It's in The Guardian, here. It's nicely set up, with sets of images in different categories of painting — not just Rothko. The experts get it wrong most of the time, perhaps because the human-made paintings aren't that great. It's pretty amusing to watch experts getting things wrong and real human art getting disparaged, so that's why I'm recommending this piece even though, generally, I'm sick of articles manufactured out of bullshitting around with AI.
From one expert's discussion of 2 abstract expressionist paintings:
"I’m not sure about the shapes and the lines in the image on the left, but it does make me think of very early Pollock, though less colourful. The one on the right could relate to many early works from some of the abstract expressionists or, perhaps, Tancredi and certain Italian artists from the 1950s and 60s. I’m sure the AI is looking at these existing works in order to create something based on them but I would still say the one on the right is real.”
Verdict: wrong
“When it’s not by a particular artist you know very well, it’s much harder to determine what feels wrong. With a specific artist you look at how they worked at a particular time, their colours, their compositions and what the feel of it should be. If it could be any artist, it’s a bit random.”
The experts are actually getting rooked into criticizing their own expertise.
पोस्ट्स
३० टिप्पण्या:
The AI expert witness approaches.
Musk is worried we could be reduced to house pet status if we don’t put some safeguards in place.
The Guardian can't tell the difference between Trump's classified documents and Biden's; how can we trust them on matters of art?
I've seen similar articles regarding wine and wine-tasting.
While not necessarily defending art critique (I think it's a little bit silly) I will say that ones mind expects something, then you will follow through with that expectation.
As an example, for those who drink soft drinks, if you like both Coke and Dr Pepper, if you expect one but get the other, the response will be 'strange' or 'yuk' (YMMV). The expectation can override reality.
It's common in subjective fields (wine tasting is a great example) for experts to fail at blind tests. An old girlfriend and I used to try it ourselves. For instance, we once brought expensive maple syrup to a diner and compared it to the fake syrup in plastic packages they offered. Neither one of us could tell which was which. Coffee beans are another example--the difference between a good and bad cup of coffee is how it's brewed and how fresh it is. The beans themselves don't matter as much as people think.
There's a lot of room for AI to screw with people.
I remember reading a long time ago that when wine tasters are blindfolded, many of them can't even tell for sure whether they're drinking white or red. This seems impossible to me - they make your actual tongue respond differently - but there you go.
I'll bet the experts would do better if the art were in front of them, not just a picture. Brush strokes and all. But all this was doing was asking them to discern whether a human mind was directing these compositions, and with abstract expressionism, who can say? I mean, when some of the works produced by humans are just the human's choice of which piece of half-painted wall (or whatever) to cut out and frame, the artist didn't actually direct the painting. Banana taped to wall - the artist had the idea that it would be art, and chose the banana and the tape and the position, but did not conceive of or create "banana," "tape," or "wall."
So, what is art these days, anyway?
“…what the feel of it should be…”. Hard enough to articulate and defend your expert opinion when the subject matter is words (literature) or ideas (legal arguments and decisions). But when the subject matter is blobs of color on canvas or a pattern of neurochemical connections from stuff in a wine glass? Please.
That said, we need to be able to differentiate red from pink, and box wine from Chateau Margaux; and on some kind of log scale of value, we do.
if you like both Coke and Dr Pepper, if you expect one but get the other, the response will be 'strange' or 'yuk' (YMMV).
I was at a dinner last weekend where, when everyone arrived, the salads were already plated, as were the desserts. Between those two items, there was a little shot glass of something pale and creamy. One of the other people at our table told his wife that it was a shot of Bailey's, which could happen at this dinner - open bar and pretty much everyone was drinking.
I like Bailey's. So, boy, am I glad that I tumbled to the fact that it was salad dressing before slugging it down!
"If it could be any artist, it’s a bit random.”
If it could be any singer/musician/poem/sentence/wine/liquor/fashion item/photograph/art object, it's a bit random.
Really brought home to me during opera intermission games way back: experts had a hard time identifying voices.
Thesis: an AI bot would have created a better MLK sculpture. There you can tell the human touch by how bad it is. Then again, if we start training AI on modern art, who knows?
Can AI generate a Norman Rockwell illustration or a Chas Addams cartoon? Abstract expressionism and, to a lesser extent, landscapes are abstracted from human feelings. On the plus side, you could probably set the parameters to proofread your work to insure that any part of your sculpture would not be mistaken for a dong.
Abstract expressionism was a con all along. More of a decorative craft, sometimes carried to a very high level, like making doilies, than art, and AI has exposed it.
