Meta TRIBE v2 predicts human brain activity
Meta AI's TRIBE v2 is a multimodal foundation model that predicts human fMRI brain activity with 70x resolution across text, video, and audio. An experimental interface by Adam Jesionkiewicz demonstrates the model's ability to predict content engagement and "virality" from raw text by simulating neural responses.
This is a major leap in AI-brain alignment, moving beyond basic sentiment analysis to full neural simulation. Trained on 1,000+ hours of fMRI data from 700+ subjects, it achieves high-resolution zero-shot predictions for new stimuli without additional training. The model demonstrates immediate utility for neuromarketing by simulating emotional and curiosity-driven neural responses in silico. Open-sourced under CC BY-NC, the TRIBE v2 foundation model turns expensive clinical research into seconds of local inference.
DISCOVERED
12d ago
2026-03-30
PUBLISHED
13d ago
2026-03-29
RELEVANCE
AUTHOR
Adam_Jesion