Convoy image heading to Rafah Crossing

https://www.reuters.com/fact-check/arabic/ICIT4BGBBBPKDMMUVCUOPZLFIU-2025-06-21/
I reviewed an image said to show the “Convoy of Resilience” en route to the Rafah Crossing. Repeated clothing patterns, distorted flags and irregular lighting pointed to diffusion-model generation. The convoy organisers confirmed the image was fabricated.

Giant anaconda in the Amazon River

https://www.reuters.com/ar/fact-check/VTFNSJV2DBKTHB4GURXO5JH72A-2025-06-03/
I analysed a viral clip claiming to show a massive anaconda surfacing beside a boat. The video showed two heads, floating objects that lacked weight and a camera view that did not match the on-screen overlay. These details confirmed synthetic video.

Gaza newborn video

https://www.reuters.com/fact-check/arabic/MLLYWEB7SBLJFF6IJTH63EJZNU-2025-05-25/
I assessed a video that appeared to show a newborn baby from Gaza smiling and raising a victory sign. The child’s fingers changed in number and length between frames and the facial movements were unnaturally smooth. These are known markers of generated video.

Cow tied to a missile

https://www.reuters.com/fact-check/arabic/BTC72XP3VZOIHD4SQIYRWSZVV4-2025-05-29/
I reviewed an image that claimed to show a Pakistani missile with a cow tied to it. Texture repetition, inconsistent shadows and distorted insignia pointed to generative image synthesis rather than a real scene.

Deepfake of Avichay Adraee

https://www.reuters.com/fact-check/arabic/535CSPX6GRIOPF6FJGWGUXUQVE-2025-03-11/
I fact checked a video presented as Avichay Adraee mocking a Syrian official. The lip movements, cloned voice and repeated facial frames matched common deepfake behaviours.

New York tunnel flooding video

https://www.reuters.com/fact-check/arabic/YU5T5BJW2RNODBZ62XPO2WLC7A-2025-08-07/
I examined a video said to show cars sinking in a New York tunnel during July floods. Repeating car designs, twisted metal rails and people who appeared to merge into the scene all signalled synthetic generation. The clip originated from an account known for AI videos.

Collapsing building during US floods

https://www.reuters.com/fact-check/arabic/76BEGAR3ZNPQVERTGZKPHZ3HQY-2025-08-25/
I looked into a video that seemed to show a building rocking and falling into a river. The motion lacked physical realism, the windows stayed perfectly intact and the riverbank warped between frames. These are common traits of AI generated disaster footage.

Aquarium tank collapse

https://www.reuters.com/fact-check/arabic/TESY7IW25NNL3MQM35SFOEXPZI-2025-08-26/
I analysed a shopping-centre video that showed a huge aquarium bursting open. Escalators stayed perfectly still while water moved around them, the glass dissolved into liquid and several people blinked in and out of the scene. These cues showed the clip was synthetic.

Qatar vs UAE football match image

https://www.reuters.com/ar/fact-check/WRYD5EZA7RIPNAOHQYOA2ZXKPQ-2025-10-30/
I reviewed an image that claimed to show phones and car keys thrown onto the pitch after a football match. The shadows did not match, several phones replaced original shoes and the watermark had been edited out. The image was manipulated.

Massive protest in Japan

https://www.reuters.com/fact-check/video-does-not-show-massive-pro-palestinian-protest-japan-2025-08-15/
I checked a video that claimed to show a huge pro-Palestinian protest in Japan. Flag motion, crowd behaviour and banner text all showed irregularities consistent with AI generated footage.

The post Further AI fact Checking appeared first on J. Rosenbaum.

I am always happy to lend my expertise to fact-checking images and will continue to do so.

https://www.reuters.com/fact-check/new-york-subway-fire-victim-misidentified-online-2024-12-27

https://www.reuters.com/fact-check/ai-generated-image-did-not-predict-new-trump-administration-2024-11-25

you can fact-check images for yourself by being critical of the media you consume. Look at all of the elements, the background as well as the foreground. See if something doesn’t fit. Gone are the days of looking for the number of fingers. Fact-checking now relies on more details than just looking at hands. Some generative AI images are too polished and sharp, some have specific artifacts, but they all have little tells. Look for things like floating parts, repeating patterns or parts that melt.

