Watch the video on YouTube

Season 2 Episode 1 – Can AI understand Art?

Show Notes (with AI assistance)

Artificial intelligence is transforming creative industries faster than almost any other technology in recent memory. Artists, musicians, authors and designers are asking difficult questions about ownership, originality and what happens when machines learn from human creativity.

In this episode of Data with Duke, we explore one of the biggest debates in AI today:

Can AI genuinely understand art? Is artistic style simply data and patterns? Or is there something fundamentally human that machines can never replicate?

Along the way we look at fascinating research showing how machine learning can identify Renaissance artists with astonishing accuracy by analysing microscopic brushstroke patterns—effectively treating paintings as biometric signatures rather than images.

We also explore:

why artists are suing AI companies the rise of copyright and licensing battles how modern creators are using AI as a creative partner rather than an enemy why mathematics and art have always been closely connected what history teaches us about technological disruption

Finally, we ask the biggest question of all:

If an AI can perfectly reproduce a master’s brushstroke, but only a human lived the experiences that inspired the original, where does the true value of art actually lie?

As always, this isn’t about fear or hype. It’s about looking at the data, understanding the technology, and thinking critically about where creativity goes next.

Links

![ABC News: Judge approved $1.5bn Anthropic settlement]

![Reuters: Thomson Reuters wins AI copyright ‘fair use’ ruling]

![CWRU Newsroom: Using artificial intelligence to tell art apart]

![CWRU Newsroom: Innovative Artificial-Intelligence Tools for Solving Art Mysteries]

![NVIDIA: Neural Network Pinpoints Artist]

![ACM Digital Library: Bridging Art and A]

![Discerning the painter’s hand: machine learning on surface topography]

![refikanadol.com/]

![Music Business Worldwide: Sony Music sues Udio again]

![The Vitruvian Man]

Transcript

Recording Date: July 26, 2026

Hello everybody, how are you? Oh my god, that’s right in my face, I’m going to have to fix that one.

Listen folks, it’s great to see you, and it’s great to see me because you’ve never seen me before—or possibly not! So listen, I am awful at recording these podcasts and these thoughts, and it’s something I want to do, and something that I’m going to keep working on. But I thought, you know what, let me try and relaunch again with a little bit of a rebrand and a face. I have a face. That’s what I’m doing.

Anyway, I’m going to try and record these maybe once a month, try and be a little bit more realistic with my time. But today I’m actually going to be coming up swinging, because I’m going to be talking about quite literally a fight that is so wild, so petty, but also actually quite incredibly high stakes if you think about it, that you actually couldn’t make it up if you tried.

What I’m going to be talking about today—which is quite current today as of recording on the 26th of July 2026—is a debate that we have not yet seen the end of yet. That’s what I think we’re going to be talking about. People are going to be coming back and watching this video, I hope at least, in a good couple of years’ time, and people are still going to be having the same arguments.

What is it that I’m talking about? Well, of course, this is the great, messy collision of raw data and human creativity.

The AI Copyright Battleground

So let me be specific. Right now, if you are an artist, a musician, or a graphic designer, I completely forgive you for looking at AI with a mix of—let’s face it—absolute fury, but also a little bit of terror, really. And honestly, that’s fair enough. We are seeing a debate that is raging across the economic, political, and creative landscape right now about what AI is doing to potentially damage some of these industries.

And listen, look at this! Look at this. Look, I’ve got tricks—keep an eye on this.

So this screen that we’ve got behind me, we have authors, artists, and musicians who are being granted legal payouts right now, where these wonderfully creative people are starting to sue these big tech giants (this one in particular I think was Anthropic), demanding what they call “Learnright” laws. Now this is a term I’ll be coming back to: Learnright laws. Basically declaring war on scrapers, the bots, and the things that are trawling the internet right now to help train all of their various models. I don’t think that’s anything to do with it, but it’s quite a bit of a war zone out there at the moment, and these cases are hitting court: $1.5 billion in a copyright settlement against Anthropic!

So my question for you today—just a rhetorical question, just to think about right now—is to think about: how does a machine see art in the first place?

Do we, when we look at a piece of art, even if it’s on a computer like we are today, look at pixels on a screen, or are we looking at something way, way, way deeper than that? And if we are looking at something deeper… let me just be very transparent with this: I’m not a massive art fan, but I respect the hustle, I suppose, from historical art all the way back to the Renaissance and through to modern art. And let me face it, my eldest son lives in Bristol, so we get to see Banksys and all that kind of stuff as we’re going through.

So listen, just before anybody asks me, I’m not an art critic. I don’t know a huge amount about art itself, but I do have the capability as a human being—and this is the important point—to be able to look at a piece of art and to feel something, to have that sort of reaction.

And the question that we’re going to try and cover off today is: can a machine do that? Can a machine look at a piece of art and understand the nuance that has created this particular piece of work?

Microscopic Art Forensics

Today I’m going to try and actually go quite microscopic. I’m going to look at how data scientists are using AI almost as an art history detective to solve 400-year-old cold cases, and how artists are actually striking back by using raw data as their new paint.

So let’s press some buttons and let’s have a look at some cases.

This is a real study out of Case Western Reserve University. When you and I—maybe most of us—look at a classic oil painting, we see color, we see subject, we see possibly emotion. But these absolute madmen in the physics and data science departments looked at a 400-year-old Renaissance painting and said: “Do you know what? That’s not art. That is a 3D elevation map.”

So bear with me for a second if you knew about this. There is a tool that was created and used called a 3D optical profiler. And what it’s actually doing… the science basically is that when an artist like El Greco or Rembrandt pressed a brush to the canvas, what they left was a physical, mechanical signature.

