Scientists have created an artificial intelligence system that constructs images of what a particular person is viewing, based on activity inside that person's brain. The technology is called Brain-IT and was developed by researchers at the Weizmann Institute of Science.
Functional MRI scans were used to identify the patterns in brain activity that occur when people view various types of pictures. The results showed that the technology recreated recognizable images of objects, food, faces, landscapes and more complex visual displays.
The technology has been described as mind reading, but Brain-IT does not sit inside someone's brain and randomly pull up thoughts. The person must be inside an MRI scanner and viewing an image.
Using the resulting brain activity, Brain-IT works backwards to create an estimated representation of what that person viewed. There is significant interest in how quickly the system can adapt to new participants.
Older methods typically required many hours of brain scan data from the same participant before they could generate useful reconstruction information. Brain-IT reportedly requires approximately one hour of learning about a new participant.
Researchers achieved that by focusing on common patterns that exist throughout all people's brains, rather than treating each person's brain as completely distinct. They broke brain activity into tens of thousands of very small components.
They identified those signals with general characteristics of visual stimuli, such as color and location, and with higher-level categories such as faces, food and sports.