Hacker News new | past | comments | ask | show | jobs | submit login
The AI-Art Gold Rush Is Here (theatlantic.com)
66 points by longdefeat 14 days ago | hide | past | web | favorite | 40 comments

When Google dropped that first batch of Deep dream images, those were compelling. The images bore an uncanny resemblance to things you see while hallucinating on psychedelics, and the images make you wonder about just how similar the processes going on inside a CNN to processes going on in our own minds. Those deep dream images were new, they invoked a strange sense of frisson, you had to bear witness to whatever it was.

But the images shown in this article just look like somebody was fucking around with random filters in photoshop. It's not interesting. It doesn't leave a lasting impression, and that's really the only metric that is universally cross-comparable when judging art.

On the other hand, https://thispersondoesnotexist.com/ is the new New and it's compelling on a whole other level. So are the GAN-generated game sprites and other pixel art we've seen over the past several months.

The artistic value of AI generated art (to me at least) will come from the obsolence of human creativity and the strong feeling of void associated with the far-future idea that algorithms and processing power are starting to surpass our species in the only remaining way that made us relevant, and looking at the pictures in the article, we are not there yet.

I am not an art connoisseur, but as a layman those pictures do not spark any emotion in me. They are like random noise that do not convey any context. Whereas there are paintings that mesmerize the mind by connecting the viewer with the painter. The day that AI art will emotionally strike people is the day when it can deduce context from our world that is relevant to us, and the art in it is the creation of the empty feeling inside us when we realize we are now being emotionally moved by an algorithm, in a way only an exceptional artist could (or surpassing talent of such artists).

I am skeptical about current machine learning methods taking AI art that far. Perhaps ML will make for example great music by generating sequences out of training sets, but it's hard to see it creating a new music genre people will vibe with in the near future.

>and the strong feeling of void associated with the far-future idea that algorithms and processing power are starting to surpass our species

The void is inversly proportional to the bitrate and error rate between human/machine communication. It ceases to exist at some point following a sufficiently data dense mind/machine merge.

Going through the images, I shake my head in disbelief. So much of “art” is selling crap to the filthy rich, who have nothing better to do with their time or money.

Maybe, I don’t “get” it. To each their own, I guess.

There is a business opportunity here, so it only makes sense that someone is exploiting it.

I hope you don’t believe that about all art. Art is one of the pillars of Homo Sapien behavior, and it is found pretty much anywhere human settlements are found. I worry that many of us are becoming too distracted with life, logistics, and “productivity” to take time out to experience the things that make us human. Maybe that’s why art has become associated with the rich - maybe they’re the only ones who have time to enjoy it.

> I worry that many of us are becoming too distracted with life, logistics, and “productivity” to take time out to experience the things that make us human.

True, but art as an industry created a bubble on its own. Nowadays it gives the perception that it is all about imposture and speculation.

Art, yes. 100k painting with no cultural significance... that's probably more about rich idle hands or money laundering.

I think it's a not uncommon view particularly about fine art as bought and sold in the galleries and auction houses of New York, London, Paris, and Tokyo.

Some art is treated like an investment, it is speculated that the price of a piece is only going to go upwards. That is a reasonable assumption if the artist is famous, especially if the piece of art reflects a certain style or creative period. Usually it is very hard to copy the work/style of an artist. That said, I don't know how these pictures are going to hold to that assumption, since 'generative art' is by default reproductible and should be easy to copy...

Some people actually like having multi million dollar paintings hanging up. But, the number of people that can afford to do so is relatively limited.

Some as you say is pure speculation, but money laundering is also a significant part of the art world. Auctions regularly move vast sums of money between near anonymous people and represent highly portable wealth.

"Art" with a capital "A" is much like religion: a pseudo-intellectual toy of possible conceptions and idealism. The big different between the fraud of Art and the fraud of religion is Art claims you can buy transcendence, whereas religion says you must give your soul and lively obedience for the potential of their idealism after death.

Randomly generated AI art is just a layer of abstraction on top of generative art. A more human form of AI created art is AI-assisted art. Art created by a human using AI tools to sculpt the art into their image which they couldn't otherwise manifest.

There is still creativity in the architecture, the training data, and the sample selection. None of the art is randomly created, but carefully curated by a human.

Question is, how can we trust the artist if they are truly legit anymore?

Generative art is truly a compositional creative medium using computers and algorithms to insert randomness and to tweak the composition in ways that use math.its the intervening of the artist in this process, the intentional choices of algorithm and training content, and the synthesis of these layers by the artist that delivers a work of art. Slapping someone else's code into a generic set of training content is just a coding excersise and the end result is maybe in the category of DuChamp calling a toilet a work of art, as a thought experiment...but I think the the lack of transparency and attribution to the original coder makes this much more an act of opportunism.

In Donnella Meadows http://donellameadows.org/archives/leverage-points-places-to... she describes the many points within a system one can intervene for various effects. It's important to note where in the generative process the artist is intervening in what is otherwise just a creative coding excersize. It's in the intervention that the most talented generative artists demonstrate their gifts.

