Meta Created TRIBE v2 — A Model for Predicting Brain Activity While Watching Videos
Meta developed TRIBE v2 — a model that predicts how the brain processes video and audio information. The system was trained on observations of neural activity a
Meta presented TRIBE v2 — a predictive model that simulates how the human brain processes video and audio. The system is trained on direct observations of neural activity and can predict the brain's response to any video content.
How TRIBE v2 Works
TRIBE v2 is trained on fMRI scan data and other methods of registering brain activity. When participants are shown video or audio is played, the model learns to predict which neurons and brain regions activate in response. The key difference from the first version is that TRIBE v2 is multimodal: it processes video, audio, and their interaction simultaneously, just as the human brain does. The model predicts activity from low levels (primary visual cortex) to high abstract levels of semantic understanding. In other words, TRIBE v2 captures not just whether the brain sees the contours of objects, but whether it understands what is happening on screen.
Practical Applications
Researchers are already applying TRIBE v2 in several directions:
- Neurobiology — studying coordination between different brain regions when perceiving complex information
- Artificial intelligence design — borrowing ideas from the brain to improve computer vision and hearing
- Medical diagnosis — detecting hearing and vision disorders, studying the consequences of stroke
- Brain-computer interfaces — developing BCI systems for people with paralysis or blindness
Why This Matters
Before TRIBE v2, neurobiology and artificial intelligence developed almost in parallel, rarely intersecting. This model creates a bridge between two fields: it shows what principles of information processing the real brain considers effective. Researchers can now look at how TRIBE v2 works and borrow ideas for their neural networks.
What This Means
TRIBE v2 opens a new direction — neuro-inspired AI that learns not merely to copy the brain's structure, but from its actual functioning. This could lead both to improvements in computer vision and to understanding fundamental mechanisms of consciousness.
*Meta is recognized as an extremist organization and is banned in the Russian Federation.