof removing the diversity and
variation that exists withincultures and giving us all just
a baseline understanding of therightness or wrongness of
something, the goodness orbadness of something.
(44:46):
If we look at life 100 years agoversus life now, I feel like
the disconnected nature ofhumanity, the localization of it
resulted in wildly distinctopinions, worldviews, art, all
(45:08):
kinds of distinction.
And the more that you exposeyourself to all of the rest of
the world and everyone elseparticipates in that as well
that distinction starts toflatten, it starts to diminish.
So I guess you could phrasethat as a question what are some
of the differences that you seebetween an artist who is a more
(45:32):
modern artist, who is exposinghimself to all of the world, all
of culture, who's absorbing allof that, versus, maybe, an
artist from centuries past whowas very local, his experiences
were very limited?
Do you see those differenceswhen you look at their work?
(45:52):
Do you feel those differences,or do you interpret it all from
your own perspective?
Matt Saunders (45:58):
Of course to see those differences and of course
I interpret it.
I feel like we can't undo theinterconnected nature of the
world and I think that we'reprojecting, maybe backwards, a
little bit of difference onthings that feel from a very
different time.
You know, if you think aboutwhatever, like Baraceli in
Florence is living in a Catholicmonoculture with near
(46:22):
parameters, never left his sidegoing to Rome once.
He never left Florence.
There's a flattening to that.
Perhaps Now we see it inrelationship to other things
that are happening.
So we don't see it as aflattening.
But I share your anxiety, butnot maybe your pessimism.
Sam Gerdt (46:45):
That's a pretty common response.
Matt Saunders (46:47):
And yeah, I, I am not.
I am someone who has A mixedrelationship to kind of the
biggest kinds of pop culture.
I'm a little out of it and I'veworked in my own work kind of
against it.
I've tended to be moreinterested in unearthing stories
and kind of digging into thearchives and countercultures.
(47:07):
But that's more who I am.
That's not a claim thateveryone should be like that.
Growing up in the sort of late20th century, the art history we
were taught about painting wasalways about the crisis of
photography, the idea thatphotography comes along and this
is, of course, a very flatdescription of this idea but the
photography replaces therepresentational function of
painting and that opens up acentury of abstraction and
(47:32):
innovation and crisis.
It's more complicated than that.
But if we if enough people arefeeling a crisis of flattening
out of AI, then there's a reallyinteresting crisis that we can
work with.
I don't think that we'retrapped.
We're trapped to live in aworld where it's going to be
full of these images.
Yes, that's going to happen nomatter what, but I don't think
(47:53):
that as artists we're not ableto push back against that.
If that starts to feelsuffocating, I found myself
thinking really quickly aboutnumber of artists who emulate
popular forms, like artists whotry to become pop stars, or the
large number of artists who getsuccessful and then try to make
a feature film and there is thiskind of artistic creative
(48:14):
impact to infiltrate or to actthe culture.
So I wonder, if an artist actsas AI, how that, what that space
might be at some point, likeimpersonating Dolly, hacking
someone's system and never mindCut that part of the video.
Sam Gerdt (48:32):
I know Matt felt like he was derailing the
conversation with thisadmittedly silly idea, but here
I am in the edit a few weekslater and I cannot help but
Google this idea of humansimpersonating AI.
So, without getting toosidetracked, here are two
interesting examples that Ifound.
Last summer, several creatorson TikTok started trending for
(48:55):
their NPC live streams.
They would stand in front ofthe camera with their hands held
awkwardly in front of them andsway back and forth like an AI
video game character.
Viewers could pay money to sendvirtual gifts to these
streamers and the streamerswould react to the gifts in the
same canned way that you'dexpect from an AI avatar.
It is absolutely bizarre.
You can go on TikTok and findthese people still doing this,
(49:21):
but apparently it's wildlypopular and the whole thing
reminds me of the living statuestreet performers in New York
City.