Think about this:

The thickness of the paint
The speed of their hand
The angle of the wrist
The stiffness of the bristles
The type of hair in the bristles

All of that got frozen in time on a canvas and immortalized into this piece of art. You might look at this and think, “It’s art, of course that’s what happens,” but when you start looking at it with a physicist’s view and a mechanics view, what we’ve actually got here is a 400-year-old physical biometric thumbprint. It’s almost like the kinetic energy of a dead genius trapped in dried oil.

And here’s where the data starts to get really, actually quite sneaky.

The Algorithmic Detective (95% Accuracy)

Back in the day, master painters didn’t really work alone. They had workshops; they had a bunch of very highly skilled apprentices who actually painted almost exactly like their master to churn out more work. So art historians have been arguing for centuries, really: Did Rembrandt paint this face, or was it some 19-year-old intern named Hans who did it instead?

So the researchers at Case Western Reserve University took these microscopic 3D scans and chopped them really nicely into these tiny, tiny little patches. We’re talking like 5 to 15 millimeters wide, really. You can’t see color potentially, you’re just looking at a tiny patch of just grey or something like that. But they fed all of these images into their Convolutional Neural Network.

They actually managed to work out 95% accuracy on this model of identifying a particular artist having painted a particular painting!

There are art students who have been studying for years who cannot look at a single piece of art and accurately—certainly not 95% accurately—identify a particular artist unless they know the piece of work for sure. But this artificial intelligence system was able to do it. The AI, looking at a patch of texture just smaller than a ladybug, knew exactly which artist held the brush.

So this leads us to a bit of an existential crisis, really, because we then need to start asking ourselves: Is artistic style some magical, spiritual, human spark? Or actually, is it something highly predictable, mathematically consistent, and just a pattern of physical friction?

Don’t answer that one yet! Don’t jump to the comments yet, because I know some of you are going to do this. But it gets even better when we start thinking about how modern artists are starting to use this concept—this mathematical consistency captured on canvas—to their advantage.

Data as the New Paint

Listen, just before I go any further, let’s have a look at the flip side, because we’ve all seen AI art. There are lots of big tech giants out there at the moment using terms like “AI slop,” including CEOs and CTOs of YouTube talking about how they want to remove low-value AI videos from their platforms. We’ve all probably seen them—like somebody types “cute cat wearing a spacesuit” into a prompt box, and it spits something out that gets uploaded to YouTube. That’s something we just accept to see nowadays.

But think about what’s happening behind me when I load this image right here. This is real-time generative art. This is generative art literally being generated day in and day out by genuine artists—people who have a view, who have an eye for this kind of work, and understand what the components are of a valued piece of art.

Some modern artists are actually starting to treat this data as paint, as their canvases, as their paintbrushes and equipment. Some are taking:

Live weather feeds
Global seismic activity
Internet traffic
Even their own heartbeats

And they’re feeding that data into custom AI algorithms to create digital sculptures that move, react, and change live!

So as for the copyright wars… well, the smart artists of 2026 aren’t really trying to ban the tech anymore. What they’re trying to do is actually reclaim the code.

What we’re seeing is a rise of creative licensing, where artists are saying: “Listen, if you were to train your models on our creative work, that’s fine. But my physical brushstroke patterns, my strum of a guitar, my synthesized work is mine. If you’re trying to train your models using it, that’s no problem, but you’ve got to pay me royalties in exactly the same way an artist receives royalties for music used in a film or TV series.”

Fine, you can use it—just pay me. And this is what’s starting to happen now within generative AI.

Math, Art, and History

It’s like history always rhyming. We like to pretend that math and art are sworn enemies, but here’s the big thing, the mind-blowing thing: math and art have actually always been married since day one!

Leonardo da Vinci was obsessed with math and physics. He wasn’t just a painter—he was an inventor, a physicist, and a mechanic.

Johann Sebastian Bach basically wrote his musical fugues as complex, symmetrical mathematical equations.

The Invention of Photography: Despite what some people claimed at the time, the camera didn’t kill painting! It moved into giving us Impressionism, because painters suddenly didn’t have to worry about capturing reality anymore—the camera did it for them. Art actually moved on as a result of the invention of photography that people were insistent was going to kill painting.

Look at The Vitruvian Man by Leonardo da Vinci—a massively famous piece of art that has much deeper meanings than the drawing itself. The geometry of the human body is what da Vinci was trying to highlight here. This is a mathematical representation of people.

This happens so often through history where a new invention comes along, threatens a piece of technology or culture, and there will always be people who say, “No, this is evil, this must stop.” In da Vinci’s time, it was “against God,” and people were demonized and removed from communities because they dared to adopt new methods. And the possibility is that we’re actually seeing this with AI art at the moment.

Conclusion & Final Question

Technology has taken a huge leap, forcing humans to be more abstract, to be more experimental, and let’s face it, more deeply human.

A machine can map a stroke of a brush, it can copy it to the nanometer, and it can guess an artist with up to 95% accuracy… but a machine will never know why we picked up the brush in the first place.

So this is my conclusion: the soul of art—whether created traditionally using paint and canvas or via a few prompts in an AI engine—isn’t in the output. It is in the human intention behind it. It’s us trying to say: “I am here, and this is how I feel right now.”

So here’s the question that I’m putting to you in the comments:

If an algorithm can copy a master’s brushstroke right down to the molecule, but a human had to live, struggle, and dream to create the original, which one actually moves you?

Let me know what you think down below, smash that subscribe button, hit the bell icon, and I will be back in about a month on Data with Duke. Bye!