Media and definitions surrounding artificial intelligence curate a specific notion that omits the large amount of human labor involved in algorithms. Robert’s writing “Your AI is a Human” sheds light on the reality of artificial intelligence. We discuss AI as eventually acting as a human or having human qualities. The fact is, these algorithms –social media platforms, chatGPT, google– all rely on, one, human input to generate predictability and, two, human labor combing through millions of pieces of content to censor inappropriate content or improve outputs.
Where does this leave us? Roberts discusses the importance of recognizing the human input and holding it accountable. Dangers arise when companies and/or the masses overemphasize the benefits of machine learning, de-emphasize their shortcomings, and don’t consider the lack of value judgments within computers (Roberts 65).
Similarly, we see these dangers firsthand (though harmless on the surface) in Salvaggio’s piece on generating AI kisses. The images generated obviously don’t look natural, but they also perpetuate biases due to human input. The training for these images lack diversity in race and sexual orientation, and they all look plain creepy. The images are spitting out replicas of patterns found among images of people kissing. These patterns come from previous images, not some new creation. The censorship, created through human labor, cannot always differentiate between what is inappropriate and what is safe (Salvaggio). I think it is fascinating that within his experiment he had AI produce photos of two men kissing and two women kissing and the women were flagged for sexual content. Women are inherently sexualized in culture and society, therefore cannot see representation outside this lens, meanwhile men do not hold his barrier so two men kissing does not code as “sexual content.” Behind all these barriers is the mass human labor force sorting through this information as Roberts describes. I wonder how this concept will be regulated as artificial intelligence improves?
Leave a Reply