Artificial intelligence has learned to read thoughts
A new algorithm for deep learning scientists from the University Carnegie Mellon can “read” a person’s thoughts by analyzing fMRI scans of the brain.
About it reported in a press release from the University.
“One of the great achievements of the human brain — the ability to combine individual concepts into complex thoughts. For example, to think not only about “bananas”, but that “I like to eat bananas in the evening with my friends,” said the study’s author Marcel Smith. — We’ve finally developed a way to see how to form complex thoughts using functional magnetic resonance imaging (fMRI) of the brain. The discovery of the link between thoughts and brain activity shows how the process of thinking in humans,” added Smith.
Machine learning algorithms have analyzed the fMRI scans of the brain. They then in sequence built a chain, which is used by the brain to create certain thoughts.
In the process of research, scientists have identified about 240 reactions of the brain to separate people, places, or physical impact. The algorithm used fMRI scans with these reactions to predict what a person thinks.
AI compared the data from each fMRI scan with the already known reactions and tried to guess the thoughts of the person. AI guessed thoughts with unexpectedly high accuracy of 87%. Moreover, AI was able to perform the operation in reverse order, having reproduced received from the person a proposal for an fMRI image.
Scientists believe that the results of the study reveal how far gone is the deep learning algorithms, which even learned to read minds.