Mapping AI issues in media through NLP methods

Using a variety of NLP methods on a corpus of press articles, we show that there are two dominant regimes of criticism of artificial intelligence that coexist within the media sphere. Combining text classification algorithms to detect critical articles and a topological analysis of the terms extracted from the corpus, we reveal two semantic spaces, involving different technological and human entities, but also distinct temporality and issues. On the one hand, the algorithms that shape our daily computing environments are associated with a critical discourse on bias, discrimination, surveillance, censorship and amplification phenomena in the spread of inappropriate content. On the other hand, robots and AI, which refer to autonomous and embodied technical entities, are associated with a prophetic discourse alerting us to our ability to control these agents that simulate or exceed our physical and cognitive capacities and threaten our physical security or our economic model.