Remember, if something is too good to be true, it probably is. Some images are very obvious, others are more challenging. The image of the woman, for example, has none of the classic generative AI “tells”. It is made using a different system that specializes in faces. StyleGAN2 is an older AI but you can test it, and your fact-checking for yourself at thispersondoesnotexist.com. The trump image was easier. The pants, the eyes, the hands, but also the composition which does not fit images of that age. The name of the “artist” was also a dead giveaway! Always check images for yourself and learn to read the tells of AI images. The best way is to practice!

https://www.reuters.com/fact-check/ai-generated-video-purports-show-apocalyptic-scenes-los-angeles-wildfires-2025-01-16

https://www.reuters.com/fact-check/ai-created-image-ice-officer-seizing-child-shared-authentic-2025-02-18

List of Additional Fact Checking Links:

www.reuters.com/fact-check/arabic/43HRUINGSZPZVIR2KPW5IDQMPY-2025-04-03/

www.reuters.com/fact-check/arabic/UFEZARY6RFJYNJSCKPPLHFN2WQ-2025-03-12/

www.reuters.com/fact-check/arabic/INZKVTZOLRLFLEYAWUXJZZZQGM-2025-04-03/

www.reuters.com/fact-check/arabic/Q6UASFE53VPILGAZ36EGLIYR6A-2025-05-08/

www.reuters.com/fact-check/arabic/DLCWXUI7BJJYZJXFEPCXKZNDG4-2025-04-23/

www.reuters.com/fact-check/arabic/3P23QTXXEBJSTHGGK2K5M7NQIE-2025-04-10/

https://www.reuters.com/fact-check/arabic/N75F36TWUZLODDN2XKQNXRKTO4-2025-05-23

Shahrouz Tariq, a researcher at the Digital Research Authority (Data 61) at the Australian National Science Agency and specialised in electronic security and artificial intelligence, said that the picture carries indications of manipulation.

“The lighting and shadows around the plane are not as consistent as expected in a normal way. There is no natural reflection of fire on the runway if we take into account the intensity of the fire. The location of the name “Yemeni” on the plane is strange and overlaps with the windows of the plane compared to its location in the real pictures of the plane.

Jay Rosenbaum, an artificial intelligence researcher and a doctorate from the Royal Australian Institute of Technology, said that the passenger windows are strangely configured and the lines of some panels on the fuselage appear to be disconnected in a way that appears in the image produced by artificial intelligence.

“The logo of Yemeni Airlines is usually under the windows, but even when it is at the top, it is not as close to the windows as shown in this picture. There are two lines for the plane door, but they don’t meet at the top. The flames and smoke look cinematic… And the plane is rather small.” (Translated by apple translate)

The post The Importance of Fact-Checking AI Images appeared first on J. Rosenbaum.

AI slop is proliferating on social media at an alarming rate. Images are generated and disseminated with political agendas, particularly in right-wing spheres. These AI-generated images often depict soldiers, sad children, or interior designs. Of particular note are the catfishing-style “I need a husband” posts featuring women with impossible proportions, ostensibly seeking partners. These chimeric creations are bot-driven posts designed to farm engagement, but they also offer a chilling glimpse into AI-driven beauty standards and the lack of diversity in AI-generated women.

Artistic evolution

this work initially started off as a commentary on beauty standards but has evolved after my research led me to find an AI slop alt right pipeline on Facebook. As generative AI is emerging as a fascist aesthetic, I have interwoven all of these themes by contrasting the classical statue aesthetic together with AI slop.

This work has been explored through projection mapping, 3D printing and video works as part of a residency for Centre for Projection Art in Collingwood Victoria. the evolution of this work has enhanced the exploration of the themes of fascism and generative AI. these works are constantly evolving, showing how generative AI is constantly changing.

LoRA generated images

Ai slop catfishing posts reflect a mechanical view of the male gaze. However, an AI cannot truly comprehend the male gaze, and in its attempt to mimic it, it creates beings beyond understanding. This research project aims to analyze the patterns in these images, explore posting methods and engagement, and examine the meaning behind the images. It culminates in a series of works critiquing both the images and their creation and dissemination methods.

