my kid could program that
October 30, 2018 7:56 AM   Subscribe

The First AI-Generated Portrait Ever Sold at Auction Shatters Expectations, Fetching $432,500—43 Times Its Estimate

CNN: A sign of things to come? AI-produced artwork sells for $433K, smashing expectations
“Edmond de Belamy" has made history as the first work of art produced by artificial intelligence to be sold at auction.

The slightly blurry canvas print, which has been likened to works by the Old Masters, sold Thursday for $432,500 -- dramatically exceeding its original estimate of $7,000-$10,000 -- at a Christie's auction in New York.

While the print is signed "min G max D x [log (D(x))] + z [log(1 -- D (G(z)))]" after a section of the algorithm's code, it was conceived by Obvious, a Paris-based trio fascinated by the artistic potential of artificial intelligence and machine learning.

To produce "Edmond de Belamy" and the 10 other portraits in the "La Famille de Belamy" series, Obvious fed a two-part algorithm 15,000 images of portraits from different time periods. After reviewing these submissions, the first part of the algorithm began generating its own portraits, trying to create original works that could pass as man-made.
More about the creation of the painting, per Christie's: Is artificial intelligence set to become art’s next medium?

Controversy, via The Verge: How Three French Students Used Borrowed Code to Put the First AI Portrait in Christie's
But for members of the burgeoning AI art community, there’s another attribute that sets the Portrait of Edmond Belamy apart: it’s a knock-off.

The print was created by Obvious, a trio of 25-year-old French students whose goal is to “explain and democratize” AI through art. Over the past year, they’ve made a series of portraits depicting members of the fictional Belamy family, amplifying their work through attention-grabbing press releases. But insiders say the code used to generate these prints is mostly the work of another artist and programmer: 19-year-old Robbie Barrat, a recent high school graduate who shared his algorithms online via an open-source license.

The members of Obvious don’t deny that they borrowed substantially from Barrat’s code, but until recently, they didn’t publicize that fact either.
On giving credit where credit is due, at Smithsonian Magazine: Christie’s Is First to Sell Art Made by Artificial Intelligence, But What Does That Mean?
Still, the debate over authorship barely begins to address larger questions of AI autonomy. Christie’s has been quick to capitalize on the Belamy portrait’s singular status, defining its medium with a heady catch-all description: “generative Adversarial Network print, on canvas, 2018, signed with GAN model loss function in ink by the publisher, from a series of eleven unique images, published by Obvious Art, Paris, with original gilded wood frame.” Obvious itself initially marketed the work as “created by artificial intelligence,” but Caselles-Dupré tells the Verge he now regrets using such definitive language.

Jason Bailey, the digital art blogger behind Artnome, explains why such phrasing is misleading, arguing that “anyone who has worked with AI and art realizes” algorithms are tools, not active collaborators or autonomous agents.
Obligatory kvetching, from The Guardian: A portrait created by AI just sold for $432,000. But is it really art?
posted by Iris Gambol (47 comments total) 11 users marked this as a favorite
 
The “slightly blurry” quality brings to mind Ecce Mono (Ecce Homo), the Spanish religious fresco restored in 2012.
posted by Iris Gambol at 7:57 AM on October 30, 2018 [6 favorites]


It would have been really cool if the painting had been bought by a robot bidder.
posted by chavenet at 8:15 AM on October 30, 2018 [5 favorites]


Great post, I'll be reading the links throughout the day, but OMG the post title...sublime.
posted by LooseFilter at 8:24 AM on October 30, 2018 [3 favorites]


But is it really art?

Art isn't art anymore, it's money laundering.
posted by Etrigan at 8:27 AM on October 30, 2018 [15 favorites]


I think "but is it really art?" is some kind of trigger for me, I cannot even begin to assault that question, let alone bring myself to answer it. It's just a question so loaded with ignorance and judgement I can't even.
posted by GoblinHoney at 8:35 AM on October 30, 2018 [3 favorites]


Interview with Hugo Caselles-Dupré, one of the members of Obvious, walking back some of the claims previously made for the painting:

Why did you say, “Creativity is not only for humans,” implying that AI was autonomously making the work, even when you knew that was a false statement?

