CPY: a_likeness_of_everything_and_nothing
《似是而非》
           |
14 June - 12 July 2025
Gallery EXIT presents 'a_likeness_of_everything_and_nothing', an ongoing project by CPY, an online artist collective. The members of CPY prefer to remain anonymous, as their personal identity adds to the misreading of pictures. Below is an essay submitted for us to publish. ___________________________ Let me start with a confession: I asked DALL·E to generate 'a self-portrait of itself contemplating its own existence' and it gave me a robot looking in a mirror. The joke, of course, is that DALL·E has never seen itself, has no self to see, and wouldn't recognize a mirror if it computed into one. But then again, neither did Velázquez when he painted 'Las Meninas', and we've been arguing about that mirror for centuries. The emergence of text-to-image models presents us with a delicious philosophical comedy--one where the punchline keeps changing depending on which theoretical framework you're holding. These models transform language into visual form through a process that would make both Plato and Baudrillard reach for the aspirin. They're trained on millions of images scraped from our collective visual culture, learning statistical relationships between words and pixels, then regurgitating novel combinations that have never existed yet somehow feel hauntingly familiar. So, are these images 'real'? Well, that depends on what you mean by real--a question that has kept philosophers employed since ancient Greece. In one sense, they're undeniably real: they exist as specific arrangements of pixels on screens, as electromagnetic patterns, as data. They have material presence in the world. They can be perceived, shared, modified. In this basic ontological sense, an AI-generated image of a 'cyberpunk sloth drinking espresso in a Venetian café' is as real as any photograph. But this is where things get interesting. These images are real without being indexical--they point to nothing that ever was. Unlike a photograph, which Susan Sontag called 'a trace, something directly stenciled off the real,' AI-generated images are pure simulacra in Baudrillard's sense: copies without originals. They're what happens when Benjamin's age of mechanical reproduction gets drunk on its own possibilities and starts making things up. The question of likeness adds another layer to this ontological lasagna. When DALL·E creates an image of 'Barack Obama riding a unicorn', it produces something that looks like Obama without being of Obama. It's a statistical average of 'Obama-ness' filtered through the model's training data--a Platonic shadow of a shadow, twice removed from any original. The likeness is both uncanny and empty, like a perfectly delivered joke with no comedian behind it. This brings us to the vexing distinction between an image and a picture. Here I'll venture a provocative claim: an image becomes a picture when it stops being purely potential and enters into a specific context of meaning-making. An image is raw visual data--those arrangements of pixels sitting on a server somewhere. A picture is what happens when someone encounters that image, frames it (literally or conceptually), and assigns it significance within a particular cultural, aesthetic, or personal framework. Think of it this way: every output from DALL·E is an image, but not every output becomes a picture. The vast majority vanish into the digital ether, forgotten moments of computational creativity. But when someone saves one, shares it, argues about it, sells it as an NFT, or prints it and hangs it on their wall--that's when an image graduates to picture-hood. It's like the difference between a joke and a punchline that actually lands. Is a picture more real than an image? In the phenomenological sense that continental philosophers love to invoke, absolutely. A picture has entered into what Heidegger might call the 'worldhood of the world'--it's not just present-at-hand but ready-to-hand, integrated into human practices and meanings. It participates in what John Dewey called 'an experience', becoming part of the fabric of lived reality in a way that a mere image, sitting unviewed in some cloud storage, cannot. But here's where our contemporary moment gets truly absurd: we live in what Paul Virilio called the 'vision machine' era, where images proliferate faster than meaning can attach to them. We scroll through hundreds of images daily, each one appearing for mere seconds before being replaced. In this context, how can any picture create or hold value? The answer might lie in what I call 'attention economics meets ontological scarcity.' In a world of infinite images, the scarce resource isn't the images themselves but the human attention and meaning-making that transforms them into pictures. Value emerges not from rarity of production (any prompt can generate endless variations) but from rarity of reception--those moments when an image arrests us, makes us pause mid-scroll, generates discourse, or becomes what the art world loves to call 'iconic.' This value is paradoxical. On one hand, it's more democratic than ever--anyone with access to these tools can create images that might have taken years of training to produce manually. On the other hand, it's more elusive than ever, as each image competes with millions of others for the scarce resource of human attention and cultural significance. The real joke might be that we're asking the wrong questions. Instead of wondering whether AI-generated images are 'real' or 'fake', we might ask what new forms of reality they're creating. Instead of lamenting the loss of indexical truth, we might explore what happens when images become pure potential, untethered from the burden of reference. After all, every era gets the images it deserves. The Renaissance got perspective and the fantasy of objective vision. The Industrial Age got photography and the dream of mechanical truth. We get AI-generated images and the vertigo of infinite possibility. The punchline? We're still trying to figure out if we're laughing or crying. In the end, perhaps Ludwig Wittgenstein had it right when he noted that a picture cannot depict its pictorial form--it displays it. AI-generated images display something profound about our current pictorial form: a moment when the boundary between human and machine creativity blurs, when images float free from their referents, and when the ancient philosophical questions about reality and representation get a software update. The ultimate irony? In trying to teach machines to see like us, we might be learning to see like them--in statistical patterns, latent spaces, and endless recombinations of the already-seen. Whether that's comedy or tragedy probably depends on your philosophical commitments. But one thing's certain: the show, as they say, must go on. ___________________________ |