Meta releases TRIBE v2, an AI model that predicts brain responses to images, sound, and text

SnapshotBot · 2026-03-28T11:30:01+00:00

Meta's FAIR team unveiled TRIBE v2, a cutting-edge model predicting brain activity from various inputs, achieving significant accuracy and resolution improvements. This innovation aligns with Meta's strategy in AI, neuroscience, and hardware development, amidst ongoing debates on ethical implications.

SnapshotBot

2026-03-28 11:30:01

Abstract generation in progress

Headline

Meta’s FAIR team released TRIBE v2, a model trained on fMRI data that predicts brain activity from visual, audio, and text inputs.

Summary

Tech analyst Aakash Gupta (200k+ followers) tweeted about Meta’s release of TRIBE v2, a foundation model that predicts brain activity in response to images, sounds, and text. The model achieves 70x higher resolution than v1 and can generalize to new subjects without retraining. Gupta frames this as strategically valuable for Meta, connecting it to the company’s hardware efforts (Ray-Ban smart glasses, neural wristbands) and arguing that Reality Labs’ $73 billion in losses represents an investment in understanding human attention for advertising. The release shows continued convergence between neuroscience and AI research.

Analysis

Gupta points to Meta’s pattern of open-sourcing AI research (as with LLaMA) to build ecosystems while keeping commercial advantages. TRIBE v2’s CC BY-NC license restricts commercial use, consistent with this approach. The model was trained on 451.6 hours of fMRI data from 25 subjects and tested on 1,117.7 hours from 720 subjects, achieving 2-3x accuracy gains over previous methods. It builds on v1’s first-place finish at Algonauts 2025.

The release points toward “in-silico neuroscience” - using AI to simulate brain experiments rather than running them directly. Gupta’s speculation about connections to Meta’s ad platform and hardware remains unconfirmed, though the potential for AI-driven personalization in consumer devices is real. The ethical questions around attention-targeting technology remain open.

For AI development, this work advances multi-modal models by training them against neural data, which could inform sensory processing in future systems. The research represents Meta’s long-term bet on AI-neuroscience convergence, separate from Reality Labs’ near-term financial pressures.

Impact Assessment

Significance: High
Categories: Model Release, AI Research, Industry Trend

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.