HC: If we knew we were going to have 400 press articles on what we do, we most definitely would have done that. But at [that] moment we were like, 'Yeah, it’s silly, okay, whatever, let's put this.' But retrospectively, when we see that, we are like, 'That's a big mistake.'

posted by verstegan at 8:38 AM on October 30, 2018 [1 favorite]


At first I thought this was completely inexplicable, but I suppose if most of the work on it was performed in 2014, that would indeed be before this paper on style transfer via a CNN (2016) which we all currently know and love, so it could have some value due to being 'first'.

As far as the aesthetics I'm not a fan, but then again I wasn't bidding.
posted by Kikujiro's Summer at 8:40 AM on October 30, 2018


Art isn't art anymore, it's money laundering.
posted by Etrigan


The news of this sale prompted a small twitter freakout on my part along these lines:
So when people talk about "art," there are at least 3 different spheres they could be talking about. The one that gets the highest-profile news coverage, and is by far the least interesting is the "shit treated as investment by moronic rich philistines" sphere.

The second and much more interesting (and sometimes overlapping with sphere 1) one is the "body of work more or less agreed to be part of the high-cultural canon" sphere. This is what you go to see in museums, by and large.

And then my favorite sphere, the biggest but least-respected and least-covered one, is the "personally meaningful, revelatory creative work done by people." Which, y'know. The watercolors you futz around with at home. The webcomics you read. Inktober stuff. Knitting. You name it.

If #s 1 and 2 are kind of defined by the presence of gatekeepers, # 3 is emphatically not, and one of the most exciting things in the art world, if you ask me, is the way the internet has been making sphere 3 stuff drastically easier to disseminate widely. It's not universal, but by and large the art worth studying and thinking about because it reflects humanity happens in sphere 3, while idiot philistines trade money in sphere 1.

Anyway, whoever bought that AI artwork is going to lose their money and is a total fool.
posted by the phlegmatic king at 8:47 AM on October 30, 2018 [5 favorites]


Ehhh, they're just in it for the Walter Benjamins.
posted by clew at 8:56 AM on October 30, 2018 [8 favorites]


All this really shows is how bankrupt the big money art world is. But I don't suppose you needed any more evidence after Damien Hirst.
posted by demiurge at 9:02 AM on October 30, 2018 [3 favorites]


I think "but is it really art?" is some kind of trigger for me, I cannot even begin to assault that question, let alone bring myself to answer it. It's just a question so loaded with ignorance and judgement I can't even.

I go by a rule of thumb: If the question has been asked, the answer is automatically "yes." Makes for some interesting conversation.
posted by klanawa at 9:09 AM on October 30, 2018 [4 favorites]


Or, another way:

Q: Is it art?
A: It is now!
posted by klanawa at 9:13 AM on October 30, 2018 [4 favorites]


I think "but is it really art?" is some kind of trigger for me, I cannot even begin to assault that question, let alone bring myself to answer it. It's just a question so loaded with ignorance and judgement I can't even.

Soooo...I'll mark you down as a 'leans yes'?
posted by NoMich at 9:16 AM on October 30, 2018 [2 favorites]


mfw when I smell something off but can't tell what it is.
posted by Going To Maine at 9:24 AM on October 30, 2018


Sigh. I hate that they ripped off someone else's code to generate the art, code which shipped with a disclaimer to not profit from work generated from it. But in doing so, they probably increased its art bona fides. There's a long, long history of copying work wholesale in the art world (see Lichtenstein, for example).
posted by zsazsa at 9:26 AM on October 30, 2018 [6 favorites]


It's only art if the AI has its feelings hurt when you give it an unfavorable review.
posted by Burhanistan at 9:27 AM on October 30, 2018 [2 favorites]


Here's Jerry Saltz's take on it in Vulture.
posted by PussKillian at 9:30 AM on October 30, 2018 [1 favorite]


If the question has been asked, the answer is automatically "yes."

