Computer Created A 'New Rembrandt' After Analyzing Paintings (bbc.com)

TechnoidNash quotes a report from Techie News: Rembrandt van Rijn was one of the most influential classical painters, and the world lost his amazing talent when he died nearly four centuries ago. And yet his newest masterpiece was unveiled only yesterday. How? By scanning and analyzing Rembrandt's works, a computer was able to create a new painting in near-perfect mimicry of Rembrandt's style. It has been named, appropriately, "The Next Rembrandt." The computer used machine-learning algorithms to create the portrait, which was then 3D-printed to give it the same texture as an oil painting. "The Next Rembrant," was a collaboration between Microsoft, ING, Delft University of Technology and two Dutch art museums -- Mauritshuis and Rembrandthuis.

Comments

What was that shit site?

[xxxJonBoyxxx]

Gimme the pic, let me look at it for the 3.4 seconds I'm interested in it, and let me get the hell out of there. The site linked from TFA (www.nextrembrandt.com) felt like a throwback to the days when people actually built "sites" with Flash. Yeech!

Re:What was that shit site?

[Tough Love]

...felt like a throwback to the days when people actually built "sites" with Flash...

It's not nearly as bad as that.
  * browser navigation keys work (doesn't get stuck in a separate mode)
  * Text can be selected and copied
  * No proprietary player brimming with exploits

Agree, it doesn't cater to the ADHD set - you actually have to play the video, or at least skip through to the end like I did - to test your art connoisseur quotient, or scope the tech out for counterfeiting potential, or whatever your immediate goal is. You have to admit, it's a slick html5 demo, and not without taste.

It also breaks the internet in various ways just like flash, for example, no deep links, you can only bookmark the top level site. The only way in is through the art show. I sure don't want the internet to end up like that, a video internet just isn't for me. But for announcing an art research project? OK, fine. At least it's not flash.

Re:What was that shit site?

[Anonymous Coward]

It's worse than that: It bitches about my browser being too old or hey, would I like a windows binary of chrome? (This is a linux binary of chrome, yech.)

Anyway the painting reminds me of a story: $GUY's father in law has a Mondriaan on the wall, and as a prank he copies the painting himself. "Funny" says father in law, walking by the painting and unaware of the swap, "always thought it had a special something, but that's gone now."

"Analyzing" and coming up with a "new" painting in the same style? Sure, good exercise for a human. Having "deep learning" do it? Bit of a tech wank. Especially so with grandiose claims attached. Yet it can't copy what makes great paintings great. Much like we can copy all we want but we won't get violins quite as good as Stradivarius' violins.

And Delft University Of Technology? They went all business-like, converted their internal IT all to microsoft, did away with the department subdomains because "too hard" in exchange, "standardised" on three kinds of supported peecee, all microsoft, and a number of similar follies. Fun fact: The IT department didn't care about no longer being able (or allowed) to run non-microsoft. The physics department, however, turned out to have a bunch of Expensive Machines that run on Unix, and now they're "allowed" to take care of those themselves again, after the departmental IT departments were cut away.

So now they're doing tech circle jerk "research" together with microsoft. Doesn't surprise me. At all.

Re:Nope, sorry

[kheldan]

Yes, but technique is only part of being an artist. There are plenty of artists out there who can create incredible knock-offs of famous paintings, but that can't create an original work of their own that anyone has any interest in; you can be technically talented, but have no talent for creation.

Re:No, it didn't.

[Calydor]

But can the AI itself then be considered art, as it has been made to make people observing it in action feel awe and astonishment at what it is doing?

Re:No, it didn't.

[alvinrod]

I suppose code could be considered art. After all Windows has elicited great feelings of despair and anguish from those observing it.

Typical, average, math

[holophrastic]

It's funny. This research spent a great deal of effort to identify the artist's "typical" approaches to individual and compound features; it then averages them to produce a work that feels like a Rembrandt.

But the only reason that it feels like a Rembrandt at all is because it is the centroid of all of his stylistic approaches -- a perfect average.

As in most cases, an "average" of many details is precisely not what an artist does. An artist's real work is in the details that defy their own averages and typical approaches. Listen to any artist analysis, and you'll hear words like "unlike in his other works...", "for the first time at that point..", "never before...", "...and yet in this painting...".

This work is very impressive, a perfect way to fool viewers and a perfect way to pay respect to Rembrandt's approach. That said, however, it is precisely the definition of not a Rembrandt. It is not the work of an artists. It is the work of a business -- which has always been the ability to reproduce copies of something (product or service) in a replicatable and bulk manner.

Re:Typical, average, math

[AthanasiusKircher]

It's funny. This research spent a great deal of effort to identify the artist's "typical" approaches to individual and compound features; it then averages them to produce a work that feels like a Rembrandt.

But the only reason that it feels like a Rembrandt at all is because it is the centroid of all of his stylistic approaches -- a perfect average.

That's one way of looking at it. But actually the methodology as described in TFA was more to combine thousands of fragments of other paintings to create a composite.

I wonder if the result is therefore a little like the musical "compositions" created by the algorithms of David Cope. Cope has been doing this procedure since the 1980s where he has a program that takes dozens or hundreds of musical works by a particular composer in a particular genre as "input," and then claims to "compose" new works in that style.

To those who don't really know much about a Chopin Mazurka, for example, the result seems pretty amazing. But to those who actually know all the Chopin Mazurkas (there aren't that many of them), the listening experience is a bit surreal, since the "new pieces" sound more like odd mashups of bits of the old ones, often joined haphazardly together in ways that don't quite make sense.

I wonder if an art historian might have a similar take on this "new Rembrandt" when examining it closely. From afar, it certainly seems to have the general style of the original artist, but perhaps the details are an odd mashup whose "joints" don't quite blend well.

