{"id":828490,"date":"2025-02-20T12:12:33","date_gmt":"2025-02-20T18:12:33","guid":{"rendered":"https:\/\/newsycanuse.com\/index.php\/2025\/02\/20\/meta-has-an-ai-for-brain-typing-but-its-stuck-in-the-lab\/"},"modified":"2025-02-20T12:12:33","modified_gmt":"2025-02-20T18:12:33","slug":"meta-has-an-ai-for-brain-typing-but-its-stuck-in-the-lab","status":"publish","type":"post","link":"https:\/\/newsycanuse.com\/index.php\/2025\/02\/20\/meta-has-an-ai-for-brain-typing-but-its-stuck-in-the-lab\/","title":{"rendered":"Meta has an AI for brain typing, but it\u2019s stuck in the lab"},"content":{"rendered":"
\n
\n
\n
\n
\n

The social network company hopes neuroscience will give it an advantage in the AI race.<\/p>\n<\/div>\n

\n

\"""\"<\/span><\/p>

Stephanie Arnett\/MIT Technology Review | Getty, Alamy, Shutterstock<\/span><\/figcaption><\/div>\n<\/div>\n<\/header>\n<\/div>\n
\n
\n

Back in 2017, Facebook unveiled<\/a> plans for a brain-reading hat that you could use to text just by thinking. \u201cWe\u2019re working on a system that will let you type straight from your brain,\u201d CEO Mark Zuckerberg shared<\/a> in a post that year.<\/p>\n

Now the company, since renamed Meta, has actually done it. Except it weighs a half a ton, costs $2 million, and won\u2019t ever leave the lab.<\/p>\n<\/p><\/div>\n

\n

Still, it\u2019s pretty cool that neuroscience and AI researchers working for Meta have managed to analyze people\u2019s brains as they type and determine what keys they are pressing, just from their thoughts.<\/p>\n

The research, described in two papers posted by the company (here<\/a> and here<\/a>), as well as a blog post<\/a>, is particularly impressive because the thoughts of the subjects were measured from outside their skulls using a magnetic scanner, and then processed using a deep neural network.<\/p>\n

\u201cAs we\u2019ve seen time and again, deep neural networks can uncover remarkable insights when paired with robust data,\u201d says Sumner Norman, founder of Forest Neurotech, who wasn\u2019t involved in the research but credits Meta with going \u201cto great lengths to collect high-quality data.\u201d<\/p>\n

According to Jean-R\u00e9mi King, leader of Meta\u2019s \u201cBrain & AI\u201d research team, the system is able to determine what letter a skilled typist has pressed as much as 80% of the time, an accuracy high enough to reconstruct full sentences from the brain signals.<\/p>\n<\/div>\n

\n

Facebook\u2019s original quest for a consumer brain-reading cap or headband<\/a> ran into technical obstacles, and after four years, the company scrapped the idea.<\/a><\/p>\n

But Meta never stopped supporting basic research on neuroscience, something it now sees as an important pathway to more powerful AIs that learn and reason like humans<\/a>. King says his group, based in Paris, is specifically tasked with figuring out \u201cthe principles of intelligence\u201d from the human brain.<\/p>\n

\u201cTrying to understand the precise architecture or principles of the human brain could be a way to inform the development of machine intelligence,” says King. \u201cThat\u2019s the path.\u201d<\/p>\n

The typing system is definitely not a commercial product, nor is it on the way to becoming one. The magnetoencephalography scanner used in the new research collects magnetic signals produced in the cortex as brain neurons fire. But it is large and expensive and needs to be operated in a shielded room, since Earth\u2019s magnetic field is a trillion times stronger than the one in your brain.\u00a0<\/p>\n<\/p><\/div>\n

\n

Norman likens the device to \u201can MRI machine tipped on its side and suspended above the user\u2019s head.\u201d<\/p>\n