I shall refer all to an assertion made at the end of this*


* disclaimer: a film in which I appear for a fraction of a second
posted by CynicalKnight at 9:31 AM on October 30, 2018


Pffft...my phone could paint that.
posted by sexyrobot at 9:43 AM on October 30, 2018 [3 favorites]


...and the computer won't see a penny of the proceeds. Typical.
posted by leotrotsky at 10:13 AM on October 30, 2018 [2 favorites]


Such idiocy. Meanwhile people starve for want of a dollar.
posted by Liquidwolf at 10:39 AM on October 30, 2018


Of course your kid could paint/program even better -- but can you or your kid orchestrate the international media hype, shmooze corrupt gallery owners, spin social media, dictate a juicy story to glossy art magazine editors, manipulate markets, interact quietly with wealthy oligarchs to help with a bit of, cough, laundry?

What's on the canvas is by far the least important part of the art world stories.
posted by sammyo at 10:46 AM on October 30, 2018 [5 favorites]


The appropriate response whenever anyone says "My kid could do that" is "Maybe... but he didn't!".
posted by Liquidwolf at 10:51 AM on October 30, 2018


So I learned about this on the 25th, from someone in the code-art circles I follow on social media signal boosting Robbie Barrat's twitter thread basically saying "is it just me or is this really fucked up?" re: the emerging fact of the sale of the piece for half a million dollars. Barrat had already been rightly salty before that, at just the news that Obvious had been putting it up for auction at all (for an estimated $7-10K sale price) and had only belatedly and grudgingly acknowledged in part the work's debt to his code and at that maybe not so much at all in the context of the actual auction.

And there's a high-minded "what is the nature of art, and is art that challenges perceptions of the nature of art the real art, and..." discussion that this can all play into; and there's the reality that what-art-is and how rich art speculators spend their money are not in lockstep in the first place; but then there's the question of "were Obvious basically shitheads about this" and that feels like a real clean obvious yes. Like, y'all, you were shitheads and it paid off and that does not make you less shitty.

Being a shithead isn't necessarily actionable, but it sure is shitty. Barrat deserves a full-throated, embarrassed apology no matter where else you fall on the rest of it, is my basic feeling. But I also don't see that as forthcoming.
posted by cortex at 11:13 AM on October 30, 2018 [1 favorite]


Also, talking about this elsewhere, someone mentioned another interesting case that isn't directly related but falls in the same territory of prestige attaching to derivative methods with a conspicuous failure to credit the prior history of work behind it: Kevin Ferguson's justifiably grumpy essay To Cite or to Steal? about Jason Shulman's gallery show in 2016 of whole-film single frame average visual distillations.

Which gets into similar territory of divorcing the question of whether two different artists/groups could have and execute the same idea (of course they can) from whether, given evidence that there's an unacknowledged, unattributed history to the execution of those ideas, it is shitty to basically be "welp, it's not ILLEGAL for me to pretend the extant work has nothing to do with my basically identical output, so you can't stop me". Which: it is, I'm gonna say, shitty indeed.
posted by cortex at 11:19 AM on October 30, 2018 [1 favorite]


Does the value of the painting go up if the program that created it is deleted?
posted by chromecow at 11:24 AM on October 30, 2018 [5 favorites]


Well, at least it's not a creepy dead-eyed monstrosity that both evokes and obscures the humanity of its subject, providing an unmistakably human form in civilized dress but also a featureless grotesque figure with grotesque proportions. That would just be a bad omen for human/machine relations for centuries to come.
posted by es_de_bah at 11:37 AM on October 30, 2018


I always tell my students that if a made thing has human expressive intention, then it qualifies as art. So my answer to the question "is it art?" is nearly always 'yes, of course.' But that's not an interesting question; the interesting questions are 'is it interesting or compelling in some way?' 'does it reward extended consideration/repeat viewings or listenings?' 'is it well-made?' etc.

So yeah, this is art, but the artists are the humans who made the AI, a group of artists calling themselves Obvious, who created a tool to make visual art. It's a really sophisticated tool, way more capable than a paintbrush, but the AI is still a tool made by people for making art. 'Obvious' appear to be removing intentionality from their visual art creation process, by putting all of their intent and control into the tool/process, and then letting that run to make the object.

