By admin 
For several months, something discomfiting has been creeping onto San Francisco’s lampposts, bulletin boards, and restaurant windows: more and more signs, for run clubs and a July 4 parade and tacos, all of which are obviously entirely produced by generative AI.
Read more Resistance 1.0 Was Not Crazy
This bothers me terribly. (Friends have called it a fixation.) But why does it bother me? A ChatGPT-generated poster for a garage sale seems legitimately dispiriting compared with a handwritten flyer, or even a haphazardly formatted Word document. What about a bodega using an AI-generated image of a hamburger instead of a random piece of clip art pulled from the internet? It’s not like most of these flyers were ever going to be high art anyway. Still, the issue nags at me constantly, whenever I pass one of these synthetic images plastered over the real world.
Last month, in part looking for reprieve from the AI onslaught, I went to what promised to be an imaginative new exhibition of Henri Matisse and other fauvists’ work at the San Francisco Museum of Modern Art. The show was centered on Femme au chapeau (“Woman With a Hat”), a foundational painting in modern art. Matisse depicted his wife with colors that were totally detached from reality: teal patches of skin, an orange-and-green neck, a phantasmagoric hat. In 1905, when the painting debuted, this was scandalous: “A pot of paint has just been thrown in the public’s face,” one critic declared.
Walking through the galleries, I was taken aback by what struck me as a scandal of its own. The museum, in collaboration with Google Arts & Culture, had used generative AI to augment visitors’ experience—in ways I found, at least initially, rather off-putting.
Part of the exhibition’s conceit was to re-create Femme au chapeau’s historical context. The museum commissioned a local artist to develop wallpaper that could have been in the palace where the painting debuted and worked with an animator to create a visual projection of the walls of Gertrude and Leo Stein’s living room, where the canvas hung for many years. I loved all of that. But an AI-generated animation of the streets of Paris circa 1905, with shadowy figures walking to the gallery and a dirigible floating above, felt more like a gimmick than historical immersion. The museumgoers around me, however, seemed awed.
Then, near the gallery’s exit, came the second AI experiment: a row of screens, each showing an expanded Matisse painting made with Veo, Google’s video-generation model. A painting of a seashore became a 3-D rendering in fauvist style; a portrait of a girl with green eyes was reimagined to place the girl in Matisse’s studio.
The fauvists painted the world with flat slabs of highly saturated hues, delighting in color for the sake of color rather than realism. Using AI to move beyond the frame and give the images “realistic” depth seemed antithetical to their original vision. I recalled childhood memories of being entranced by various Matisse works at the Museum of Modern Art in New York City, marveling at his Dance and The Piano Lesson. The AI videos alienated me rather than inviting me in.
Yet maybe I’m just a grump? I am at least open to the possibility that my response is not so different from that of the pot-of-paint conservatives in the early 1900s. So I reached out to the museum’s chief curator, Janet Bishop, to learn more about these AI experiments—how they’d been conceived, what effect they had on the show, and what role AI broadly has to play in traditional art museums. We met at the galleries, where Bishop told me that curators wanted to “give a sense of what early-20th-century Parisians or visitors to the Grand Palais would have experienced in approaching this imposing exhibition hall and then going into the sprawling 1905 salon.” The AI installations resulted from extensive human research and oversight. I came away, to my surprise, not unconvinced of these experiments’ utility—though still a bit skeptical of their tastefulness.
This conversation has been edited for length and clarity.
Matteo Wong: How did your collaboration with Google Arts & Culture come about?
Janet Bishop: It was early in 2026. We were fairly far along in terms of planning for the exhibition. We had secured all the artwork by that point and were in the process of developing a couple of the ideas that we had for how to situate visitors, such as by re-creating the gallery where Matisse’s painting made its debut at the 1905 Salon d’Automne in Paris and the successive scandal.
Our original idea for the opening gallery, where we wanted to bring our visitors back in time to turn-of-the-century Paris, was to project a lantern slide that is an image of the Grand Palais. Then our museum director, Christopher Bedford, and head of education and community engagement, Gamynne Guillotte, got back from a really inspiring trip to the Google Arts & Culture offices in London and proposed the idea of setting still images in motion.
Through our curatorial research, we had postcards, still images, interior shots, exterior shots, and other material to work with. So in what became weekly meetings with Google Arts & Culture and project team members, we reviewed the images we had and developed this film.
Wong: What does having the animated film provide to a visitor that projecting the lantern slide and displaying postcards wouldn’t?
Bishop: We have some archival displays throughout the exhibition. But what we’re programming for in this entrance gallery is to give people a sense of what it felt like to approach the Salon d’Automne in a more immersive way.
Read more Ukraine Planned to Swarm Moscow Airports With AI-Guided Drones
Wong: Someone could look at this AI animation and its imperfections and perhaps find it kitsch. Were you worried about that at all?
Bishop: “AI-generated” makes it sound like AI did it by itself or that this was quick and easy. The project was very labor-intensive, and everything is historically accurate. One moment that was especially exciting to me was when we were looking at a draft, and there were all of these figures that were moving away from the building, and I said to the Google engineer, “Oh my gosh, it would be so amazing if we could have them walk toward the building,” and she said, “No problem.”
Wong: Is it harder now to get people to come here and engage with the art, especially younger folks? Or is this something that would have been just as helpful decades ago?
Bishop: You were talking about your experience as a child going to MoMA, and now we recognize that people have a lot of things that command their attention. One of the challenges in getting people to art museums is just pulling people away from their screens.
People are used to experiencing content in a really wide variety of ways. I know that it is important to our director that, given our proximity to Silicon Valley, we pay attention to developing technology and how it might enhance visitor experience. I loved the idea of the original lantern slide as a projection, but I can say that I think this film does something different in terms of giving our visitors a sense of the energy of Paris around the turn of the last century.
Wong: I’d also like to talk about the other AI collaboration with Google in the exhibition, near the exit. Can you tell me the origins of this one?
Bishop: The thought for what we called the “Experiment” was to enter into the world of a select number of Matisse paintings and imagine the context for those paintings beyond what is represented on the canvas. For instance, with The Girl With Green Eyes, we don’t know who the sitter was, but we do know that it was made in Matisse’s studio. So in the Experiment, you see the figure get up and walk across the studio. The details of the studio are consistent with his actual studio, which—we know what it looks like. And she looks out the window across the Seine.
Wong: Clearly a lot of visitors are enjoying this, but I wanted to ask why you thought it was important to help people imagine the world beyond the frame. My initial reaction, if I’m being honest, was that there’s a frame for a reason, and that the paintings aren’t naturalistic for a reason. What was so shocking 100 years ago was that these were, like, just flat planes of color, and that the colors were the focus. To turn that into a three-dimensional seascape seems almost wrong or off-putting.
Bishop: It’s part of our interest in engaging with different sorts of learners, people of different ages. They’ve been very popular with kids, and these images of paintings might encourage people to look more closely at the actual paintings. We’re experimenting, and I think experimentation is very definitely what we embraced with these features within the exhibition.
Wong: You mentioned earlier that you were excited by changing the historical AI video to have visitors walking toward the palace, rather than away from it. Was there a similar moment of excitement here?
Bishop: The big surprises came for me in the first piece, where in terms of the ability to conjure an experience, the use of AI as a tool made us more effective in achieving our goals than would have been possible without it. So that was really actually quite exciting. I think the Experiment is a little bit less integral to the story but an engaging feature that I’m happy that we pursued.
Read more The New History Wars