Imagine a model which predicts our brain activity.

One that will open doors to more accurate neuroscience research and an understanding of how cognition works.

This is what TRIBE-V2, by Meta is designed to do.

It included 500-1000 hours of fMRI recordings from 700+ human participants who are exposed to things like movies, long form video, audiobooks and spoken language. Real cognition at scale.

Collecting fMRI data is slow, expensive and unforgiving as it needs cleaning, aligning and quality control checks, but Meta did it anyway.

They created a disciplined model that can predict brain activity, respond to new stimuli , works across new languages and tasks and requires NO retraining.

Why does this matter.

It will enable neuroscience experiments at scale and in minutes not months, predicting how our brain will respond to stimuli.

It becomes a benchmark for the human brain, a foundation of how future models are trained

It brings us a step closer to brain-computer interfaces

It may answer questions about how perception changes with age, how neurodiversity manifests, how diseases alter our information flow.

It will support research into neurodegenerative diseases, sensory processing disorders and the effects of stroke or traumatic brain injuries.

And in case you are wondering, TRIBE-V2 does not read your thoughts, or infer personal intent or beliefs. It responds to external stimuli, not internal.

SOURCE
See more: https://ai.meta.com/blog/tribe-v2-brain-predictive-foundation-model/

BESCI AI OPINION

Meta have released TRIBE-V2 for academic and non-commercial research, which enables reasearchers to download and run for free, but prohibits commercial deployment without permission or profit making products as-is.

The training daa has not been made public.