This is what the composer Steve Reich did, following John Cage's philosophical lead, in creating process-based musical composition techniques like phase shifting (e.g., Come Out, It's Gonna Rain, or--most closely resembling this piece of visual art--Piano Phase). His succinct explanation is worth taking 3 or 4 minutes to read, and I think his characterization of composing a process, feeding that process a small bit of material, and then letting the music create itself as the process runs, is very similar to writing code and then running it, with the artwork itself as the outcome of the process of the software. It's all still human-created art.
posted by LooseFilter at 11:38 AM on October 30, 2018 [1 favorite]


> Robbie Barrat's twitter thread basically saying "is it just me or is this really fucked up?"

I just looked at the github repo, and it's licensed under BSD.

"EXTRA: NO OUTPUTS OF THE PRE-TRAINED MODELS MAY BE SOLD OR USED FOR-PROFIT OTHERWISE." was added five months ago.
posted by Leon at 11:43 AM on October 30, 2018 [1 favorite]


Also, I think the market for 'fine art', and the hype/publicity, and etc., are all cultural concerns separate from the much more interesting creative ones that Obvious have introduced by making an AI paintbrush. All the rich people playing these collector games will die soon enough, and the work and tools will abide. Ars longa, vita brevis.
posted by LooseFilter at 11:43 AM on October 30, 2018 [1 favorite]


creative ones that Obvious have introduced by making an AI paintbrush

That's kind of the point of contention: Obvious didn't make the paintbrush, Robbie Barrat did. If anything, it's more like they bought a mold and then sold the resulting cast for a huge amount of money while only begrudgingly acknowledging the mold-maker when called on it.
posted by Pyry at 12:29 PM on October 30, 2018


Again, the code's licensed under BSD. As far as I can tell, Obvious have met the terms of the license. In fact, they've gone beyond the terms - when the author contacted them asking for acknowledgement beyond what BSD requires, they did that.

It honestly annoys me that some guy makes the choice to license under the most permissive license there is short of public domain, then walks it back, and it's the people who took him at his word that are getting the flack.
posted by Leon at 12:35 PM on October 30, 2018 [1 favorite]


I wonder if Robbie Barrat realises he's still publishing the version of the code without his ad-hoc license addition. (It's at b475eb0b if anyone's interested).

That's a consequence of publishing the whole repo of course... that's going to be an interesting one when the lawyers get ahold of it.
posted by Leon at 12:42 PM on October 30, 2018


under the most permissive license there is short of public domain

This would have been legal under GPL and even AGPL, because the painting is a product of the tool, in the same way that if I produce an image in Gimp I could sell and even claim copyright on it.

But most people are not arguing about the legality-- you can both believe what Obvious did is completely legal and at the same time be irritated that they are getting boatloads of money and publicity for essentially running someone else's photoshop filter.
posted by Pyry at 12:45 PM on October 30, 2018


Obvious say they used their own training data, and for the moment I think I have to take them at their word. So what makes this tool unique, as against all the other open source tools that people use to make money? (Gimp is a good example).
posted by Leon at 12:49 PM on October 30, 2018


Apologies for glossing over a point of substance w/r/t creation of the tool(s) in question. (Though that raises a primary challenge to contemporary ideas of authorship and ownership, etc., going forward, since so much creative work is unavoidably collaborative, given the sophistication of our tools now.)

The most interesting thing in all of this to me is the tool itself, and how it may be used going forward. From this musician's perspective, it's really compelling to see a typically non-temporal medium like visual art devise a collaborative, process-based means of creation. That's kind of been the sole domain of music (as a medium, intrinsically) for most of human cultural history.
posted by LooseFilter at 12:49 PM on October 30, 2018


So what makes this tool unique, as against all the other open source tools that people use to make money? (Gimp is a good example).

It's the difference (which is, again, not a legal difference) between using Gimp in a genuinely creative manner vs. just running the 'GIMPressionist' filter on an image. If Obvious had convinced the art world that an otherwise unremarkable photo run through a stock Gimp filter was worth $400k, then irritation with both the artists and the high art world would be an appropriate reaction.
posted by Pyry at 12:59 PM on October 30, 2018


> Obvious say they used their own training data, and for the moment I think I have to take them at their word.
I'm not sure we do based on how strikingly similar works were created using Barrat's model as-is.