Call me when AI can drink itself to death or cut off its own ear.
There was a hilarious Murphy Brown about exactly this. Her handyman who is constantly fixing her house and painting the rooms wanted to be an artist. He finally got an exhibition at some fancy snooty place. He was so proud. The intellectuals walked in and the room was empty! They turned on the light, and there is nothing there. They begin to theories as to his message. All that was in the room was a light switch. They conjectured the meaning of the light switch. Eventually, they concluded we are the masters of our own lives. It is within our power to turn the light on and turn the light off. finally the blue collar guy. The artist walked in and said no you dummies. I painted the mural on the ceiling. That is where the art is! I think that sums up the modern world of art today.
Here you go… https://archive.org/details/murphy-brown-s03E13-eldin-imitates-life
I'm not really buying the underlying concept here.
If you think of the AI pictures as forgeries or copies, then I'm not surprised experts can't tell them apart just by looking at them (i.e., no use of chemical, x-ray, or other physical techniques). After all, the AI is working from the originals, so why shouldn't the result be a good approximation of the originals?
Aren't forgeries of "lost" artworks made by humans nearly impossible to detect by sight alone? Why should AI forgeries be any different?
“We ask the experts to tell the difference between human and AI art."
I could tell the AI art because it’s better. We’ve known since last year that meatsack artists are doomed, and this article just shows that meatsack “experts” are going to follow them into extinction.
Here’s a better one: is it a Pollock or was it painted by Suda the elephant?(
"The experts get it wrong most of the time, perhaps because the human-made paintings aren't that great."
Or perhaps because the whole edifice of art critique is a bit of a scam.
I did my own analysis; I guessed which one I liked and which I didn't. I was right 100% of the time.
Turing Test. Don't know if people know that one of the most important techniques now for machine learning is using two computers against each other. This one tries to fool that one, this one learns how not to be fooled, the first one corrects its attempts. After a while they're both really good at it - better than your "experts".
"Computer, come up with a dramatic sense of ennui w/o involving people."
One of the authors of "Freakonomics" told the story about switching wine labels at a Harvard or Yale event and nobody telling the difference between the vintages.
But I know for sure some can tell the difference. My MIL has a nose sensitive enough to detect a gas leak in her house that took the gas company sensitive equipment to detect. My stepson and daughter both have more sensitive taste buds than I, and I rely on them for cooking advice.
The current hilarity surrounding AI art derives from the fact that aesthetic relativism, by devaluing originality, complexity, intensity, and meaning, has made much art into something merely decorative and algorithmic.
Ironically, extremely bad art is less likely to be generated by AI. No AI could come up with Boston's disastrous sculpture.
Documentary about a probable Jackson Pollack painting found in storage:
"Who the #$&% Is Jackson Pollock?" (2006)
https://www.imdb.com/title/tt0487092/
Watch it for the reactions. "Artsy fartsy" became an expression long ago.
MikeR, anyone who does that with two AIs should watch the old SF movie "Colossus: The Forbin Project". Spoiler - it does not end well for humans.
My favorite AI art was a Dalle2 piece titled "Girl With A Pear Leering."
A modern model of empathetic correlations.
The amazingly original point of abstract expressionism when it was a novel art genre was its sponteneity, its improvisation, and the requirement of an interpretation by the viewer rather than having a narrative presented by the painter. That, and it was the first US-originated artsy movement, raking in cash for American artists, whereas Paris had been the center of the art world's originalism and financial success, up to WW II.
I, for one, saw a Wahlberg movie - Contraband, I think? - wherein a Pollock was stolen by treating it like a house painter's floorcloth. It was rolled up in a messy, paint-splotched bundle and tossed in a house painter's van where it remained secure from the police and other criminals seeking it. Now, that's the way to treat an abstract expressionist art piece!
When is AI going to ruin music-playing as a career??
William said...
Can AI generate a Norman Rockwell illustration or a Chas Addams cartoon?
I don’t know how I can share it here, but I just generated four Addams-like cartoons. Then I generated four Rockwell-like paintings. All in minutes.
Tim, you paid too much for low-quality, last run maple syrup. Aunt Jemima and other maple-flavored corn syrups have a much, much stronger maple flavor than high quality maple syrup. The only type of real maple syrup that comes close to the flavor of fake maple syrup is the lowest quality, last run syrup that is normally only sold in bulk for flavoring, sweetening, etc. of other food products.
टिप्पणी पोस्ट करा