The artist René Magritte completed a painting of a pipe and coupled it with the words “Ceci n’est pas une pipe.” Magritte called the painting La trahison des images, “The Treachery of Images.”
Magritte’s assumption was almost diametrically opposed: that images in and of themselves have, at best, a very unstable relationship to the things seem to represent, one that can be sculpted by whoever has the power to say what a particular image means. For Magritte, the meaning of images is relational, open to contestation. At first blush, Magritte’s painting might seem like a simple semiotic stunt, but the underlying dynamic Magritte underlines in the painting points to a much broader politics of representation and self-representation.
Reflect on the relationship between labels and images in a machine learning image classification dataset? Who has the power to label images and how do those labels and machine learning models trained on them impact society?
I think the labels we give to images and their meaning in the machine-learning classification dataset have a strong mutual connection between them. Labels we give to images reflect how we humans perceive this world, and image classification data sets indicate how the artificial intelligence perceives this world portrayed by us humans. When we are defining the images in the AI datasets, the AI is also defining us (our images). The power dynamics between labels and images are mutual. This has a big impact on society because of the ways of producing these datasets. Although many people think these datasets are unbiased, they might actually have different opinions and judgments towards people or organizations who have produced these datasets. For example, older datasets like ImageNet use words that may oversimplify or incorrectly describe people. The names we give images come from society’s ideas, which determine and reflect how artificial intelligence will see the world.
In real situations throughout history, powerful people have decided how things should be named and labeled. They had the rights and privileges to influence the structure of society and determine the stories we hear. However, according to our reading’s reference to Magritte and the physiognomists’ approach, anyone who can recognize and name the images should have the power to label images. Our labels towards one image can be relative, for we are living in a diverse world with different people holding different opinions and perceptions. The meaning and naming of images can be discussed. Just like when we face art, everyone has a different understanding of it. Our understanding of it not only represents the work but is also a manifestation of self-representation. This kind of diverse presentation and discussion should be encouraged, so as to prevent AI from ignoring individual voices, reduce bias, and make the database more diverse and inclusive.
My sketch: https://editor.p5js.org/jz5821/sketches/oNgDMsiGd
Screen Recording 2023-10-03 at 5.28.38 PM (1).mov
I trained an audio model on Teachable Machine including the background noise, clapping sounds, and sounds of rubbing hands. Users would trigger a line of text “If you’re happy clap your hands” with its corresponding emoji when they clap their hands, and so as rubbing their hands.