This AI computer can read your mind
Machine-learning technology probes deep into brain to identify song subjects are listening to.
Mind-reading technology may be closer to reality than you might think.
An international team of scientists, including researchers from the D'Or Institute for Research and Education in Rio de Janeiro, have used a Magnetic Resonance (MR) machine to read people's minds and identify what song they were listening to.
Six participants listened to 40 different pieces of music ranging from classical to rock, pop, jazz and other types while being monitored by the MR machine, which was linked to a computer.
The computer was fitted with special software which learned to identify the 'neural fingerprint' of each song, or the specific brain patterns associated with it. It did this by searching for musical features such as rhythm, tonality, dynamics and timbre.
After the tests, the computer was given the option of two songs and asked to identify which one the participant had been listening to. It was accurate 85% of the time, which is significantly more accurate than previous studies.
The researchers then made the task more difficult by providing the AI with a choice of 10 songs where there was only one correct answer. In this case, the computer correctly identified the song 74% of the time.
The findings are described in a paper published in the journal Scientific Reports.
So-called 'brain-decoding' techniques, such as these, may pave the way for research into reconstructing auditory thoughts and inner speech. This may prove hugely beneficial for people with locked-in syndrome, as it could lead to the development of advanced brain-computer interfaces that would allow people with that syndrome to communicate.
Sebastian Hoefle, an author of the study, speculated that one day "machines will be able to translate our musical thoughts into songs."