There's a huge amount of possibility in marrying Deep Learning techniques to make novel works of art. One of those possibilities is grifting rich dudes while putting in zero effort. I can't really get too mad about that, ultimately
posted by ReadEvalPost at 1:06 PM on October 30, 2018


Art isn't art anymore, it's money laundering.

- Savonarola, commenting on the Medicis, circa 1490
posted by clawsoon at 4:35 PM on October 30, 2018


It honestly annoys me that some guy makes the choice to license under the most permissive license there is short of public domain, then walks it back, and it's the people who took him at his word that are getting the flack.

This seems like a really narrowly code-centric lens on it all, though. Barrat was fine with what the Obvious folks presented themselves as wanting to do: some sort of "democratizing", research-sounding project extending the work his code was doing.

That they in turn pestered him for help with using his code, produced output congruent with stuff that had already previously been generated, declined to be at all transparent about any of that until a fine point was put upon it and then only grudgingly and tersely, and wrapped that all up in a pile of self-aggrandizing marketing for their design group, culminating in this auction, is good golly far worse behavior than being publicly salty about being exploited in that fashion.

"What we did is technically permissible" isn't a defense of any kind of good ethics, it's a defense of technical permissibility. That they're allowed to be self-serving shitheads hardly means they aren't being self-serving shitheads.
posted by cortex at 5:50 PM on October 30, 2018 [3 favorites]


Man gives away paint by numbers book for free, answers questions about his free book, and is not happy when someone sells the painting made by painting by numbers. Yeah, that's his bad.
posted by runcibleshaw at 7:55 PM on October 30, 2018


surely there are things that exist, but are not works of art
posted by thelonius at 10:53 PM on October 30, 2018 [2 favorites]


I guess we deserve all this.
posted by bongo_x at 12:54 AM on October 31, 2018 [1 favorite]


cortex: It's not code-centric, it's "what the author said he wanted"-centric. He gave his code away for free. They took it, added value (marketing, selecting their preferred output, and (maybe) some training data) and sold a product based on it. If they're wrong, so is every company on the planet that uses Apache for free.

Now that may actually be the case (I don't think it is, but you can make a case for it), but singling out for censure these three guys who have just done what everyone else does, from Amazon and IBM to the littlest startup, is... well, it's just today's unthinking Internet pile-on.
posted by Leon at 3:35 AM on October 31, 2018


But, again, that is a very code-centric (or license-centric, if you prefer) lens on it. It's narrowing the aperture of perceived wrong-doing down to the point where the only light it can let in is that reflecting off the question of "was the code involved licensed as open source", and saying that because the answer is yes, the only thing to say here is that no wrong-doing was done.

No one, not Barrat or anyone else, is disputing that he licensed his code as open source, making it available for use. That does not perforce mean that every use of it will be ethical; that does not mean that every complaint about unethical use of it is equivalent to demanding that open source licensing be arbitrarily revocable.

If I invite someone into my home because they tell me they need to go to the bathroom, and they proceed to take a shit on my living room floor, film it, and sell it to a wealthy scat fetishist, they didn't technically violate the terms of our agreement but they sure as hell are in the wrong. Reducing that to, "hey, you told them they could come in and go to the bathroom, this is on you" would be viewing the situation through a very narrow lens.
posted by cortex at 9:09 AM on October 31, 2018


I don't think this sold for the price it did for its art-ness. It fetched its price for its first-ness. Does anyone believe this is the last piece of AI art to come along? No? Then this piece will always have value as the first.
posted by aurelian at 8:09 PM on October 31, 2018


surely there are things that exist, but are not works of art
posted by thelonius at 10:53 PM on October 30


Name one in particular and we'll find out I guess.
posted by klanawa at 10:38 AM on November 2, 2018


« Older A veritable treasure trove of old footage   |   Condensed Milk Toast Newer »


This thread has been archived and is closed to new comments