Researchers at Kyoto University in Japan say they have used deep neural networks to decode thoughts; specifically, more "hierarchical" images that have multiple layers of color and structure.
They note the technique enables artificial intelligence (AI) to detect objects, and not just binary pixels as engineered in previous methods.
"These neural networks or AI model can be used as a proxy for the hierarchical structure of the human brain," says Kyoto University's Yukiyasu Kamitani.
The project involved 10 months of showing three subjects natural images, artificial geometric shapes, and alphabetical letters for varying periods while their brain activity was scanned. Afterward, a computer decoded the data to produce visualizations of the subjects' thoughts, and Kamitani notes the team discovered the method could be used to reconstruct visual imagery a person produces by thinking of some recalled images.
He says the AI encountered more difficulty rebuilding images when decoding brain signals from a subject remembering those images.
View Full Article
Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA
No entries found