A team of Russian scientists trained an AI programme to guess key features of videos watched by a set of volunteers based on patterns of their brainwaves.
The research conducted by technology firm Neurobiotics involved showing volunteers a set of videos with key visual stimuli such as a waterfall or a first persona video of extreme sports. During this viewing, the participant’s brainwaves were observed using a none-invasive electroencephalogram (EEG) with each stimulus showing a discrete pattern of brainwaves.
The data from these patterns was fed to an AI to analyse, with the programme being trained to spot specific signatures and to identify and characterise these signatures as corresponding to specific stimuli. Finally, the results of what the AI has identified was decoded to see the AI’s interpretation of the visual stimulus, with the results and the corresponding original images being displayed in the video below:
While the recreated images are far from perfect, the ability of the AI to distinguish overall colours and more distinct shapes and patterns represents a significant leap forward if thought interpreting AI technology.
Scientific Britannicus 