Social media giant Meta recently announced that they have developed a new device that can input text by reading neural signals from the human brain. This research achievement is detailed by Meta's scientists in two studies, utilizing advanced brain scanning technology and deep learning AI models to successfully decode the brain signals of people while typing, even reconstructing complete sentences.
Specifically, this technology relies on a device called magnetoencephalography (MEG), which can capture the faint magnetic signals emitted by the brain. Compared to brain-machine interface technologies that require implantation, this device has the advantage of not needing invasive surgery and can operate without directly manipulating the brain. However, the device also has significant drawbacks: it weighs nearly half a ton, costs up to $2 million, and can only be used in specially shielded rooms to prevent interference from the Earth's magnetic field. Additionally, during use, subjects must keep their heads still, as any movement may result in signal loss.
Nevertheless, this technology still demonstrates extraordinary potential. According to researchers, the system can detect the key presses of a "skilled" typist with up to 80% accuracy. Although this accuracy is not perfect, it is sufficient to construct complete sentences by decoding brain signals. To achieve this, the research team designed a deep learning system called "Brain2Qwerty," which learns the key presses by observing thousands of characters input by users.
While the current technology is still a considerable distance from practical application, Meta's researchers are optimistic about this discovery. They believe that this research validates the theory that the human brain follows a hierarchical structure in language formation, which could greatly benefit further research in artificial intelligence. Jean-Rémi King, head of Meta's Brain and AI team, stated that understanding how the human brain works may provide new insights for the development of machine intelligence.
Key Points:
🧠 Meta has developed a non-invasive brain-machine interface device that can input text through brain signals.
💰 The device weighs half a ton, costs up to $2 million, and requires use in special environments.
📊 Currently, the accuracy is at 80%, but improvements are needed, and there is still a way to go before practical application.
