AI image generators are making images that look like photographs, and it’s pushing you to ask what part of your work is skill, taste, or just access to a tool like Photoshop. That question hits even harder when a prompt can produce something that passes at a glance, whether it’s going on your website, a client deck, or a social feed.
Coming to you from Blake Rudis with f64 Academy, this plainspoken video treats today’s AI panic as a rerun of older art fights, not a one-off crisis. Rudis starts in the 1800s, when painters dismissed photography as mechanical and “too easy,” like the camera was doing the thinking. That complaint sounds familiar now, only the target has shifted from lenses and film to datasets and prompts. The useful part is how he keeps the focus on behavior: who feels threatened, who adapts, and who tries to freeze the definition of “real” at the moment they learned it. You end up watching the current argument with a little more distance, even if you still feel the heat when you see AI images getting traction.
Then the video lands on a concrete turning point: the release of the Kodak Brownie camera in February 1900 and what it did to exclusivity. Suddenly families could make pictures without training, without a darkroom, and without begging someone else to operate the gear. Pros didn’t shrug it off, they worried that ease would flatten the craft into “anyone can do this,” which is basically the same fear you hear today about automated image-making. Rudis doesn’t romanticize the old gatekeeping, but he also doesn’t pretend the pressure wasn’t real when the market got flooded with competent results. The tension he sets up is practical: when the tool becomes common, you have to separate “I can operate the tool” from “I can make something worth revisiting,” and those two skills don’t move together.
The video also jumps to the early 2000s, when digital capture and Photoshop got framed as cheating, with film positioned as the last honest method. If you started on film, you probably remember how quickly people stopped debating outcomes and started policing process. Rudis admits he was part of that resistance, which keeps the point grounded instead of smug. He frames the pattern as predictable: a new tool arrives, the prior generation calls it inauthentic, serious makers redefine their role, then the tool becomes normal and the fight moves on. That arc forces you to ask where you are in it right now, especially if your gut reaction is to defend the old lines instead of testing new ones.
Where it gets sharper is when Rudis pulls algorithms into the same frame as authenticity. He isn’t only talking about AI versus cameras, he’s talking about what platforms reward when the goal is quick reaction and constant scrolling. AI images often aim for instant punch and fast comprehension, which can do well in that environment, while human work can be slower, stranger, or more personal in ways that don’t read in half a second. That creates a real tradeoff: do you chase the kind of impact that performs immediately, or do you build a body of work that’s harder to copy and harder to forget, even if it travels slower. He hints at how some high-end fine-art shooters are already moving away from hyper-real looks toward more expressive choices, which is a risky move if you’ve built your identity on realism. Check out the video above for the full rundown from Rudis.
1 Comment
I find this article fundamentally off.
The novelty that drives social contagion in imaging has never been the art itself, it’s been financial accessibility. Cost reduction is the real catalyst, not some sudden philosophical shift. That was true with photography in the 19th century, and it was true again in the transition from film to digital.
During the film-to-digital shift, most of the loudest resistance didn’t come from people deeply fluent in film processing and printing. Many critics fundamentally misunderstood where image value lived. JPEG became the battleground because they mistook the rendered image for the photograph itself. In reality, data was the image. RAW files were the negative, and bit depth was the digital equivalent of chemical latitude. Once photographers understood that RAW preserved constraint and interpretive authority, digital didn’t cheapen photography; it accelerated it.
Phones followed the same pattern. Because small sensors couldn’t sustain image quality, AI was introduced to simulate photographic appearance with computational fill-ins for missing physical capacity. That wasn’t an artistic expansion; it was a commercial solution.
AI, however, is not photography.
Photography requires an originating event encoded as RAW data—light interacting with a sensor at a specific moment in time. AI models do not generate that. They mimic the statistical appearance of RAW values, but they cannot create them. That’s a limitation, not an evolution.
You can see this clearly in repeatability. You can’t produce an identical photograph twice because time is non-repeatable. AI also can’t produce the same image twice—but for the opposite reason. It has no time, no event, no causal anchor. Its variability comes from probability, not temporality. AI does not know time.
That’s why the claim that “the tool is deemed inauthentic” misses the mark. This isn’t a moral rejection of a new process. The issue is ontological. AI does not create in the photographic sense. It reprocesses existing representations. At that point, the historical analogy collapses.
Simply put, AI is massively disturbing cognition, while art traditionally does the opposite. That doesn’t mean AI is meaningless or useless. It has value in certain contexts. But when a tool removes constraint entirely, it undermines the conditions that allow meaning to accumulate. What remains is optimized output content, not art.
AI isn’t evil, it’s just flat. In the context of art, AI doesn’t reinterpret reality; it reprocesses representations of reality through probability.