The artwork I am working on problematicizes this relationship between real and fake, the generated chimeric women and the relationship between classical sculpture and fascism. I trained a Learn On Reconstruction and Attention model, also known as a LoRA, on this AI slop and set up a production line. In this work I generate figures as marble sculptures, transform them into 3D models, and then manipulate them in Blender and AI video systems or 3D print them. This grotesque imitation of generative AI producing mediocre content is linked to the fascistic fascination with marble sculptures and the inherently misogynistic nature of how these images are generated and disseminated over time. . These works are both a product of the algorithmic gaze and a method for subverting it.

This work will take many forms including video works, sound and 3D printed works. I also intend to create more works with 3D printing, projection mapping and even interaction.

The grotesquerie is deliberate, and the plastic relates not just to the plastic nature of the generated images, but also to the fragility and toxicity of plastic and fascism. These works subvert the male gaze by being deliberately unappealing, raising questions about the ethical implications of using AI to manipulate and control the human form. AI derives its power from the collective choices of users, creating a bottom-up dynamic that makes it more difficult to regulate and control. The artworks show the problems of letting these continue unchecked, challenging the ideas of perfection. Generative AI and fascist ideals both promote a narrow and potentially harmful vision of human perfection. Fascist ideals were aspirational, while AI beauty standards are presented as attainable through technology, making their impact potentially more insidious. This work subverts those ideas of perfection, challenging what perfection means to us and to a world inundated by generative AI.

Paper and talk

Fascist Fascinations

Greek and Roman statuary, captured in marble, embodied godlike perfection. Fascist governments lauded these edifices for celebrating purity, whiteness, and idealized forms. Mussolini’s pallazo della Civilta Italiana, with its 28 travertine marble statues, exemplified this. Nietzsche’s Übermensch, centered on an idealized vision of human potential achieved through willpower and morality, epitomized this concept, often linked to physical prowess and racial superiority, and used to justify violence and discrimination. AI’s new idealizations of Eurocentric beauty and perfection, akin to those of Mussolini and Hitler, mirror these exclusions. AI slop has found a home in right-wing circles, showing idealized forms of Trump, the IDF, and sensationalized natural disasters . Generative AI, with its kitsch and avant-garde nature, combines the two to create a pastiche. Clement Greenberg proposed that kitsch is schlocky and cartoonish, opposing avant-garde modernism. However, John Gantz argued that fascism is inherently both avant-garde and kitsch, making it difficult to separate the two. At its core, AI slop and propaganda are closely entwined, a prime target for fascist exploitation.

“Fascism is always both: avant-garde and kitsch, and it’s difficult to fully divorce one side from the other. The avant-garde abstraction into attractive notions like “power” as such obscures just how ugly and stupid power really looks like in practice” (Ganz 2022)

This work has evolved to become an extended video piece. The video works are being shown at Fed Square Melbourne and Bunjil Place Narre Warren. It will culminate in a showing of projection mapping and 3D printing at Collingwood Yards. Slopocene Icons is about the horrors of AI slop and its ties to fascism. It is unapologetically a work of grotesquerie and destruction

The post Slopocene Icons appeared first on J. Rosenbaum.

Through this work I hope to explore the connection between AI bias and interpretation of identity. What are the most common terms that the AI uses and why? are people represented faithfully by these systems or does the AI frequently miscategorize and underrepresent them? I want to encourage users to explore why AI sees them in certain ways and interrogate that connection. What does it mean when the AI changes who you are?

I want to encourage people to discuss the biases inherent in these systems and consider how much they trust them. In a way, I am encouraging “AI hallucinations” by reducing the information that the AI system receives. This creates a strong basis for it to wildly experiment and make up the missing data. As users of generative AI rely on these systems to help fill in incomplete data, I feel it is fitting to show how they fail to extrapolate from missing data. Especially when it comes to classification of living people.

Future Directions

I also want to expand this to environments to explore the inherent colonialism of generative AI. AI systems are notoriously biased, and generative AI systems are particularly Eurocentric. This directly contributes to the colonialist narrative of generative AI systems as they continue to expand across the internet and the utilities we rely on. This colonization of generative AI is an area I want to interrogate in this work as an installation in one of the many beautiful parks in Melbourne. Can we de-colonize a colonial structure by training images on local flora and re-interpret the space? what would that look like?

