The heptapods don't teach Louise Banks a new language—they teach her to read the way a transformer model reads.
Watch the moment she first decodes a logogram: not as a linguist parsing grammar, but as someone learning to intake simultaneous information streams and extract meaning from their geometric relationships rather than their sequence.
This is the actual mechanism in the film—the aliens' writing system forces her brain to process input the way a neural network does, collapsing temporal linearity into spatial simultaneity. By 2016, this infrastructure was operational: AlexNet had won ImageNet in 2012, and transformer architecture's conceptual problem—how to let systems learn relationships across entire datasets at once instead of parsing them word by word—was already the dominant research direction in machine learning labs.
The film's central mistake—that understanding their language grants Louise precognition—accidentally reveals the truth: she doesn't see the future. She sees causality as a geometric structure instead of a temporal arrow. That's what parallel processing is.
The aliens themselves are almost irrelevant, just the narrative machinery required to make Americans accept that learning to read non-linear systems would rewire human consciousness.
Find the sequence where Louise studies the heptapod logogram on her whiteboard (around 45 minutes in) and ask yourself what you're actually seeing: Is she decoding a message, or learning to parse a data structure the way a CNN parses an image?
Eric Heisserer's 2017 interview with The Verge about the screenplay's development mentions he spent months researching cognitive linguistics but almost nothing on cutting-edge AI—yet the film's visual language mirrors exactly how convolutional neural networks were being visualized in academic papers from 2014-2016.