Close×
January 30, 2023

Magic Avatar, Vanity, and Digital Resignation…

This article is a translation of a text first published in Eductive’s French edition.

At the end of 2022, we have seen the viral emergence of thousands of selfies on social media feeds. Generated by artificial intelligence (AI) from databases of artistic works, these magic avatars increase the already critical big data that are cumulating in storage spaces. This trend reopens the discussion on personal data ownership and the potential capacity of AI to overtake some professions and expertise.

Following the success of Lensa AI, AI art apps entered the App Store’s Top Charts of the most downloaded apps. They allow users to turn their selfies into sci-fi or anime characters, among others. This new trend reminds me of the advent of augmented reality filters introduced by Snapchat in 2016.

The frenetic use of filters was at its peak when I received on my smartphone, a picture of my grandmother who had replaced her face with a cat’s.

Technological breakthroughs in art from 2015 to today

In 2015, DeepDream, a program created by the Google engineer Alexandre Mordvinstev, was initially intended to detect objects in images in order to automatically classify them. Based on a model of a neural network called “Inception” (named after Christopher Nolan’s sci-fi movie), the tool powered by pixelated images of insects and mammals would create surreal representations of animals.

Many images of a red canine superpose each other, forming cartwheels. The result looks like fractals.

Artwork generated by the DeepDream program (Image source: Martin Thoma, “Deep dream white noise”, Wikimedia Commons)

In 2016, a teacher at Cornell University was working on an algorithm that could detect artistic styles. His program named Style Transfer could transfer an artwork such as the Mona Lisa painting from a Renaissance style to a Cubism or Post-Impressionism style.

A version of The Mona Lisa that appears painted with broad, expressive brushstrokes, in the midnight blue, black and yellow hues of VanGogh's The Starry Night.

Mona Lisa by Leonard da Vinci in the style of Starry Nights by Van Gogh (Source: Pjfinlay, “Mona Lisa the starry night”, Wikimedia Commons)

In 2018, the STYLEGAN project by NVIDIA engineers based on Generative Adversarial Networks (GAN) could generate images of people who do not exist. This system involves 2 networks opposing each other. One is the discriminator, having in its database millions of face images, while the other one is the generator that will create an image according to specific criteria.

Image of a human face created by the random face generator on the website This Person Does Not Exist

In 2011 appears DALL.E from Open AI that can create different types of images from a text description and the GPT-3 language model. In fact, Big Sleep by the artist Ryan Murdock uses GAN and DALL.E.

In 2022 appears Guided diffusion that can generate images massively used to create notebooks and sketches for artists. It is followed by a myriad of communities of coding artists who will adapt these tools to create some free and open ones.

With the rapid breakthrough of algorithmic technologies in art over only 7 years, will AI also take over our creativity?

Who better to answer this question than an artificial intelligence model such as ChatGPT? Here is an excerpt from the dialogue.

Me: “ Will AI take over human creativity?”

ChatGPT: “It is difficult to predict the future with certainty, but artificial intelligence (AI) will probably play an important role in human creativity in the future. Indeed, AI can be used to help people create new ideas and find innovative solutions to complex problems. However, it should be noted that IA will never completely replace human creativity. Indeed, creativity is a complex process that involves not only the capacity to find new ideas, but also to evaluate and effectively implement them. It requires a large amount of reflection, intuition, and hard work, and these qualities are unique to the human mind. In other words, although IA can be a valuable tool for human creativity, it will never completely replace the human capacity to create and innovate.”

Personally, I find the answer comforting. It confirms that we have control over our creativity, but that the art enhanced by algorithms will allow us to push the limits of our inventiveness.

Let us not forget that without data, AI is nothing. The new black gold of algorithmic technologies remains big data that we generate by using apps such as Lensa AI.

Digital resignation

I am always the first to raise the flag when it comes to data protection. However, I could not resist the temptation to use an app that generates images from 20 of my selfies uploaded from my images folder. This for the only purpose of seeing myself as a hero in Avatar or a warrior in Crouching Tiger, Hidden Dragon.

