You might cringe at the prospect of someone utilising artificial intelligence to read your thoughts, which are probably the only aspects of human nature that are private and unavailable to others. A novel technology that can read thoughts and convert them into a continuous stream of text has been studied and published by University of Texas researchers.
The study describes how the authors trained a semantic decoder to be able to interpret a subject's brain activity while the subject listened to or silently imagined stories and watched silent videos, producing text that precisely corresponds to what was heard, thought, or watched using functional magnetic resonance imaging (fMRI).
The decoder is an uninvasive device that picks up new information from the brain activity that is monitored by an fMRI scanner as the subject consumes hours of podcasts. In doing so, the system learns how to analyse and connect the input data with the scanned brain activity in order to understand the subject's future ideas.
After watching four silent movies, listening to a new narrative, or imagining one, the person, the decoder was able to decode their brain activity and produce writing that corresponded to their ideas.
It's not as sinister as it sounds for AI to read minds. People who are conscious but unable to speak, such as those with physical disabilities, could communicate more effectively if they were able to decode one another's thoughts. This study shows that linguistic brain-computer interfaces can be used to do this.
How accurate is AI at reading minds?
Despite its ground-breaking nature, this study is far from complete. The semantic decoder produces a continuous stream of text, but it does not give a verbatim account of the speaker's thoughts. However, the system used in this study is able to decode continuous language for extended periods of time that contains complex ideas rather than just words or simple phrases.
And it's not as efficient as you might think. According to the study, only around half of the time did the semantic decoder properly produce text that corresponded to the subject's thinking.
Given that this is the first successful trial of a non-invasive semantic decoder that doesn't require surgical implants, half the time is still a rather outstanding outcome.
Can this AI read my mind?
Though the semantic decoder developed by the researchers at UT of Austin is capable of deciphering and reconstructing a person's thoughts to display them in text, it won't work on anyone.
For starters, the system requires an fMRI scanner to capture an individual's brain activity, both during training and testing, though it could be done with other technologies in the future, like functional near-infrared spectrocopy (fNIRS).
The UT researchers respect people's right to their own privacy in their minds and want this technology to only be utilised by those who can benefit from it. They demonstrated in the study that the semantic decoder only functions with consenting people who wish to interact with it, reducing the possibility of this technology being misused.
The system was unable to successfully decipher ideas from people it had not been taught on or from people who had changed their minds after training.