This is very much a work in progress at the moment and I hope to update it with more results as I explore and experiment. This work will need further testing with more people and locations.

The post Fractured Reflections appeared first on J. Rosenbaum.

In ‘Memento Mori,’ I explore the consequences of our increasing reliance on AI and technology. I particularly explore this impact in the context of environmental degradation. By capturing human presence and obscuring them through repetitive tracing, I highlight the ephemerality of human presence in the face of environmental harm. This artwork serves as a reminder of the need for accountability and awareness in our relationship with technology, urging viewers to consider the environmental impact of their actions.

Inspiration

Memento Mori comes from the Latin remember you will die. It is a recurring theme in art over the centuries. In this work, I use it as a memento for the earth, remember our environment can die and our world can die. I drew from the work of Rembrandt re-using his etching plates and reconfiguring the artwork each time but in the process completely destroying the ability to reproduce the previous work. The plates were a limited resource, much in the same way our natural resources are limited. As this is a plotter work I also drew inspiration from one of the pioneers of generative art, Vera Molnar. Molnar’s work was a particular inspiration during some of the more challenging aspects of this work’s development, forcing me back to the drawing board.

Evolution of the work

Memento Mori started as a plotter-based artwork where a plotter would draw and re-draw over and over again rectangles representing a human presence and thereby occlude the glass plate with presences that become meaningless over time. This work has now evolved into a projection-based artwork with an evolving aesthetic where human presence is reflected in photographs of nature, stamped into rectangles with bright outlines. Over time, the nature images are slowly replaced by generative images. The generation is made using a lightweight local AI and trained on photographs taken by the artist. The presence of a person moving into range of the camera is tracked via a purple rectangle that fixes and fills in with nature and becomes bright green.

The outlines fade into the work until the nature is occluded by the chaos of humanity’s presence and the machinic signature of the neon green overwhelms the work. Over time, viewers will see the work become more chaotic and involved, showing how our actions impact the environment, especially over time.

Plotter Development

This work began, as all my work does, in code. And in the idea of segmenting and drawing the outlines of humans seen by the AI. I began with webcam capture, then segmentation, edge detection, png generation, svg tracing then finally sending to the plotter via gcode. There were multiple problems in the execution of this work, from the wrong plotter to the overall aesthetic so this work evolved in appearance while maintaining the overall impact and concept of the piece. From segmentation I moved to YOLO and drew bounding boxes instead which allowed for a more aesthetic final result.

Fort Earth Exhibition at 45 Downstairs

FORT EARTH is a curated exhibition + project providing the opportunity for visual artists and creators to develop personal or collaborative artworks with sustainability and the environment in mind. Artists came together to create a massive exhibition using environmentally friendly materials.

Memento Mori was shown in an experimental capacity to test the system under a large influx of viewers.

The post Memento Mori appeared first on J. Rosenbaum.

AI system interconnectivity

Our project employs a trained classifier that assigns these words to users, followed by Stable Diffusion and controlnets to generate an AI impression of what a person embodying those words might look like. However, because we should never just accept what an AI has to say about us, Self_Portrait incorporates a second stage where users can choose to either embrace the assigned terms or modify them to better reflect their true selves. The system then generates a new portrait overlays it onto the viewer’s face using augmented reality.

Central to this artwork is the use of AI. It is through the intricate workings of AI, just like the complex nature of gender, that we examine the assumptions AI makes about us. Each day, various AI systems make snap judgments about us based on our online behavior, writing style, music preferences, and more. These systems create individual profiles that may not always align with our self-perception. This profiling, along with reinforced biases, can have far-reaching consequences in domains such as job applications, the justice system, and financial services. While our work cannot encompass the full breadth of AI’s access and overreach, it serves as a reminder of how easily we take for granted the labels assigned to us by different AI systems.

