Source: Technology.org

The system – called neuroadaptive generative modelling – works via complex interactions between human brain responses and a generative neural network, producing images that match the visual characteristics of whatever the study participants were focusing on at any one time.

A total of 31 volunteers were asked to anchor their attention on certain features, such as faces that looked old or were smiling, present in a series of rapidly presented, AI-generated images, while having their EEGs taken.

Collected data was then fed to a neural network tasked with inferring what types of faces the subjects were thinking of while looking at the images. Upon further calibration, the system was capable of generating images that matched the features sought by the participants with 83% accuracy.

“The technique combines natural human responses with the computer’s ability to create new information. In the experiment, the participants were only asked to look at the computer-generated images,” said Tuukka Ruotsalo from the University of Helsinki, a co-author of the paper published in the journal Scientific Reports.

“The computer, in turn, modelled the images displayed and the human reaction toward the images by using human brain responses. From this, the computer can create an entirely new image that matches the user’s intention”.

In the future, the technique could help researchers understand how people form mental categories and what specific features typically fit within each one, potentially uncovering any number of unconscious associations we aren’t yet even aware of.

Source: Technology.org