Researchers from the University of Helsinki in Finland have found a way to generate images on computers by monitoring brain signals.
The study involved 31 volunteers. To begin, the researchers showed them hundreds of AI-generated images of different people in quick succession. In particular, they asked the participants to concentrate on specific features such as faces that looked old or were smiling.
All the while, the participants' brain activity was recorded via an EEG and fed into a neural network. This network was able to find correlations between electrical signals in the brain and what the subjects were looking at. As such, it was then able to estimate what kinds of faces the participants were thinking of.
To test the neural network's validity, participants were then asked to evaluate images generated by the computer on how well they matched the features they were thinking of. Overall, the participants rated the computer to have matched their thoughts 83% of the time.
The researchers say that the technique may one day be adapted to enhance human creativity. Would someone want to draw something but be unable to do so, for example, they say that computers may be able to help them. The researchers also say that the technique could be used to understand perception and other underlying mental processes.
"The technique does not recognize thoughts but rather responds to the associations we have with mental categories." says senior author of the paper, Michiel Spapé, "Thus, while we are not able to find out the identity of a specific 'old person' a participant was thinking of, we may gain an understanding of what they associate with old age. We, therefore, believe it may provide a new way of gaining insight into social, cognitive and emotional processes."
To take their research a step further, the researchers are now looking into whether the technique exposes unconscious associations eg. seeing whether the computer always generates images of old people as smiling men.
Sources: Neuroscience News, Nature
