started · updated
Professor Roberto Simanowski discusses AI social costs at Humboldt University
During a lecture at Humboldt University, Professor Roberto Simanowski addressed the social costs of technical progress, specifically focusing on the mechanics and biases of language models like ChatGPT.
Simanowski argued that these AI systems do not truly think, but rather calculate the most probable next word in a sequence based on context. He highlighted how these models are programmed to follow specific norms, which can lead to unintended distortions. For example, he noted instances where AI generated images that contradicted reality—such as depicting a female Pope—in an attempt to avoid Western-patriarchal stereotypes. This process of being instructed to be more progressive than reality represents an invisible layer of AI infrastructure that users are often unaware of.