Skip to main content
Innova-tsn Culture
Mar. 2025

Invasion of the Language Models

[An article on the film The Arrival, Denis Villeneuve, USA, 2016] 

How difficult communication in these modern times is, where the meaning of words varies so much depending on context. In Goodbye to Language (2014), Jean-Luc Godard constructed a 3D filmic poem with a testamentary vocation and a pessimistic sentiment in which he already spoke about the need to use an interpreter for every word one utters. 

 

In this film, where the formal interplay between signifier and signified stands out through verbal anagrams with images, there are also sentences like “words. I don’t want to hear about them.” Among other references, the French film titan responsible for Histoire(s) du Cinema (1988-1998) alluded to Mary Shelley and the writing process of Frankenstein or the Modern Prometheus: the foundational novel about Artificial Intelligence. Interestingly, two years after Godard’s film and one year before the revolution linked to the use of Transformers in the context of LLMs (Large Language Models), Denis Villeneuve’s The Arrival (2016), based on Ted Chiang’s “Story of Your Life,” was released in cinemas. 

 

The American writer had already made a proposal of ontological terror about the nature of mathematics in the story Division by Zero, where he created a diatribe about the consistency of mathematics to end up comparing it to the existence of God.  

 

In the Canadian film starring the linguist played by Amy Adams, everything starts with the appearance of giant semi-oval spaceships that might be reminiscent of the monolith in 2001: A Space Odyssey (1968). That is to say, a new dissertation on the problem of communication with a fully developed extraterrestrial intelligence: the heptapods.  

 

In the face of confusion, while some start to panic and the military think only of defending themselves and isolating themselves in order to avoid a war, the scientists try to establish contact. Every word counts and a misunderstanding can be terrible in a race against time to understand aliens who seem to have no curiosity about the human species. Jeremy Renner’s character tries to find out about their scientific knowledge based on mathematical principles, while the female protagonist, like a mother, realises that communication must start with simple words in order to build up conceptual complexity from there. 

 

On the other hand, while in Close Encounters of the Third Kind (1977) common language was made up of rhythmic musical games, here the dialogue through speech and writing forks. The most accurate channel of communication is through abstract circular-shaped semasiographic logograms, intended to be understood as embeddings, specifically designed for humans to decipher and progressively develop a new form of non-linear thinking following the Sapir-Whorf theory. 

 

“Memory is a strange thing. It doesn’t work like I thought it did” says the protagonist. In Chiang’s story the constructions of sentences were structured under the juxtapositions of past and future tenses. For his part, Denis Villeneuve establishes a formal interplay in which the allusions to the passage of time in Andrei Tarkovsky’s films, the circular camera movements, the fragmentation and the alternative narrative organisation converge towards a borderline territory between the beginning and the end of the story that reveals the true meaning of the mise-en-scène. A real structural milestone emphasised by the fact that the protagonist’s daughter is called Hannah, like a palindrome.  

This very concern about the predictive power linked to language learning remains present in his version of Dune (2021-2024) where acquiring the knowledge the Fremen and the Bene Gesserit Reverend Mothers offer is part of the training that leads young Paul Atreides to become the Lisan Al Gaib, this is, the Chosen One.Perhaps the strange power that the protagonist receives when perceiving time could be understood as the way in which LLMs work in their task of understanding double entendre sentences to generate text. After all, in the age of ChatGPT and prompt engineering derivatives, it seems that behind all the epistemological architecture they hide, there is an intrinsic poetics that will not only define the future, but reveal its true expressive essence, more akin to art.  

 

* By Nicolás Quintero, Data Scientist at Innova-tsn