Art is the conveyance of meaning through a work, not the method by which the work was created (although that can sometimes contribute to the meaning).

No, today, "Art" is anything an artist can convince a source of cash that the cash they want is for this argument claiming some object/event/scribble/thought is "Art".

I'm a professional artist, where are these sources of cash I can convince to give me money instead of spending all my money from my day job on materials?

Let's say I have a machine which looks for mathematical theories. Upon finding and verifying some statements, I take some of the results and publish it. Who is the author and does it matter?

Isn't the credit important at some point? For e.g. it matters to me to decide if I should watch for your future work vs the subject itself.

If you're referring to predictors of future quality, does it still come from the human? Who is the author?

On a similar question, for the combination of credibility + relevance, do you look to metaphorical Google News, NYT, the journal from which the article is sourced, the science team being reported on, or the individual scientists on the team?

I imagine in the future of credibility we might discuss the forecasting abilities of specific models.

Guiding a machine to generate art is no different than guiding a paintbrush rather than finger-painting or using a jackhammer rather than a hammer and chisel for sculpture.

I for one am excited for what crazy new stuff artists will create with better tools!

There's a slight difference in that mastering mechanical movement is no longer necessary to create art. I imagine there will still be a place for traditional artists and their unique talents, much like how music synthesizers have not gotten rid of our desire to appreciate live instrumentalists. Rather, some new people appreciate the DJ, while most still appreciate the real-breath/finger performance.

Mastering mechanical movement hasn't been necessary for a long time.

The paintings remind me of Francis Bacon’s works from the 60s


It's a neat experiment for sure, and I agree with others here that the high price that piece fetched was from it being the 'first'.

IMHO, AI-Art will never ever produce the commentary in art that humans can. Depending on the time, medium, patron, subject, artist, etc etc, symbols can make impactful statements in art.

This AI stuff is neat, but it seems like a mash up of other art slammed together to make something new. I think the only commentary that it can make is that it was made by a machine.

Takes away room from serious art, in my opinion, like from this oil bym Brianna Lee, Metamorphose https://i.imgur.com/VeLbX27.jpg

Any conversation that gets anywhere near art on hacker news makes me deeply depressed about the community here.

The emperors new clothes.

I think the 450k GAN painting was a bargain. There is no doubt that neural art will increase in quality and quantity over the years, maybe even surpassing us in stylistic insight. That painting is the first of its kind, much like owning a first-in-style classical painting from the middle ages. Either that or it becomes a worthless parlor trick (but I deem the chance of this low).

I think it's very likely to become a worthless parlour trick. That one piece may have value as The First of Its Kind [tm], but the others will be irrelevant.

They've used a lot of signifiers of Expensive Vintage Art with a Modern Twist - dark backgrounds, exaggerated flesh tones, rich-looking clothing, aggressive modern distortions - to suggest value. But the neural networks that may (perhaps?) be being fooled are the ones inside the brains of buyers.

By the end of the year this art will be available as an instagram filter and/or from a web page. Which is why it will be worthless. Neither the software nor the imagery are special enough to survive that.

The real art - if that's the word - is in how this is being run as a startup and is trying to turn investor money into a hype train.

> I think it's very likely to become a worthless parlour trick. That one piece may have value as The First of Its Kind [tm], but the others will be irrelevant.

That's technically not disagreeing with the OP. And I do think that the first one will be highly valued, independent of the fact that it'll be available. Just like the Banksy Image that was shredded will be worth so much more now; any images from now on that pull this trick won't be as highly regarded as well.

Other than that I agree: AI art is so easy to produce that the sheer quantity will make it uninteresting for collectors.

But it is a worthless parlor trick. Hell, I could “create” pieces of art like that in Photoshop in a few hours if I really wanted to. How is that in any way anything beyond the bare minimum for creativity or artistry? Why is using a GAN to do that any more artistic? This kind of artwork is the literal definition of “derivative”.

But hey, people are free to waste their money if they like.

Here's an important piece of art from the early 20th century: https://en.wikipedia.org/wiki/Fountain_(Duchamp)

Cheap parlor tricks have a long and glorious history in art, and sometimes become important works of art.

That significance is because it was an originator and trail blazer of the cheap parlor trick.

I think the artistic value in that one specific Picture does not lie in the image itself but that the artist was the first one to put an AI-generated Image on display and proclaimed it art in such a way that it found a buyer. Now that it has been done, other AI-created images are boring I think. (AI-assisted generation of Artworks which still have a substantial amount of a human artists creativity and craft involved is another matter and could find its niché (similar to other tool-assisted art like procedural art)

This sounds like every argument against Mondriaan.

It would only be an argument against Mondriaan if Mondriaan could easily be turned into a Photoshop plug-in.

I tried to make an auto-Mondriaan machine once, and I gave up after a few days. It's one of the hardest things I've attempted. It's incredibly easy to splash some black lines and primary fills around, and unbelievably hard to get the proportions and density right.

It was an anonymous bid, and the "creators" of the piece apparently met at a blockchain conference. I know what conclusion I drew from that, but I'll leave you to draw your own.

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact