10. Canvas of the Future

The inspiration is the policy implications. This abstract character is actually a Digital Still from a generated variation of the original Selkie. It reminded me of a sculpture, possibly a sort of Golden Globe or other trophy. It is layered on Saint-Armand handmade watercolour paper made by artisans since 1979. I chose that paper as a background because it’s important to me to understand the socioeconomic impacts of adopting Generative AI as an art form. Who will benefit? Who will close shop or lose gigs essential to pay the rent? That is the Canvas of the Future that is playing out before us.

If such a prize were to be handed out, what would be the judges’ criteria? How would they define what art is? Should they be interventions along the socio-technical pipeline and value-chain of data collection and algorithmic design? Recently, artists in the UK were warned that “overtly political or activist’ statements could lead to a withdrawal of funding”.Given the important environmental and socioeconomic implications of generative AI, shouldn’t AI art be required to vibrant and legitimate Digital Democracies? The AI for Good conference and the Information Telecommunication Union (ITU) in partnership with Shutterstock are holding a contest and the results will be known this spring of 2024. It will be interesting to see what criteria they choose. Will the environmental impact of AI be one of them?

The following best practices when curating art projects involving AI ethics have received wide-spread support globally. For the Art Impact AI petition, a Manifesto of sorts, signed by over 2000 people (artists, AI scientists, cultural workers, political economy and legal experts, etc), co-creating Responsible AI with artists means supporting and creating spaces for: 

1. Art that acknowledges the political dimensions of AI and contributes to the co-creation of responsible, equitable and sustainable AI and data governance.
2. Scenarios that foster a sense of “agency”, avoid dystopia or fear and panic.
3. Interactivity and immersion that facilitates learning together.
4. Projects that put inclusion as an added-value in design (for juries and funders, this could mean for example assessing who is leading and benefiting the most from the project, add that as a question).
5. Public art in all sorts of spaces, neighborhoods (less so in institutions).
6. Projects that recognize the plurality of knowledge sources in co-construction processes, and give credit.
7. Art that values authenticity and concrete objectives for their project such as Reconciliation, Human Rights and Sustainable Development Goals. 
8. Policies that respect the intellectual property and moral rights of artists as well as cultural intellectual property.

9. Whale's Breath

Inspiration: The original work (16mm film and watercolour) and the Digital Still that is then extensively overpainting digitally using Fresco are about self-care and that important cycle of effort followed by recuperation; about transparency and vulnerability. When overpainting the Still, I integrated Words contributed to our data set and that were related to water.

How: WHALE'S BREATH is made from an animation film of a woman doing yoga using a 16mm Bolex, an analog painted still, both which prompted an AI-generated variation, of which I pulled a Digital Still to paint over using digital paints. The generated outcome respects the initial intention of analog work done more than 10 years ago and follows the same yogic flow of movement similar.

Policy and Research: When do AI-generated images become Art? If you look at this digital painted still, what is its foundation? The work I started doing years ago, the layers I added to an algorithmic variation, or is the entire foundation of this corrupt because the use of generative AI in the process? For me, the answers to these questions vary depending on the painting. Some of the 12 pieces showcased in the Algorithmic Frontiers Exhibit are more on the scale of “corrupt”, or at the very least I wouldn’t dare register them for copyright. For WHALE’S BREATH, I would. I think I can rightfully claim that I have authorship on both the image and the video. What do you think?

8. Peace & Doves

Inspiration:Peace and Doves is inspired by the Words related to peace contributed by to our data set. The video below ends with a slow, visual poem about understanding our own strength.

How: This piece is entirely AI-generated. My prompt was "A white flying Peace Dove in the style of Pablo Picasso". RunwayML's platform as a "Train your own" option which I used so this image is a result of AI "Art" that is trained* using my art (non-generated, real old style paint, pencil, analog film). *Partly trained. Learn more in the section How does Generative AI work. The video includes non-AI art as well as a number of text+image to video generated pieces. As in other videos, I often export a still and work on it in digital painting apps such as Adobe's Fresco or Sketches.

Policy and Research: As I was working through numerous versions of a Peace Dove, I was also mulling over the following questions: What are the implications of asking a Generative AI system to copy in the style of a deceased renown and wealthy artist who used women's bodies for his inspiration? Is it less of a steal if I prompt an AI to imitate his style and sell it on a t-shirt? I put a sample t-shirt below. Let me know your thoughts! Is there such a thing as an AI Art Robin Hood and if so, would that make it ok to use art works without permission and, at a global scale, what would that imply?

7. Speak

INSPIRATION: This Digital Still was inspired by…frustration! Fear mongering tech millionaires are given so much power while concrete policy solutions put forward by experts in AI policy are barely heard. SPEAK and the video are my shout out to humans working to make AI less biased, to make digital economies more equitable and sustainable, to those who have a position of power and seek the systemic changes that are required to make AI and more holistically, digital economy, work for more than a handful of people, cities and countries.

HOW: I used my art a visual prompt in text-to-video by RunWayML. Every time, the generated video goes off course, I take a digital still (one image from the video) and digital paint over it. I then edited those images and videos together. The generated part in this image is the porcelain doll, the rest is digitally painted. Speak is an image exported (Digital Still) from a video generated using a watercolour as an image prompt. The intention was to animate the abstract portrait, but the AI model couldn't follow the contour of the eyes, face, mouth, and other details. It generated an 3D rendering of a porcelain-like doll instead which I painted overusing a digital painting app to keep more in spirit with my own intent.

POLICY AND RESEARCH: The making of this video coincided with a period in which I felt that not only was this AI not listening to my prompts, but that gender equity wasn't deemed a priority in AI strategies despite a solid body of research and policy work that demonstrates it should. This was created in the fall of 2023. It is now February 2024 and the technical improvements on Generative AI platform are noticeable, those in policy, less so.

I advocate for a diligent oversight of the multifaceted effects of generative AI and advocate for pre-emptive measures to comprehend and mitigate adverse outcomes. This could involve establishing a cross-departmental team or inter-ministerial committee dedicated to ensuring that the perspectives of marginalized and equity-seeking communities are considered in AI governance. Such initiatives could be implemented prior to the enactment of regulations or in jurisdictions resistant to regulatory measures. These efforts can be undertaken by governmental bodies, the private sector, or non-governmental organizations.

Click here to download and read the letter I sent to Ministers of Digital Transformation, Economy, Finance and Labour (all men). Feel free to use it as a template to write to your own MP.

6. My Body,
My Data.

Inspiration: I made this watercolour with algorithmic variations for the cover of a presentation for the Canadian Fertility and Andrology Society's Gatekeepers' Dilemma and Ethics and Law Summit early 2023. For this particular conference, I was looking into uses of AI in reproductive health.

How: The Digital Print is a variation of my watercolour using Dall-e's out-of-frame tool. This means that the middle part is my watercolour while the outer edges were generated. I prompted it with words such as “an abstract watercolour on the theme of reproduction and fertility”. The video: I made a watercolour when I heard that a friend of mine was pregnant with her first baby. I prompted PhotoLeap with low-level variations which means the variations are closer to my work than a variation at the other end of the scale. The video is a compilation (using VideoLeap) of variations of that Matryoshka-inspired watercolour. The backdrop is from my 16mm handpainted animation worked + generated variation.

Policy & Research: AI can help discover new cures, saves lives but it could also resemble a Black Mirror episode especially for women, women of colour, Indigenous women and trans men. The lack of data in health translates to AI systems that don't understand womxn's bodies and health. If men are designing the AI tools to save lives, will the lives saved have a gender and a colour?

5. Algorithmic Molds

Inspiration: Algorithmic Molds reflects on identity and how technologies can enable, or censor, how we tell our stories, and Generative AI will augment which stories are told. As discussed under other pieces, namely SPEAK, freedom of expression is not equal for all users in Generative AI platforms. If I chose this piece because it is a positive representation of who I feel I am (I'm a sea person), don't be fooled. Most representations had to be discarded or heavily transformed. The video includes other examples of selfie variations, some where you'll notice are repainted, adding layers of digital paint and ink, and others highlight what variations of my selfie look like when prompting for an older woman.

The video below also includes a collage of failed attempts: a drawing I made of a young girl into RunWay ML or VideoLeap, and their algorithmic variations undress her, literally take her pants off, another grows breasts and another ends up in a small white bra. A woman of colour with silver hair and wrinkles is digitally white washed and age-washed. I spent a couple hours adding darker tones to her skin, beautiful age lines and silver in her hair, in seconds, those were all gone. This causes harm.  

In this 2min video, I explain AI Art: https://youtu.be/l58ldy5BHc8?si=kc_DsMuG3EgkV52K

How: This Digital Still is a blend of analog and digital, of a 16mm film, a photo of acrylic gel skin on window, and an algorithmic variation using the video+image prompt (no text) in RunWayML. The percentage of AI in the output is very low.

Policy & Research: Some technologies help us tell our stories better, some limit everyone, some limit only some of us. I felt censored so I pushed the limits and explored how I could impact the generated outputs. It turns out that redressing biases and sexualization every step of the process requires more funds, more time and involves giving more of our own art for platforms to use for their benefit. People who are racialized, marginalized, sexualized or otherwise made invisible or represented negatively by Generative AI must input more, to protect their AI generated-identities. They must input more to protect their self-expression from being censored, and from existing bias in society becoming worst.

Some say AI simply reflects bias in society but a recent report from Bloomberg shows that Generative AI is more biased and risks making discrimination worst.  This Exhibit hopes to increase awareness of these issues and support policies (AI design, funding, and more) and regulations that address them proactively.

What do you think is an acceptable solution to this problem? Share your ideas with us in the "discussion" section.

4. Selkie 2:
The Sea Lioness

Inspiration:The Words written on her chest reflect the meanings attributed to "Imperfect" by our contributors: ache, war, bruise, bias, fear, stain, sabotage, anxiety, anger, etc.  The Sea Lioness video reminds me of motherhood and the fierce Lioness that leaps out when your young is under attack; of hurt provoking anger; of constructively channelled anger that leads to action and positive change. 

How: I used my painting as an image prompt. A video was generated, when prompted with "a seal woman dives into the ocean", producing a rather interesting algorithmic interpretation. Technically, the controls in the Generative platform, RunwayML, include text, image, video, angles, speed, etc. They are still limited.

Policy and Research: The concept of control is pivotal in intellectual property law, serving as a cornerstone for safeguarding artists' economic rights. In the video I created, I used an original image of mine to influence the video's final output. Currently, the extent of control I have over the result makes it relatively straightforward for a court to identify the creator. However, my concern is that this clarity may soon diminish.   

For instance, an AI-generated film named "AI Surobuin" was granted copyright in South Korea, primarily because it was classified as a Composition Work, where the creators exerted significant control over the final product. This control was manifested not only through the editing process, combining generated images, text, voices, and music, but also through the detailed customization of the AI's generative capabilities, similar to the techniques we use in 'Explor'art', our Retrieval Augmented Generation (RAG) chatbot (in Explore & Play section). 

Recently, OpenAI unveiled a preview of its text+image-to-video generator, capable of producing compelling videos up to 60 seconds long, surpassing RunWayML's current 18-second limit. OpenAI's strategy involves developing a base model that allows others to train specialized mini-generative AI systems, thereby enhancing user control. This approach is akin to OpenAI's existing GPT Store, which resembles an App Store.  

With the potential tsunami of videos this new generator will produce, how will the protection of copyright be affected?  