What\u2019s more, says King, the second a subject\u2019s head moves, the signal is lost. \u201cOur effort is not at all toward products,\u201d he says. \u201cIn fact, my message is always to say I don\u2019t think there is a path for products because it\u2019s too difficult.\u201d<\/p>\n

The typing project was carried out with 35 volunteers at a research site in Spain, the Basque Center on Cognition, Brain, and Language. Each spent around 20 hours inside the scanner typing phrases like \u201cel procesador ejecuta la instrucci\u00f3n” (the processor executes the instruction) while their brain signals were fed into a deep-learning system that Meta is calling Brain2Qwerty, in a reference to the layout of letters on a keyboard.<\/p>\n

The job of that deep-learning system is to figure out which brain signals mean someone is typing an a, which mean z, and so on. Eventually, after it sees an individual volunteer type several thousand characters, the model can guess what key people were actually pressing on.\u00a0<\/p>\n

In the first preprint, Meta researchers report that the average error rate was about 32%\u2014or nearly one out of three letters wrong. Still, according to Meta, its results are most accurate yet for brain typing using a full alphabet keyboard and signals collected outside the skull.<\/p>\n

Research on brain reading has been advancing quickly, although the most effective approaches use electrodes implanted into the brain, or directly on its surface. These are known as \u201cinvasive\u201d brain computer interfaces. Although they require brain surgery, they can very accurately gather electrical information from small groups of neurons.<\/p>\n<\/p><\/div>\n

\n

In 2023, for instance, a person who lost her voice from ALS was able to speak<\/a> via brain-reading software connected to a voice synthesizer. Neuralink, founded by Elon Musk, is testing its own brain implant that gives paralyzed people control over a cursor.<\/p>\n

Meta says its own efforts remain oriented toward basic research into the nature of intelligence.<\/p>\n<\/p><\/div>\n

\n

And that is where the big magnetic scanner can help. Even though it isn\u2019t practical for patients and doesn\u2019t measure individual neurons, it is able to look at the whole brain, broadly, and all at once.\u00a0<\/p>\n

The Meta scientists say that in a second research effort, using the same typing data, they used this broader view to amass evidence that the brain produces language information in a top-down fashion, with an initial signal for a sentence kicking off separate signals for words, syllables, and finally typed letters.<\/p>\n

\u201cThe core claim is that the brain structures language production hierarchically,\u201d says Norman. That\u2019s not a new idea, but Meta\u2019s report highlights \u201chow these different levels interact as a system,\u201d says Norman.<\/p>\n

Those types of insights could eventually shape the design of artificial-intelligence systems. Some of these, like chatbots, already rely extensively on language in order to process information and reason, just as people do.<\/p>\n

\u201cLanguage has become a foundation of AI,\u201d King says. \u201cSo the computational principles that allow the brain, or any system, to acquire such ability is the key motivation behind this work.\u201d<\/p>\n

Correction: Meta posted two papers describing its brain-typing results on its website. An earlier version of this story incorrectly said they had been published at arXiv.org<\/em>.<\/p>\n<\/p><\/div>\n<\/div>\n<\/div>\n

Antonio Regalado
Read More<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

The social network company hopes neuroscience will give it an advantage in the AI race. Stephanie Arnett\/MIT Technology Review | Getty, Alamy, Shutterstock Back in 2017, Facebook unveiled plans for a brain-reading hat that you could use to text just by thinking. \u201cWe\u2019re working on a system that will let you type straight from your<\/p>\n","protected":false},"author":1,"featured_media":828491,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2452,51398],"tags":[8966,141864],"class_list":{"0":"post-828490","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-brain","8":"category-typing","9":"tag-brain","10":"tag-typing"},"aioseo_notices":[],"_links":{"self":[{"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/posts\/828490","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/comments?post=828490"}],"version-history":[{"count":0,"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/posts\/828490\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/media\/828491"}],"wp:attachment":[{"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/media?parent=828490"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/categories?post=828490"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/tags?post=828490"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}