The second example I found isfrom 2018 and it's a startup
strategy and it's called pseudoAI.
There were several casesreported where startup companies
would present themselves toinvestors or customers as AI
powered companies, but theprocesses that were supposed to
(49:46):
have been AI powered wereactually offshore contractor
powered.
The idea was to reduce or delaythe initial development costs
of implementing AI poweredservices, so, by taking
advantage of cheaper offshorelabor.
Startups could more efficientlyprototype their processes.
Once they scaled, they couldhand things over to the AI
systems.
(50:08):
What I found interesting aboutthis example is the people that
they interviewed for thearticles that I found who were
doing this work were expressinghow much they hated it, how
mundane it was, how much theycould not wait for a bot to take
their job.
So in answer to Matt's questionof what would it look like, I
kind of think imitating AI,imitating the computer, looks
(50:31):
boring, it looks bland, it looksunappealing, and I think that
says something about theconversation we're having here.
The experience of being humanseems to me to be far better
than the experience of being acomputer when you look at the
(50:56):
students coming in today the artstudents coming in now and you
compare them to maybe when youwere first starting as a student
.
What are some of the biggestgenerational differences that
you see when answering the samequestions?
Like you know, what is?
(51:16):
Creativity, what is art?
Matt Saunders (51:17):
That's a really good question and there are
definite differences.
There are differences about theassumption of what art does and
where it lives.
Certainly, I have one reallygreat student in particular
whose highest goal is to be onInstagram and those are
fundamentally different ideasabout what you're actually
(51:39):
producing and what's circulating.
A sense of what rolemateriality plays in a work is
quite different.
I think the thing that I quotethe most is something that a lot
of artists quote I'm not evensure if it's possible, but it's
always attributed to Ed Ruschaand it's this idea that some
work you look at and you say wow, huh, but that he loves work
(52:02):
that you say huh, wow, you know.
That made so much sense to mewhen I was told that, because I
believe that work should bedifficult and should open up and
should be inscrutable andshould be complicated to read,
and I think that thatfundamental sense of values has
shifted a bit.
I think that the legibility,efficiency, speed of consumption
(52:22):
, delivery of a message, what Iwould call the kind of more
illustrative quality of work,which was a very pejorative term
when I was in school, is now apositive for a lot of people how
quickly we consume these images, how quickly we read the thing
and judge it.
Sam Gerdt (52:40):
I hadn't thought too much about this shift in favor
of the illustrative until Mattmentioned it, but it's something
that I also recognize.
It's an evolution that makessense when you consider how our
world has changed over the past30 years.
We live in an attention economy, in a digital world.
Now.
The demand for things likegraphic design and digital
illustration has skyrocketedsince the mid-90s, and that
(53:03):
demand tugs on everything.
The way we look at art ischanging too.
The pace at which we consumeand demand new content is
mind-numbing.
Compared to previous decades,we look at art on four-inch
screens instead of a canvas.
In having a conversation aboutcreativity, it's hard to ignore
(53:23):
the fact that our collectivedefinition of the word itself
and the standard by which wemeasure it has changed radically
, and I would argue that it'snot changing for the better, but
for the worse.
Matt Saunders (53:38):
And there are ideas that come up around the AI
question, around rightness,around function, around
efficiency that I think, forsome of my generation, raised
some hackles.
In relation to the creativity,I fundamentally don't think that
art is problem solving.
I feel like it's makingpropositions which may or may
(53:59):
not mean anything to anyone, andif they mean something to
someone, then they become agreat proposition.
But I don't feel like I'm everI mean so consciously on some
level.
You, of course, are, but Igenerally don't feel like I'm
judging my work in terms of doesit do what it should do?
It's more what did it do and isthat something worth keeping?
(54:19):
Is it a hope that it arrives atsome provocative, unexpected
state?
Maybe that's the jump over thefence that I started out by
talking about as creativity.
