To create a strong AI (one which replicates human intelligence) would necessitate an understanding of the relationship between language and thought; but human languages encompass more than written and spoken communication, they are also manifest in images, sounds, artefacts, places and practices. The complexity of language is therefore immense, perhaps intractable, leaving us with the question: how can computational natural language understanding (NLU) and computer vision hope to untangle the complex situated and cultural encoding of human language? Understanding the historical trajectory of AI, while also evolving a methodology and related methods to understand the ontological status of artificial language understanding, is the key theme of this essay. By cutting Hitchcock’s films Psycho and Rear Window into small ‘language games,’ the case is presented that tensions between Turing and Wittgenstein’s understanding of language are still manifest in contemporary algorithmic processes and rhetoric. Using image auto-tagging, verbal summaries, scene prediction and style transfer algorithms, practical and theoretical experiments address the limits and potential of symbolic representation and machine learning to understand the language and thoughts of human and other animals. The essay evidences what happens in specific cases when machine-driven effort is engaged to translate embodied and cultural meanings embedded in film, a ground truthing method articulated here as a form of language game.