ARTIFICIAL INTELLIGENCE AND/FOR THE SOCI

Escaping artificial intelligence has become a complicated task. Escaping the discourse about AI has become even more so. Whether we see in it a promise of progress, the vector of a dangerous regression, a fad that will eventually pass, a simple tool, discussions about AI are omnipresent. And when we ignore it, AI often affects our lives in ways we do not notice. Because AI has become integral to the discourses and the practices of contemporary societies, social scientists are being pressed to position themselves with respect to it. 

The idea behind this course is that to understand what is at stake with AI, we first need to understand what AI is. It is useful to explore its origins, its history, the movements that have gone through it. It is also necessary to understand in concrete terms what a contemporary AI algorithm does. This means that we need to grasp, even intuitively, the difference between an expert system and a connectionist approach; why the Transformers architecture has allowed progress in the study of content; why neural networks are said to be “better at prediction than explanation”. Understanding what certain now-ubiquitous terms mean (such as RLHF, train/test/dev sets, AGI, zero and few-shot learning, ...) is also important for those who want to study contemporary societies.