If it's not what I expected itto be and it seems better, then
that's the work that I cherishand put out into the world and
(54:41):
that's kind of my approach tomaking.
But I think that I take veryseriously that I don't.
That doesn't seem like animmediately shared value.
Sam Gerdt (54:52):
As an artist yourself , your work tends to be a
mixture of material and digital.
Yeah, mixture of physical anddigital.
You definitely do a lot ofexperimenting with both.
Do you have a?
Matt Saunders (55:05):
favorite?
Oh, 100%.
The digital is secondary andcomes in almost in terms of its
flexibility in showing the work.
Most of the digital things I doare digital videos which are
made out of real.
The work is always made withreal materials and then scanned,
(55:26):
and I'm very conscious thatbetween the scan and the
experience of the installationor the projection in the gallery
there's a lot of things you cando in the digital space.
That's a kind of digitalmateriality and I do play with
that and experiment with that.
But everything starts on thepage.
It starts with, in my case, abrush of some kind, because I
(55:47):
hate pencils.
Sam Gerdt (55:47):
But yeah, again it's.
Will an artificial intelligenceever express anything like that
?
I mean even just like I alwaysstart this way because I hate
that.
It's like, well, wait a minute,what do you mean?
You're gonna have to explainthat to me.
(56:09):
It's that, it's that diversity.
I think that that makes it so,so appealing, anything so
appealing.
And as soon as you, as soon asyou remove that element, I'm in
marketing now, and so, like,what is marketing except to pull
(56:30):
on an emotional string?
I just don't.
I don't see how we are doingany better, accomplishing that
by being overly dependent on onthese tools, when all that they
do seems to be reductive, andmaybe that's like you said, I
tend to be a pessimist withthese things.
Even you talk about thegenerational differences, your
(56:53):
students.
I've talked to other professorsbefore and they've said exactly
the same thing.
I have, I have, students whosegoal, whose aspiration is to be,
is to be, on social media.
I bristle at that because itseems to me to be foolish in so
many ways.
But the flip side of that is,I'm also belonging to a
(57:16):
generation who just doesn't getit in the same way, and that's
why I asked about, you know, thegeneration that you're teaching
now it seems like they're.
They're living more of adigital life, and so the the
conversations that they'rehaving are going to be digital,
and the the ways that they'rethinking and the issues that are
(57:38):
affecting them are going to bedigital.
I, I'm, I'm with you.
I much prefer the material, butI also don't have any interest
in living in that digital world,in that digital space there's
such a diversity of spaces inwhich to live.
Matt Saunders (57:55):
Yes, it's would seem horrendous to me to be to
be big on social media, but Iextend that also to the idea of
genres and and something that Ialways thought about visual art
is that it's behind a lot ofother types of art making, like
within, you know, within music.
(58:16):
We really understand thediversity of different types of
music and what genres differentthings fit into, and it's
absolutely commonplace to say Ilike this and I don't like this,
but we do seem to often havethis kind of fatalitarian
impulse to define all the art ofa time and I've never seen it
that way.
There's so much art that Irespect and I respect people's
(58:40):
desire to want to do that, but Ihave absolutely no interest in
making that kind of work.
For me at the moment, ai isvery much in that category.
But the world changes and new,you know, new genres come and
new places to live come and I amcurious about next generations
(59:03):
and what options they have andand I'm trying to understand it,
which is maybe why I'm soopen-minded right now you talked
about the flatness ofBotticelli, living in Florence's
whole life, hardly ever leavingexcept for the one time, and
how.
Sam Gerdt (59:16):
That is a flat existence in its own way.
And how, the the differencesyou know.
From our perspective, we see.
We see that existence with allof the.
You know thousands and millionsof other existences and we can
appreciate each of them fortheir diversity.
Looking at all of them togetherand how, now, with the, with a
(59:39):
more global view, global mindset, there there seems to be a
global flattening in some sense.
It seems like there's a wholelot of value in understanding
where you live and where youbelong and what your particular.