To develop the classifier, I collaborated with chatGPT, leveraging its expertise to compile a list of words associated with gender. We used Stable diffusion to generate batches of faces using these terms, enriching the dataset. This process aimed to shed light on the AI’s perception of gendered words, rather than solely relying on my assumptions. Through techniques like stable diffusion we can engage in a dialogue with the generative system, revealing its assumptions and biases on a broader scale. Incorporating AI into various stages of development, from ideation to programming, allows for a comprehensive examination of AI’s limitations.

About labels

At its core, this work revolves around labels and the choice to accept or reject labels imposed upon us. It was inspired by a thought-provoking conversation I had with my family about labels. In the past, society would often impose labels upon us, leading to ostracization. However, today we have the power to claim our own labels, which can help us find communities and connections. While labels can still be sources of exclusion, we now take a more active role in choosing, accepting and embracing labels. Self_Portrait encourages us not to passively accept labels bestowed upon us by unknowing and indifferent machines. It encourages us to consciously choose our own labels. Witnessing people interact with the system, laugh at the assigned labels, and then carefully select their own demonstrates the true essence of this work—the genuine care and curiosity we have for the meaning these words hold for us.

Collaboration

Ironically, as a design lecturer, my expertise in user design experience is limited as I usually work with code in the terminal. Fortunately, I had the privilege of collaborating with Melanie Huang, an exceptional technologist and creative. Together, we transformed my ideas into a cohesive and engaging artwork. The collaborative journey was truly wonderful, and I am forever grateful for Melanie’s contributions to the project.

The post Self_Portrait appeared first on J. Rosenbaum.

In Gender Tapestry, J. Rosenbaum investigates the fascinating connections between facial recognition, gender classification, and color perception. Similar to how different computer monitors present colors in varying ways, J posits that individuals perceive and experience gender uniquely. They explore the idea that our life experiences, cultural influences, and personal perspectives shape our understanding of gender.

In their innovative approach, J equates color with gender, seeking to provide individuals with a personalized blend of colors instead of conventional gender classifications.

J’s artistic practice thrives on orchestrating multiple AI systems within a complex network of multi-modal interaction. Throughout their research, J critically examines the limitations and exclusivity of gender classification. While Gender Tapestry initially aimed to revolutionize the perception of gender classification, it has also led J to a more enlightened understanding of gender itself—a better way to appreciate its fluidity and individuality.

To participate in Gender Tapestry, users have the opportunity to submit an image of their face. This image is processed through a multi-label classifier capable of working with various pronouns. Instead of receiving a traditional gender classification, users are presented with a custom mixed color based on the RGB color model. This unique color becomes a representation of their gender within the project.

The custom color is then utilized to create a simplified, Warhol-esque image that includes both light and dark variations. These generated images are incorporated into a diverse library of images—a dataset used to construct a vibrant mosaic of GAN-generated artwork. However, the user’s image serves a disruptive role within this process. It is introduced into the GAN to influence and modify the resulting output, adding an element of personal expression.

As more individuals are classified and participate in the project, additional colors are introduced into the library. This growth prompts both the GAN’s generated results and the mosaic itself to expand in complexity and diversity. The evolving mosaic becomes a dynamic representation of the collective experiences and identities of those engaged with the system, continually transforming and reflecting the richness of the gender tapestry being woven.

Gender Tapestry at ACMI

Gender Tapestry at Execute_Photography

As part of Photo2024 Gender Tapestry will be showing at Execute_Photography at RMIT Gallery. This is a chance to see the work as it operates longterm over the course of two months. Installation view photographs, execute_photography, RMIT Gallery, Melbourne, 2024. Photo by Christian Capurro.

I will be showing Gender Tapestry at NeurIPS 2023 and Execute_Photography. In the meantime, you can explore your own gender color via the website and see what the AI has to say about you.

Gender as a 3D colorspace

Try Gender Tapestry

You can receive your own gender color classification online!

Gender Tapestry

The post Gender Tapestry appeared first on J. Rosenbaum.

