Artificial intelligence that can read thoughts and convert them into text is introduced by Meta
Meta has revealed an AI system that can convert ideas into written language. It sounds like science fiction, but the tech giant has managed to bring it to life. The research project aims to decode brain activity and convert it into digital text, opening up new horizons in human-machine interaction.
A groundbreaking system that reads thoughts has been discovered through research by Meta's Fundamental AI Research Center (FAIR) in Paris in partnership with the Basque Centre for Cognition. Using a sophisticated scanner, the researchers were able to decode up to 80% of the characters that 35 participants thought they were writing. The results were astonishing: the generative system was able to capture not only words but also the semantic context of thought.
- How does this "mind reading" work?
The researchers used magnetoencephalography (MEG) and electroencephalography (EEG) technology, a synergistic system based on Brain2Qwerty, an AI model that interprets the brain’s magnetic signals as if they were keys being pressed on an invisible keyboard. It works like a real-time interpreter that can non-invasively read brain activity and translate thoughts into written words.
During the test, the volunteer sits inside a MEG scanner helmet, which detects the magnetic signals generated by neurons firing in the brain. The AI model then analyzes which signals correspond to which keys. After proper training, the machine is able to predict which letters the person intends to type.
“By taking 1,000 images of the brain every second, we can pinpoint the exact moment when thoughts turn into words, syllables, and even letters,” Meta explains in detail. “The study shows that the brain generates a series of representations that start from the abstract meaning of a sentence and gradually translate into concrete actions, such as moving fingers on a keyboard.” This is called “dynamic neural codes,” a special neural mechanism that links successive representations, maintaining each one for long periods of time.
Despite the revolutionary premise, the project still has several practical limitations. First, the scanner must be placed in a specially shielded room to block out the Earth’s magnetic field, which is a trillion times stronger than brain signals. In addition, even the slightest head movement can distort the signal, so it is essential to remain still. Finally, although this research was conducted on healthy volunteers, more work is needed to explore how useful it is for people with brain injuries.