As individuals we can't livethese incredibly diverse lives.
We are going to have our ownflatness to us, as you know,
(01:00:02):
beings living in a particularplace at a particular time, with
particular beliefs andexperiences.
There's flatness in that.
But understanding where youbelong and then staying there
and developing it and developingthat conversation with other
unique individuals, whilemaintaining your integrity, that
seems to be to me the goal ofany artist today to know who you
(01:00:27):
are and and what, what yourpurpose is and and what your
belief is, and then taking thatand interfacing with other
people who are doing the samething.
The threat from AI that I thinkI've talked about could be
restated as the antithesis ofthat idea.
People know less of who theyare and where they belong.
(01:00:51):
They know less of theirindividuality, their uniqueness,
and they see themselves asbeing just one tiny piece of
this big, more fists blob ofpeople and culture and just a
mishmash of everything.
Yeah so, yeah, so I wow.
Matt Saunders (01:01:14):
It's such an interesting thought yeah, so
dark, I mean it's so dark andwith you on that.
I think it's a well, this is no.
Sam Gerdt (01:01:29):
I was just gonna say this has been one of the
weirdest conversations about AIthat I've ever had.
Matt Saunders (01:01:36):
I feel like I leave with more questions than I
came in with and I keep comingback to the, to the film idea
that came up and and the sort ofthe history of of that medium.
Since it's relatively short, youcan look at it and think about
the way that studiosconsolidated the ability to make
(01:01:57):
films and you have an era thatgoes through all the studio
codes and then there's a momentwhen the tools change and indie
cinema becomes, you know,accessible and then that gets
consolidated and you know we'reliving in a moment of kind of
consolidation around televisionand film, but also in opening up
(01:02:19):
of all the tools, including AI,that can make movies.
I have we haven't talked aboutthis, but I do have I've
encountered film students atHarvard who are using AI to edit
their clips and even generatethings, and you know, I think
there's some curiosity aboutthat, but that's part of where I
have hope about this is.
I feel like right now, all our,all our anxiety is up and all
(01:02:43):
our eyes are focused on one ortwo big platforms, and it seems
very flattening.
But I do.
You know, I have to assume thatthings will come along that
will change those dynamics andcreate radically different
versions of AI creators andmaybe a kind of underground and
(01:03:05):
off, off cycle, and I could betotally wrong.
But there's the basic questionyou started with about
creativity, but for me thatimmediately gets into questions
of the marketplace and the toolsof production, the means of
production for that, and yeah,and I think that we're at the
real infancy of accessibilityand and use of AI.
Sam Gerdt (01:03:29):
So hopefully goes and hopefully people mess with it
in the right ways yeah, well,for my part, I, I put my, I put
my money on the artists and theinnovators and the technologists
who are doing that breaking,who are doing that challenging
(01:03:50):
and who aren't, who areresisting the urge to just join.
You know, go go with that flowof of where these tools might
take us, especially in theirinfancy, because, over and over
and again, again, I feel likethat gets validated in
everything that I see the oneswho are actually adding value to
the world or the ones who are,who are trying to break AI who
(01:04:16):
are?
trying to recreate it intosomething different than what it
is, not the ones who are usingit, right?
Yeah, yeah, well, that is aninteresting discussion.
I love that, matt, I thank youfor.
Thank you for talking to me.
(01:04:37):
It was, it was a real pleasure.
You are an incrediblyinteresting person to talk to.
I love your ideas and maybemaybe in time we reconnect and
see how you know how we feelabout what we've said, how we've
changed, how we've been changedand, and at some point I'm sure
(01:05:01):
we're both going to have toengage with AI a little bit more
, break some, break down some ofthose walls that that we have,
those those resistances thatexist, and and I I would love to
see, like, what happens withthat yeah, yeah, no, thank you
very much.
Matt Saunders (01:05:20):
It was really I'm .
It was surprising that youreached out and I've really
enjoyed the conversation.
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