Conceptual piece

‘Frankenstein’s Telephone’ is a conceptual artwork created using artificial intelligence to examine how computer vision expresses gender classifications. Each of the images in the sequence  are snapshots of that document how various neural networks have generated an image based on a prompt phrase. The steps taken to create the sequence are as follows:

1. Select an image caption from the Broadly Gender Spectrum Collection. 
2. Enter the caption into ATTN GAN, a generative adversarial network that creates an image based on the caption.
3. The image generated by ATTN GAN is then uploaded to DeepLab, an artificial intelligence algorithm that segments and classifies parts of an image as people and types of objects.
4. The resulting abstract segmented and classified image is then fed to SPADE-COCO, a generative adversarial network that generates a new image based on the segmented labeled image.
5. The resulting image is then fed to IM2TXT, a neural image caption generator. The final screenshot is the desaturated image with the resulting caption. 
6. REPEAT using the resulting caption until the caption no longer references a person or people.

Research discussion

Through this process, I want to consider how neural networks and machine vision deal with gender and how the design of these networks and technologies reflect the bias and perspective the humans who designed these artificial intelligence systems. All AI systems are trained using a dataset selected and curated by people and therefore reflect inherent bias. What is interesting is interrogating how the resulting AI responds to its training parameters and generates outputs that evolve and shift over time.

This work transforms the game ‘telephone’ into an AI version. Players pass on a message by whispering to their neighbor, who then conveys their version.

This corrupts and changes the original meaning in a humorous way. In ‘Frankenstein’s Telephone’, I am exploring the shifts pertaining to gender and personhood. The sequence of different AI systems reveals aspects of how AI classification works and, potentially, where bias occurs and which parts of the chain are weakest.

Red, assigned in DeepLab does not represent a gender. It simply stands for “person”. I believe that the simplicity of that classification category is what we should aspire to. These works reveal that the researchers responsible have intentionally programmed some of these AI systems to ascribe gender. The discomfiting images of computer generated humanity calls to mind the story of Frankenstein’s monster, a narrative that many queer people, especially transgender people, identify with. The real monsters are the people behind the scenes that unconsciously perpetuate bias. Humans create and label the datasets used in these works and humans are responsible for the gendering and misgendering that appears. The captions use gendered language written by humans and interpreted by machine. This is where the bias comes from, this is where the mistakes and misgendering occur. These machine generated conceptions of humanity, with their unexpected forms and eyes and mouths could be seen as modern monsters, but no less deserving of their label “person”.

Video

a screen recording of the interactive site in action

While more images are available on the website, this is a video slideshow of images in sequence.

Exhibition History

Frankenstein’s Telephone ran as an interactive project until 2023 and exhibited as part of Future Histories, And/Or and AI Perceptions of Gender and is still available in the virtual gallery space on Spatial.

The post Frankenstein’s Telephone appeared first on J. Rosenbaum.

Artist Statement

A series of marble faces, generated by AI, as it learns to create and update its bias’ on gender. First it is trained to generate masculine marble faces, fixed, immovable. Then marble feminine faces are added. It learns to change, a transgender neural network, updating its knowledge and its experience as it goes. Then the marble starts to give way, non-conforming self expression, color, joy, emerging as the gender becomes unfixed, non-binary. Through this evolving conceptual artwork I am exploring how my machine creates images representing gender and if it holds onto its trained bias. I am examining what the gender shift looks like and what transgender and non-binary self expression and self aware aesthetics are beyond biological essentialism. The datasets used by the machine are created by me using 3D modeling and echo the idealized masculinity and femininity of classical statuary with a more inclusive bent. I am purposely training bias into my machine and then attempting to unbias it and see how the artwork evolves as the machine learns that there are multiple genders and gender expressions.

The works you see on the wall and the screen are moments in time in the training of the Deep Convolutional Generative Adversarial Network (DCGAN) as I manipulate the dataset it relies on to challenge its bias. As it learns it generates sample images to show its progress and understanding. The samples are entirely represented in the video to represent the full training of the neural network to date. The works on the wall consist of the sample rendered into a mosaic of a face. They are chosen by the application. This is part of my ongoing research project and is an evolving conceptual piece.

Dataset Development

My work is classically inspired in aesthetics to create a resonance with history and historical images. There is a satisfaction to tying the latest in technology with the depth of historical artwork and the anonymizing sense of marble. I desire to create an imagined history for gender minorities, bringing visibility to those largely unrepresented in art history. This involves crafting a historical context for a modern mode of artistic creation.

