For several years, the implants proposed by Elon Musk for connect the brain to a computerNeuralink, have taken much of the press for their results. But now a contender has emerged: MarkZuckerberg.
As detailed in studies published by Meta, the authors used a latest generation brain scanner and an artificial intelligence model of deep learning to interpret The neuronal signals of the people while writing, guessing which keys they were pressing with precision high enough to allow them to rebuild complete sentences.
But do not expect the potentially revolutionary gadget to reach the market … never. This is because the platform is built on a prohibitively large and expensive magnetoencephalography scanner, which Detect magnetic signals in the brain. The advantage is that you can scrutinize your mind without having to place a device, such as a brain-computer interface, within your skull, an invasive method that prefer other mental typification techniques.
The problem is that it is so difficult to handle as a magnetic resonance machine, weighs around half a ton and costs 2 million euros. Not only that, but the scanner can only work in a protected room that cushions the Earth’s magnetic field, which would otherwise prevent our weak brain signals from being detected. And while using, the subject cannot move the head at all, or the signal spoils.
Thus, the conditions are many and too important to allow the device to be marketed. And yet it is an undeniable achievement achievement, and Meta believes that you can use what you have learned here to give an advantage in the development of other AI models.
“Trying to understand the architecture or the precise principles of the human brain could be a way of informing the development of artificial intelligence. That is the way”, Says Jean-Réli King, leader of the Brain & Ai Meta.
According to the study, the system is able to correctly detect which keys clicks An expert typing up to 80 % of the time. That is not perfect, but it is precise enough to build complete sentences from brain signals, according to King’s team. But the success rate increases through the deep learning system called Brain2qwerty, which can learn which keys a user is pressing after observing it for several thousand drives.
Even so, the target system is not likely Provide a direct route to the practical applications of technology. However, the authors are excited, since what they have discovered seems to confirm the theory that our mind forms linguistic signals in a hierarchical way, which could be a “holy grail” for research in AI.
“Language has become the base of the AI - King concludes -. Therefore, Computational principles that allow the brain, or any system, acquire such capacity are the key motivation behind this work”