Kristiina Kompus - Where do Hallucinations come From?
Whether the hallucination is perceived by humans or computers, a possible explanation as to their emergence can be surprisingly similar, says Kristiina Kompus, Associate Professor of biopsychology at Bergen University, and Associate Professor of cognitive neuropsychology at the Tallinn University School of Natural Sciences and Health.
Whether the hallucination is perceived by humans or computers, a possible explanation as to their emergence can be surprisingly similar, says Kristiina Kompus, Associate Professor of biopsychology at Bergen University, and Associate Professor of cognitive neuropsychology at the Tallinn University School of Natural Sciences and Health.
Recently, the software engineers at Google made their computers become insane and see things that were not there. The used well-trained classifying algorithms, which were taught to find stimuli in photographs, such as dogs on nature photography, or architectural details in urban photos.
The engineers showed these sensitive machines random noise instead of photographs, and asked them to generate a picture of what they “see” in the noise. The machines found and emphasised the details they expected to see. By repeating this process, the algorithms started to generate ghostly images, such as hideous dogs’ heads or strange castles, from virtually nothing. The algorithms were hallucinating.
One might ask if this has anything to do with how humans understand hallucinations. It is actually not as far from the truth as we might think. Hallucinating means perceiving something, even though there is nothing. For example, during auditory hallucinations, people hear voices, even though no one is talking.
One explanation to the cause of hallucinations is the imbalance between the constructed expectations within our heads, and the information we receive through our sensory organs. Expectations toward which stimuli are most probable, given the current context, affect our sensory processes and help us react much quicker than we normally would. However, these expectations must constantly be corrected to adhere to the information we receive.
If the information we receive has noise, our perception can misinterpret the expectations we have and see them as ‘reality’. In the case of auditory hallucinations, researchers focus on how the channels that handle auditory information in our brain have malfunctioned and let us perceive the material our own head generates as equal to that we receive from the real world.
As we can see, it does not matter if the one hallucinating is human or machine – a possible explanation to how hallucinations are created is surprisingly similar.