I used the following multistep method to create the dataset:

1. Mold 3D models of faces or bodies in Daz3D.
2. Create a camera attached to a null object in the 3D workspace that allows stable rotation.
3. Create a large animation timeline of 1000 frames.
4. Create a keyframe at each 100 frames to morph the 3D model into another form.
5. Create an animation sequence where the camera moves around, up, and down to capture as many angles as possible per 100 keyframes.
6. Render the output as images.
7. Flip the images in a bulk image editor to increase their number and present a different view for the GAN.

DCGAN Version

In my case I trained these as single works in the GAN rather than a grid, the gridded training appears to require more epochs per train as the dataset is small. I am using 1000 images per gender. I created the images in the dataset using 3D modeling and renders to gain different angles and a variety of physical features as the figure morphs between different sculpts and appearances. The marble texture recalls classical sculpture, but also the rigidity of stone of being cisgender and of the opinions of some people about the supposed immutability of gender. The addition of color symbolizes the breaking free of gender constraints, of leaving the stone behind and gaining warmth and color. As the series progresses I anticipate adding further transgender aesthetics in full color, having completely shed the restrictions of stone to realize your self. 

Installation Images

These works are work in development, first shown at Future Histories as a mural and a video work showing the training process. This work was also a finalist in the Midsumma and Australia Post Art Prize

The video shows the full sequence of the DCGAN training and the disruption of data:

ProGAN Version

The next neural network that I used to build on this style of work was ProGAN. With a progressive growing GAN, each sample builds on the one before it, so that instead of generating thousands of individual images, it works with a grid of images and builds on each face in the grid, dividing the pixels until the image is more and more resolved. This process allowed for works which were more suited to a smooth video transition. While the processing is more intensive, each video shows the transition with little to no pixel disruption. This aligns well with the projection work I developed for the Gertrude Street Projection Festival. I took extra steps, dividing each grid into separate parts and carefully disrupting the training. If live manipulation in the DCGAN training was labeled disruption training, this form could be more accurately termed subversion training, as the new data is stealthily renamed to subvert and replace the existing data.

Judith Butler discusses subversion in her book Gender Trouble where she explores making gender trouble through subverting and displacing masculine presences. 

I used ProGAN to create a looping projection based work at Gertrude Street Projection Festival. and Epoch Loop

The images and video below were taken by Bernie Phelan for GSPF 2021

This video shows the full looping project installed for GSPF.

Knox Immerse Library Installation

Set in stone was chosen for a large scale installation at Knox Library for Knox Immerse festival in 2024. I reimagined the mosaic as a large scale site-specific work for the space and integrated an Augmented Reality component.

I programmed the Augmented Reality to show the video as a transparent overlay onto the mosaic with some 3D cubes textured in faces to further activate the space. As this mosaic was larger than any I have previously made I also made a custom python algorithm to generate the mosaic at the correct size for printing. I integrated images from several different GAN training sessions to ensure there was enough variety in the mosaic images.

The post Set in Stone appeared first on J. Rosenbaum.

Picture a face manipulated in Snapchat, adorned with a computer-generated, melty mask—set against an artful background. My fascination with Snapchat led me to scrape images from Instagram, curious about what a neural network could create. The Generative Adversarial Network, instead of blending Augmented Reality components into nightmarish forms, astoundingly detected commonalities and erased these components altogether!

Enter the world of Generative Adversarial Network grid at mode collapse. Working with a DCGAN until it reached mode collapse provided intriguing results, with converging grids making for visually captivating works. During training, I experimented with Snapchat’s face picker, leading to both horrifying and hilarious outcomes. The process unfolded in a detailed Twitter thread as I navigated the intricacies of Snapchat’s features.

Moving forward, I employed style transfer on some images for upsampling and utilized resizing GANs and upsampling algorithms to ensure the works reached print size from their initial 128px.

Artworks

If you find joy in this project, spread the word, and when you use the lenses, tag me or use #snapchatgan so I can revel in your wonderful creations!

Snapchat codes

This work is based in snapchat and has several lenses to try out.

Installation photos

this work was exhibited at Melbourne Fringe Festival in the Trades Hall for the length of Fringe

Video

The post Uncanny Valley appeared first on J. Rosenbaum.