Despite being aware of some abuses in the use of our personal data, why do we keep using this type of app?

Users waive their fundamental rights and freedoms when using magic avatar apps, but these rights remain widely unknown. Yet, nearly 13 million people in the United States downloaded the app.

It is important to know that in today’s world based on image and physical appearance, several apps and platforms are designed to boost our ego. LinkedIn, for example, is a great tool for networking, but also to view the number of shares and thumbs-ups following a post of our achievements. Notifications are developed in such a way that they generate a sense of fulfilment and recognition when we realize how many people have viewed our page. The effect is equally gratifying when, 20 minutes after submitting 20 selfies to the magic avatar tool, you discover more than 50 selfies enhancing your sense of self with a perfectly regular and balanced facial structure.

Resigning ourselves for self-rewards

Digital resignation is the condition produced when people desire to control the information that digital entities have about them but feel unable to do so (Draper, N. A., & Turow, J, 2019). According to Nick Yee (2014), a social scientist who has studied self-representation and social interaction in virtual worlds and online games for more than 20 years, if Lensa’s appeal is substantial, there is probably a vanity factor in play. It is no secret that people enjoy seeing pictures of themselves looking good. This appeal factor highlights the real issue of the lack of transparency in personal data management by companies developing this type of app.

Being cautious and resisting our ego

AI and cybersecurity specialists state that personal data are generally provided when users create an account where they can add information such as their name, email address, or even their home address. Nothing new here. The trouble is Lensa AI communication concerning the 20 selfies we have to upload to access the final works of art. What happens to those photos and data generated using this app?

Screencapture of a tweet from Prisma Labs (@PrismaAI), made on december 6, 2022, at 9:37. The tweet is a reply from @PrismaAI to one of it’s own tweets. The tweet says: «As soon as the avatars are generated, the user’s photos and the associated model are erased permanently from our servers. And the process would start over agin for the next request.» When the capture was made, the tweet had 15 likes and 4 replies (not visible).

Screenshot of a tweet by Prisma Labs mentioning that as soon as the avatars are generated, the user’s photos are erased permanently from their servers. (Source)

According to Prisma Labs, as soon as the avatars are generated, the user’s photos and data are erased permanently. But how can this information be verified?

Photorealistic illustration of Florence Sedaminou Muratet’s face and shoulders

Portrait of Florence Sedaminou Muratet generated by Lensa AI

I could discover the 60 portraits of myself generated by AI from the uploaded photos on the app. The use of AI-augmented artistic tools will never cease to amaze us. That being said, the use of personal data remains an issue that should be taken seriously. These new technologies highlight a digital vulnerability that we must address within the different college programs.

A large number of these apps are developed with the sole purpose of obtaining our data that will afterwards be used at our expense. The issue here is not to focus on the replacement of artists by AI, but to develop strategies to protect ourselves, numerically speaking. It is also about the role of teachers who will have to discuss and clarify these topics in class. One of my colleagues who is a Philosophy and Ethics teacher has introduced the question of self-perception and hypersexualization of the images of women generated by Lensa AI. The issues related to the protection of personal data and digital resignation could then be addressed in class. In the field of education, there is always a way to ask questions and guide reflection. For my part, I have never been as close to my students as when discussing these digital tools. We are all in the same boat, and we navigate together.

About the author

Florence Sedaminou Muratet

Inspired by the conviction that education is the most powerful tool for changing the world, Florence Sedaminou Muratet has forged a solid academic career in anthropology and education. Cumulating over 20 years’ experience, she has not only taught French as a second language and Humanities, but also integrated interculturality and digital pedagogy into her courses at an early stage. Her major contributions include the redesign of educational programs, which led her to skillfully integrate multidisciplinary approaches and digital technologies, thus significantly enhancing the learning experience.

As a digital pedagogy consultant with Collecto, Florence stands out for her pioneering work in implementing emerging technologies in education. She is particularly committed to exploring the applications and implications of artificial intelligence in society, actively helping to shape the future of digital education.

Subscribe
Notify of
guest

0 Commentaires
Inline Feedbacks
View all comments