In 2017, Facebook (now Meta) had a bold idea. With a brain-reading hat, you can input using just your thoughts. A few years later, the company actually built something similar, but the catch means it’s not close to everyday use.
Meta’s brain-type system uses artificial intelligence (AI) and neuroscience to analyze brain activity and predict keys that people are pressing based solely on their own thoughts. But there’s a catch. The system requires large, expensive machines and only works in highly controlled lab settings.
How Meta’s Brain Typing AI Works
According to a recent blog from MIT Technology Review, the technology relies on a special brain scanner called a Magnetic Electroencephalography (MEG) machine that detects small magnetic signals generated by brain activity. The scanner is so large and sensitive that it must be stored in a special room to block interference from the Earth’s magnetic field.
Meta researchers trained an AI model called Brain2QWerty to analyze these brain signals. As a volunteer entered into a keyboard, AI has learned to match patterns of data to specific characters. Over time, the system has become accurate enough to correctly predict the letters a person had thought about about 80% of the time.
The experiment was attended by 35 volunteers at a Spanish research centre. Each person spent about 20 hours inside the scanner, entering sentences while the AI was studying brain activity.
Challenge: Why is this not becoming a product in no time?
Despite its advances, Meta’s brain typing system is far from becoming a consumer product. There are several reasons for this:
• Huge and expensive equipment: The MEG scanner weighs 0.50 tons, costing around $2 million, making it unrealistic for everyday use.
• You still need the head completely. If a person moves slightly, brain signals become unreadable.
• Still not accurate enough: the system is cutting edge of this kind, but about 32% of the letters are wrong.
Rather than trying to create a product, Meta research team, led by Jean-Rémi King, wants to learn more about how human intelligence works. Understanding how to build structural language in your brain can help improve AI systems such as chatbots and language models.
Meta’s brain typing AI is not ready for the real world, but the brain computer interface (BCIS) is moving forward quickly. Some of the most effective systems today use electrodes embedded in the brain, allowing paralyzed patients to control computers or “speak” via synthesized voices.
Companies like Neuralink, founded by Elon Musk, develop brain implants that can restore movement and communication for people with disabilities. Meta research focuses on non-invasive methods (no surgery), but there is still a long way to go before brain-controlled typing becomes practical.
For now, Meta’s brain typing AI remains an exciting scientific achievement, but is likely to remain in the lab.
