Meta’s AI research team, in collaboration with the Basque Center for Cognition, Brain, and Language, has made significant strides in decoding human thoughts, developing an AI model capable of reconstructing sentences from brain activity with 80% accuracy.

Advancing Non-Invasive Brain-Computer Interfaces

Unlike traditional brain-computer interfaces that require surgical implants, Meta’s approach utilizes magnetoencephalography (MEG) and electroencephalography (EEG) to measure brain activity without invasive procedures. The AI model was trained using brain recordings from 35 volunteers as they typed sentences.

When tested on new phrases, Meta claims its AI can accurately predict up to 80% of written characters using MEG data—twice as effective as EEG-based decoding methods. However, the technology remains limited, as MEG requires a magnetically shielded environment, and subjects must remain still for precise readings. Additionally, it has only been tested on healthy individuals, leaving its effectiveness for those with neurological conditions uncertain.

Mapping the Brain’s Language Processing

Beyond decoding thoughts into text, Meta’s AI is shedding light on how the brain transforms thoughts into language. By analyzing MEG recordings at the millisecond level, researchers have identified a “dynamic neural code” that links different stages of language formation while retaining access to prior information. This discovery helps explain how humans construct sentences effortlessly in real time.

Future Prospects and Challenges

While the findings hint at a future where non-invasive brain-computer interfaces assist those unable to speak, the technology is still far from practical application. Accuracy must be improved, and the logistical constraints of MEG devices pose a challenge for real-world use.

Meta is actively investing in advancing this research, announcing a $2.2 million donation to the Rothschild Foundation Hospital and collaborating with institutions like NeuroSpin, Inria, and the CNRS in Europe to refine the technology.

For now, thought decoding remains an exciting but experimental frontier—one step closer to bridging the gap between human cognition and artificial intelligence.