3. FISH

INSPIRATION: The Saint-Lawrence River, or the K’taqmkuk in Mi’Kmaq, and the realities of fishing communities. 

HOW: Digital Stamp (no AI) made using a cut out of a satellite image of Lac St-Pierre, courtesy of Pierre-André Bordeleau, RIVE, UQTR and the Sketches painting app. The short video below was generated using a text and image prompt in RunwayML. 

Policy and Research: Amid disruptive climate changes, overfishing, and unsustainable business models, communities that used to rely on fish for their survival are now facing scarcity.

1. Click here for recent policy work presented at the United Nations titled Gender Equality and the Environment in Digital Economies.
2. Click here for an interactive documentary (in French only) co-created with artists to define Smart Oceans.
3. Click here for a sociolegal study of the feasibility of a data trust for the Saint-Lawrence River (in French only).

2. FEET IN THE WATER

INSPIRATION: I grew up in a small town in Gaspésie that adopted my mother and I. When my mother died, I was living in Japan. She lay on the floor for 3 days. During those 3 days, I was by the ocean for a final camping trip with my then husband, knowing we were going to divorce soon. I wrote that story and started making a short film about it because losing her was the biggest loss I’d ever experienced. That is the art that went into the machine.

HOW: Feet in the Water is a Digital Still of a short video using 16mm film I made of my mother-in-law's feet walking on watercolour and an algorithmic variation generated using RunwayML.

POLICY & RESEARCH:  To create this Digital Still, I used both a video and an image (my analog art) to prompt the video variation. I read the User Agreement and had a conversation with ChatGPT to help me better explain the complex implications. RunwayML states that it the user (in this case me) still owns the original artwork, but by sharing it on the platform, I accept that the platform will use it to get better and make money from it. The short story is: artist inputs artwork and helps train the Generative AI platform; platform monetizes and shares with their partners. Keep reading, it gets spookier. ChatGPT explains RunwayML’s user agreement as follows:

"We're not going to claim we own your original artwork or the new pieces created from it. But when you share your art with us, you're giving us permission to let our technology learn from it. This could mean your art helps improve how our platform works, helps us understand different types of art better. Also, we might make money from your artwork by using it to enhance our platform, show it to others in different ways, or create new art based on it. Artwork or data shared with a Generative AI platform could indirectly enhance AI models that might be used in defense contexts, due to the broad applicability of AI advancements across various fields. While your specific artwork isn't likely to be used directly by the military, the improved capabilities of AI systems, developed through exposure to diverse datasets including artwork, could be applied in defense-related areas such as image recognition. It's important to be aware of how the technology developed by platforms you contribute to might be used, including potential military applications, by reviewing their terms of service and understanding their collaborations.”

Then I asked ChatGPT to clarify what happens when I input more of my art to create algorithmic variations. Here is the answer.

“When it comes to algorithmic variations or derivative works generated by the service provider based on your original artwork, the situation becomes more complex and depends on specific copyright laws, which can vary significantly by jurisdiction. Generally, copyright for derivative works can sometimes be held by the creator of the original work, but in the case of algorithmic or automated transformations made by a platform, the platform might claim some rights over those new creations, especially if the transformation is substantial and involves a significant amount of creativity or originality by the platform's algorithms.”

If you re-read that last sentence carefully, you’ll notice that ChatGPT implies that machines are creative, a bias that reflects the interests of commercial viability and close partnership with Microsoft. Some of the handful of large tech companies that are developing and deploying Generative AI platforms have presented arguments before the Federal Trade Commission (FTC) in the U.S. and asking governments around the world to change intellectual property laws for their benefit. It is also noteworthy that the FTC started investigating deceptive changes in user agreements. Artists are fighting back.

What are the policy implications of women, women of colour, Indigenous women, non-binary folks, differently abled people or otherwise marginalized and equity seeking people, feeding Goliath size Generative AI platforms? Should benefit distribution mechanisms be implemented and enforced?