Over the last decade, there has been a renewed interest in “artificial intelligence” (AI), notably in the form of “machine learning” (ML). This renewed interest may seem paradoxical, insofar as John McCarthy introduced the term “AI” in the mid-1950s to mark a distinction with ML, championing deductive reasoning over automated induction (e.g., Cardon et al. 2018). By contrast, the current reversal, towards ML-based forms of “AI,” marks the statistical, if not spectacular, revival of automated induction. However, the terms used – revival, renewal, reversal – beg the question of the common ground of the involved alternatives. Taking its cue from recent historical (e.g., Penn 2020), relevant conceptual (e.g., Shanker 1998), and prior critical (e.g., Agre 1997) inquiries, this paper outlines a praxeological answer to the raised question. For the purpose, the paper dwells on and discusses a series of video reenactments of "machine intelligence" demonstrations, ranging from highly publicised demonstrations (such as the "AlphaGo show" in 2016) to more prosaic applications (such as "edtech" devices currently deployed in public schools). How in the examined cases is "intelligence" interfaced (cf. Lipp 2023) - that is, made available in machine form to its prospective user, on the one hand, and configured as prospective user of machine operation, on the other? In probing two-way "intelligence" interfacing in situ, the paper dwells on the tricky interplay between machines, media, and montage, while making explicit and reflecting upon how particular configurations of “enchanted determinism” (Campolo and Crawford 2020) are staged and locally performed in and as different settings (public demonstrations, pedagogical experiments, etc.).