An AI that can output sentences without implanting human thoughts in the brain will be developed

A research team at the University of Texas at Austin has developed an AI system that can non-invasively scan the brain and verbalize thoughts. In the experiment, MRI was used to scan the brain, but it has been shown that it is possible to verbalize thoughts even with devices smaller than MRI.

Semantic reconstruction of continuous language from non-invasive brain recordings | Nature Neuroscience

https://cns.utexas.edu/news/podcast/brain-activity-decoder-can-reveal-stories-peoples-minds

Technologies that read and output human thoughtsalready exist , but existing systems require devices to be implanted in the brain. A new system developed by a research team at the University of Texas at Austin has made it possible to verbalize thoughts using brain data scanned by MRI.

The research team scanned brain data while listening to 16 hours of podcasts, and trained AI using the scan data. AI is tuned to recognize the content of thoughts as a whole rather than recognizing words one by one. ” is recognized.

The figure below shows the results of ``experiment to read brain data when listening to the subject'', the left side is ``transcription of the speech actually heard by the subject'', and the right side is ``AI thinks It was read and written down.” Looking at the figure, you can see that although the sentences are not reproduced according to the word order, the contents of the sentences are reproduced.

The following is the result of thinking reading when the subject is allowed to view a short silent movie. AI believes that the subject 'i see a girl that looks just like me get hit on her back and then she is knocked off' detected.

According to the research team, although MRI was used in this experiment, in theory NIRS , which is smaller than MRI, can be used to read thoughts. The research team states that NIRS can be used to read thoughts to develop a portable thought reading device.