Tech
An AI “mind-reading” tool can reconstruct what you’re looking at from a brain scan
Publication: MIT Technology Review
Author: Jessica Hamzelou
Source published: 01/10/2026, 10:32:24 UTC
Briefing created: 02/10/2026, 10:24:20 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 02/10/2026, 10:20:24 UTC.
Briefings
Researchers at Weizmann Institute developed an AI tool using high-resolution fMRI data to reconstruct visual content from brain scans and predict neural activity based on visual input. The system uses an encoder-decoder architecture trained on 9,000 images per subject, achieving reconstruction accuracy where the model's output matches the original image structure and content. The tool requires only one hour of fMRI data per new subject, compared to 40 hours previously needed, and can generalize across individuals by identifying shared brain regions. The work was presented at the Cognitive Computational Neuroscience conference in New York last month.
Key points
- The AI tool uses an encoder-decoder architecture to predict fMRI brain activity from visual inputs and reconstruct visual content from brain scans.
- Training data consists of 9,000 images per subject, with approximately 70% of the data being from images never shown to a person in an fMRI scanner.
Context and limitations
The tool requires one hour of fMRI data per new subject, whereas previous methods required 40 hours. The reconstruction accuracy depends on the specific structure and position of the original image, and the tool may not perfectly replicate all visual details.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 02/10/2026, 10:24:20 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeieabhbhawjf4oucwoeq4vkbnwge34py2m5kzcmyvxgxhebnghhrmm