Hacker News Reader

Comment by bambax | Hacker News Reader

> So, would people stop making art?

An analogy that TFA curiously doesn't touch is the advent of photography. What did it do to painting?

Painting became less and less figurative as photography became more and more accurate. The point of painting was to represent reality; but as photography became so much more accurate, that point became moot.

And so, painting went into a different direction. Represent not reality, but feelings, what the artist feels when looking at reality; something that's out of reach of photography.

AI is trying to conquer this as well. Human art needs to go further. Something weird that AI cannot touch.

View on HN

Art has already been there for ages. I went to see Copenhagen Contemporary's exhibit of light & space and I'd never actually FELT art like that. It was quite exhilarating. Suddenly i understand what people say they feel when they connect with art.

That's not what people mean when they talk about AI art though. AI art is Commercial Art. What Horkheimer and Adorno famously called "The Culture Industry".

I think the current output of AI art is mostly commercial but I'm sure an artist will come along and use it in some way different that exceeds the current use. Probably something that doesn't use text prompts or some other clever way.

It also pushed art to the niche. Painted images explicitly for children has been a massive industry for decades- also one that mature AI generative art could absolutely dominate.

I'd love to see this technology aimed at concept art for games. I don't think this would hurt employed concept artists, I do think it will be a tool concept artists can use to quickly thumbnail significantly more versions of a vision.

That artist will then refine the results into something supremely engaging to humans.

Really I see this as an incredible tool for artists to help refine truly innovative ideas.

If anything, painting has become more stylistic and figurative since photography; impressionism was a direct response to the photorealistic style of painting popular during the renaissance.

It freed artists of having to focus on realism, if they wanted to do so.

I don’t think that’s true though. There’s no reason you can’t prompt an AI to make an image that makes people feel things. One of the very first things I tried was “Mt rainier by Monet” and it certainly made me feel quite serene just like any human made nature art

I'm still wating for music to react to recording technology in the way painting reacted to the photographic camera.

Listen to some drum and bass.

I'd argue that Hyperpop is a more direct response to the commercialization of music.

I don't understand what you're saying here. Music has been in a dialogue with recording technology more or less since the beginning of recording technology.

fair enough. I can barely put it into words; but I'm still trying.

Music can be performed. Or rather, performing music is way less boring for an observer than performing painting or sculpture. Of course, speedpainting exists, but performative art is more resistant to being obsoleted by technology.

I'm starting from the viewpoint that the whole point of music is observing the performer do it.

in contrast, the whole point of paiting or a sculping is the final completed artifact.

What I'm trying to get at, is how recording technology "defeats" the purpose of music as its performace; and so, irremediably splits it from it's source (its origin, the composition of the music). I suppose improvisation remains; but highly constrained within prearranged structures (a la "AB[improvisatino goes HERE]C") like in jazz.

I do not understand how hyperpop nor DnB are supposed to be a response to what I'm trying to get at (then again, I'm just now barely conveying my point... I hope.)

Art that human beings actually care about is not the commercial art many people are referencing here. Art is not pretty pictures. Art is not a expensively produced media. Art is a human communication concerning the realizations of finite life. However, Art being Art, it does not communicate these issues in immature direct language, Art communicates through richly layered metaphor.

Real Art, that which moves one's soul, is beyond the capabilities of artificial intelligence for a large number of reasons, chief of which because it is a communication between beings aware of finite existence.

Note, I am not saying the consumer markets will not be flooded with cheap mimicry of art as pooped out by ignorant noncomprehending AIs. That will most definitely happen. And a generation of would be fine artists will never pursue their innate vocation thanks to the misconception that AI is capable of creating Art (which it cannot.)

However, you do not have to fall for the stupidity most are caught. If you have an artistic vocation, realize that indirection and metaphor are your human creative super powers that AI cannot touch. Metaphor requires comprehension and often complex interwoven comprehensions, which is so far out of AI’s reach it is laughable the general pubic believes otherwise. AIs are idiots when you know them well. They have zero capacity to create Real Art composed of complex metaphors embodied in a form other than how they appear.

> Real Art, that which moves one's soul, is beyond the capabilities of artificial intelligence

It’s easy to test. Show people “real” art and AI generated art and see how often they can tell which is which.

That won't work. Art is also not a popularity contest.

Then who is the arbiter of “real” art?

Nobody. Does it move your soul? Does it feel important that it exists? Does it matter if someone destroys it? These are answers nobody can collectively make, and that is why Art is simultaneously a grand fraud and the frontier of human reasoning. It cannot be defined because it exists on the edge of reason. The moment it can be created without a human communication behind it, it ceases to be Art.

Some DALL-E art moves my soul. It’s important to me that it exists. It would matter if it was destroyed. So, I guess that settles it.

Human curated DALL-E art. The AI just generates, the human curators, even if it is you, created the art. Recognition the generation communicates something within is the distinction, that's how "found art" became a thing.

I suspect you are going to repeatedly move the goalposts to keep the human in. Perhaps that's right, perhaps not but perhaps also ask yourself why you feel the emotional need to.

That is the nature of Art, it is a dynamic frontier, the edge of human communication and understanding. It is where our communication languages combine to speak what we've not said yet because we are yet maturing as a species and learning how to speak to ourselves.

I've no idea if that even means anything. I'll say it then, I think that's a load of tripe.

I've been saying this in the copyright discussions. There is a human in the loop doing both prompting and curation. The copyrights should be assigned to the human, at least in part.

Maybe a GAN generating images from random seeds without conditioning would make uncopyrightable content.

You are.

Each person is individually.

Great, then I say AI art is real art.

And that is perfectly fine. We are going to see a devaluation of image art, simply because ordinary people will cease to respect the imagery, believing no effort or human was required to create it. Currently, laypeople respect effort and that becomes their proxy for initial respect of image art. However, Art has not been merely the Art Object for over a hundred years. What is in Art museums today, discounting the historical art that serves as the mileposts leading to our current location, is the cultural identifier the Art Object presents for an idea. The object itself is merely the next milepost, something concrete people can point to, in the unknown frontier of our most complex communications to ourselves we've spoken to date.

"Art being Art, it does not communicate these issues in immature direct language, Art communicates through richly layered metaphor."

True, as is seen in the "banana with duct tape".

> They have zero capacity to create Real Art composed of complex metaphors embodied in a form other than how they appear.

Wait 5 years until the first android AI is being raised as a child and develops first hand embodied and social knowledge.

We've been 'waiting for 5 years' on this one for centuries; like so many unrealised technologies, it's apparently always just around the corner.

We lack fundamental science, such as no theory at all for artificial comprehension. That "first android baby" within 5 years is theater and fraud. (Which, incidentally could be Art.)

Well, "artificial" does come from "art" (ars + fex = skill + make)

> From Middle English artificial (“man-made”) via Old French (modern French artificiel), from Latin artificiālis from artificium (“skill”), from artifex, from ars (“skill”), and -fex, from facere (“to make”).

Regarding the artificial baby - I was thinking along the lines of putting the Gato [1] model in an android body. Nothing wild technologically, just give it the experience of having a family, friends, things to do, like we do for kids. This will induce the agent to form the same kind of value judgements as people make. It won't be just trained on the whole internet, but have access to the real world + human society, the best training dataset.

I believe the lack of access to the real world, the human society and a humanoid body is what keeps AI models from being more human-like. We can't become a developed human without society, for example. Why would AI be able to? The next step in AI is not about the neural net architecture but the curriculum, especially real world exposure.

[1] https://www.deepmind.com/publications/a-generalist-agent

Reply on news.ycombinator.com