Computer code and DNA sequences used to train AI can trigger human-like activity in a brain scanner.
April 20, 2026
Original Paper
Brain Score Tracks Shared Properties of Languages: Evidence from Many Natural Languages and Structured Sequences
arXiv · 2604.15503
AI-generated illustration
The Takeaway
High Brain Scores in AI models appear when they process any structured sequence, including Python or nested parentheses. These models predict human brain activity just as well as those trained on actual English or Spanish. The scientific community previously thought this alignment meant AI was finally mimicking human linguistic processing. This evidence suggests that the human brain and AI are simply responding to the same underlying mathematical structures found in all logic. AI development might be converging on a universal grammar that transcends human biology entirely.
From the abstract
Recent breakthroughs in language models (LMs) using neural networks have raised the question: how similar are these models' processing to human language processing? Results using a framework called Brain Score (BS) -- predicting fMRI activations during reading from LM activations -- have been used to argue for a high degree of similarity. To understand this similarity, we conduct experiments by training LMs on various types of input data and evaluate them on BS. We find that models trained on va