💡AI Tool Predicts Brain Activity from Images, 70% Unpaired Data
AI Can Guess What You're Looking At from Brain Scans
TL;DR
A new AI tool can predict brain activity from images with remarkable accuracy, using 70% unpaired data. It could enable communication for the 'locked-in' and aid in understanding brain function.
A new AI tool can predict brain activity from images with remarkable accuracy, using 70% unpaired data. This breakthrough could enable communication for the 'locked-in' and aid in understanding brain function. The tool's encoder and decoder models, trained on diverse images, can work on scans from new individuals with minimal calibration. The decoder needs just one hour of data compared to 40 hours for other tools. The approach raises questions about mental privacy and the potential for involuntary analysis of brain activity.
Key Points
The AI model has two branches: one predicts image structure, the other content, improving accuracy.
The encoder can predict brain activity from images, trained on as many images as desired.
The decoder needs only one hour of data, compared to 40 hours for other tools.
The tool outperformed previously described ones in a comparison test, showing remarkable accuracy.
The team plans to move beyond images to video and audio, aiming to reconstruct thoughts and dreams.
Why It Matters
If you're working on brain-computer interfaces or studying brain function, this tool could revolutionize your research. It allows for more accurate predictions of brain activity from images, even with minimal calibration for new individuals. The ability to analyze brain activity from EEG would be a 'game changer', making it easier to study brain function and potentially aid those who are 'locked in'.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 3,568 builders reading daily.