Researchers at the University of Helsinki have developed a technique in which a computer model visualizes perception by monitoring the signals of the human brain. In a way, it is as if the computer tries to imagine what a human is thinking. As a result of this imagery, the computer is capable of producing completely new information, such as imaginary images that have never been seen before.
The technique is based on a novel brain-computer interface. Previously, similar brain-computer interfaces have been capable of one-way communication from brain to computer, such as moving individual letters or a cursor.
As far as is known, the new study is the first where both information and computer presentation of brain signals were modeled simultaneously using artificial intelligence methods. Images that matched visual characteristics that participants were focusing on were generated through interactions between human brain responses and a generational neural network.
The study was published in September in the journal Scientific Reports.
Neuroadaptive generative modeling
Researchers have called this method neuroadaptive generative modeling. A total of 31 volunteers participated in a study that evaluated the effectiveness of the technique. The participants were shown hundreds of AI-generated images of diverse-looking people while their EEG was recorded.
Subjects were asked to focus on certain features, such as faces that looked older or were smiling. Given a rapidly rendered series of facial images, subjects’ EEGs were fed to a neural network that anticipated whether an image was matched as the subjects that the subjects were looking for. .
Based on this information, the neural network adapted its estimate to what kind of faces people were thinking about. Finally, the computer-generated images were evaluated by the participants and they almost completely matched the characteristics that the participants were thinking about. The accuracy of the experiment was 83 percent.
“Technology combines natural human responses with the ability of computers to create new information. In the experiment, participants were asked to view only the images created by the computer.
The computer, in turn, models human response to images using displayed images and human brain responses. From this, the computer can create an entirely new image that matches the intent of the user, ”says Tukke Rutstalo, a Finland research fellow at the University of Helsinki, Finland and an associate professor at the University of Copenhagen, Denmark.
Subliminal mood may be exposed
Creating images of the human face is only one example of the possible uses of the technique. One practical benefit of the study may be that computers can enhance human creativity.
“If you want to attract or do something, but are unable to do it, a computer can help you achieve your goal. It can just be the focus of attention and predict what you want to create, ”says Rustomalo. However, researchers believe that the technique can be used to gain perception and understanding of the underlying processes in our brains.
“Technology does not recognize ideas, but we react to associations associated with mental categories. Thus, when we are unable to ascertain the identity of a specific ‘old man’, thinking of a participant, we can understand what they are concerned with in old age. Therefore, we believe that it can provide a new way of gaining insight into social, cognitive and emotional processes, ”says senior researcher Michel Spape.
According to Spapé, this is also interesting from a psychological point of view.
“One person’s view of an elderly person may be very different from another’s.” We are currently uncovering whether our technology can uncover subliminal associations, for example if computers always tell older people to smile, smiling.
Neurodeceptive modeling to create images matching perceptual categories
Brain-computer interfaces enable active communication and execution of a pre-determined set of commands, such as writing letters or moving cursors. However, they have so far not been able to estimate more complex intentions or to adapt to more complex outputs based on brain signals.
Here, we present neuroadaptive generative modeling, which uses a participant’s brain signals as a response to optimize an infinitely generative model and generate new information that matches the participant’s intentions.