On the other hand, if they managed to make those "seams" invisible so the whole thing looks coherent, that'd be a much greater achievement. Also, it's unclear how much tweaking was involved in this project -- Cope was known to listen to hundreds of the "AI composed pieces" and himself choose the best of the lot to exhibit... presumably the few that actually made the most sense.

The project is a work of art, not the output.

[dsmatthews9379]

What they have done is Art, but what they have produced is just a beautiful artefact. It is the act of pushing the envelope in visual communications that is the artistic act, the Next Rembrandt is just the evidence, in the same way a video recording is the evidence, a document, that records a performance art event.

Re:No, it didn't.

[Tough Love]

Because art is all about the artist attempting to elicit an emotional response from the person experiencing their work, and since no so-called 'AI' has actual emotions, it can't understand art, and therefore can't 'create' art.

Even if I accept your narrow definition of art as "elicits and emotional response", I can't accept your non sequitur: having emotions is not a precondition for eliciting an emotional response. Just consider a flat tire to know that.

Were are all the Rembrandts?

[rsilvergun]

Something I often wonder is why we don't hear about 'new' geniuses in art. It's always the same guys from the 1700 and 1800s (once in a while an author from the 1900s sneaks in). Maybe they're all doing hard science now that it's a thing.

Re:Were are all the Rembrandts?

[AthanasiusKircher]

Something I often wonder is why we don't hear about 'new' geniuses in art. It's always the same guys from the 1700 and 1800s (once in a while an author from the 1900s sneaks in).

Uh, well, you might want to have a look here, where there are links to probably a couple hundred painters of the 20th century who are regarded as significant. There are many people in the art world who consider a lot of them to be "geniuses."

Now, your objection might be: "Hey, most of that stuff looks like crap compared to Rembrandt. Where's the good stuff like what he did?" Certainly many people feel that way about various artistic movements of the past century or so.

And for that, you'd have to blame German Romanticism of the 19th century, and "the cult of the genius." It's interesting that you use that word "genius," but artists weren't always seen that way -- Rembrandt wouldn't have been referred to that way in his lifetime. Artists used to be viewed a craftsmen, with particular skills -- some more talented than others. Centuries ago, being a painter meant learning the latest skills -- paint and materials weren't as good as they are today, so you needed to exploit the technology they had. Knowledge of perspective or the ability to manipulate and represent light and shadow took time for artists to figure out -- and so there was a gradual evolution toward the ability to create more realistic representations. The best painters were the ones who had incorporated this new knowledge to hone their craft.

That all started to change with the "cult of the genius" and the "Aesthetik" movement that began in late 18th-century Germany and spread during the 19th century. The emphasis went away from craftsmen participating in guilds or groups of people with knowledge, and more emphasis was put on the individual creator ("genius") who was supposed to demonstrate individual expression.

All of this is to say that there are hundreds of artists that could easily paint a Rembrandt-style painting today, probably as good as this computer did if not better. And there are probably thousands with the talent to learn how to do it easily, but they have no desire to.

Why? Because that would just be "derivative." It has already been done. The mastery of skill and technique is considered only what one does in school -- to become an "artist" today requires innovation. Painting a realistic expressive scene with interesting chiaroscuro like Rembrandt is a great "exercise," but the "geniuses" of the 20th and 21st century wouldn't be caught dead doing that as their own output -- unless they were doing it ironically or something.

TL;DR -- Rembrandt wasn't a "genius," since that whole concept of artist as "genius" hadn't yet been invented in his lifetime. And there are plenty of modern folks people call "artistic geniuses" -- but usually because of their unique innovations or individual ideas of expression, which are often quite different from the goals of someone like Rembrandt.

Re:No, it didn't.

[AthanasiusKircher]

But can the AI itself then be considered art, as it has been made to make people observing it in action feel awe and astonishment at what it is doing?

It really depends on your definition of "art." I can "observe" a waterfall or an erupting volcano "in action" and "feel awe and astonishment at what it is doing."

Are they "art"?

I know what you mean, though. For most people, "art" has something to do with a special kind of "aesthetic" appreciation. If you are able to appreciate the actions of a computer program in that aesthetic fashion, then sure, why not call it "art"? For some, there are certain things that are always "off-limits" for that kind of aesthetics -- this is where you get that big debate with Roger Ebert that happened several years ago where he said that video games could never be "art." That's because he was convinced (as many art fans are) that certain kinds of things are open to aesthetic appreciation, while others aren't. Almost everyone has some sort of boundary there -- Duchamp's famous presentation of a urinal as a piece of "art" shows how one can ultimately push that boundary to absurdity.

Or perhaps not. Perhaps "aesthetic appreciation" is really just in the mind of the observer, and if the observer declares it to be "art," then why not? That's part of Duchamp's point.

Of course, I doubt you wanted to dredge all of this philosophical baggage up with your comment. But it really gets at the heart of what "art" is, and how one knows whether to consider something to be "art."

Name-dropping

[Tony Isaac]

This is a classic case of name dropping to make the software seem more amazing than it actually is.

Lots of students in art classes can copy the work of a famous painter, and even imitate the style. Some can even do so convincingly. But that doesn't make the student a "new Rembrandt," in fact, most such students will never be recognized as brilliant. There is much more to being a master artist than just imitating the style of a great painter.

Re:FTFY

[phantomfive]

Your "top list" has Yoko Ono on it.
I like Yoko, but if she's a top artist, then we've fallen far from the days of Rembrandt.

Re:Nope, sorry

[hairykrishna]

It's not just digital. There is a physical, 3d printed, picture with a texture which is presumably mapped from Rembrandt brush strokes.