27 June 2020
From ancient times to today, the most impressive thing that artificial intelligence has done is: a computer that is very good at chess.
Nowadays, a variety of software can do more and more things. Not only can artificial intelligence be able to chat with us through Facebook Messenger, but also independently operate the Curiosity Mars Rover when human engineers snore. Now, a team of scientists at Carnegie Mellon University has developed an artificial intelligence that can do something that was once thought impossible: reading the human brain.
The teams new software uses a new method to guess the internal activity of the human brain. The data is first collected by MRI technology, then by observing the brains activity patterns when generating these data, and finally by reverse detection to predict human thinking.
The research leader MarcelJust explained: "A major feature of the human brain is the combination of independent concepts to form complex ideas. It thinks not only of bananas but of me I like to eat bananas with my friends at night. We finally found a way to read this complicated idea from the imaging signals of MRI. The correspondence between the idea and the pattern of brain activity shows us that the idea is How to build it."
The study found hundreds of different patterns and interactions, including social and physiological interactions, involving more than 200 different events, and the sentence structure of discourse. Using this data to train the algorithm, so that the algorithm learns to detect the same pattern, can accurately predict 90% of a persons speech content.