Meta AI releases Brain2Qwerty v2, a non-invasive decoder that translates raw neural signals into full sentences in real time

To help accelerate neuroscience breakthroughs, we're releasing the full training code for Brain2Qwerty v1 and v2, and our partner, @bcbl_, is releasing the v1 dataset.

Learn more and explore the artifacts here: https://go.meta.me/42ed9c

We trained Brain2Qwerty v2 on ~22,000 sentences from 9 volunteers, each recorded for 10 hours wearing an MEG device while typing.

By using end-to-end deep learning on raw brain signals from MEG devices and fine-tuning LLMs, the system effectively bridges the gap between noisy neural data and coherent language.

The results are promising: - Avg word accuracy of 61% across participants - 78% word accuracy and 50%+ of sentences decoded with ≤ 1 word error for the top-performing participant - Performance scales log-linearly with data volume

Highly recommended reading.

What an impressive use of LLMs and deep learning.

Achieves "real-time sentence decoding from non-invasive brain recordings, approaching levels of accuracy previously exclusive to techniques that require brain surgery."

@bcbl_ For clarification, Brain2Qwerty v1 was published earlier today in @NatureNeuro.

It's hard to read Meta working on something like this without tacking on "...and think how much better we'll be able to target ads like this!"

@AIatMeta @Nature The last thing I'd want is Zuckerberg to have access to my brain even if I was disabled.

@AIatMeta @Nature @Nature would have been great, but this is @NatureNeuro, here is the link: https://www.nature.com/articles/s41593-026-02303-2

@AIatMeta @Nature soon I can code at full bandwidth

@AIatMeta @Nature That is so cool.

@AIatMeta @Trace_Cohen @Nature @grok has neuralink ever published in a journal like nature?

@AIatMeta @Nature if this can be miniaturised.. its gonna be crazy

@AIatMeta @Nature "No need to talk, we KNOW you dislike the Party, the scanner flagged you. Yes, that's also why your carbon pass and all your social scores had been terminated. Now please step on the circle and look at this red light..."

@AIatMeta @Nature I don't trust zucc with this power.

@AIatMeta @Nature 9 volunteers, 10 hours each, and it's already decoding semantics in real time

that's surprisingly small dataset for something this complex...

@AIatMeta @Nature Stop it. We know you are only doing this to optimize advertising. It's gross.

@AIatMeta @Nature @threadreaderapp please #unroll

@AIatMeta @Nature @Grok can you help clarify what this image shows? Also, is 10 hours per volunteer enough for robust training?

@AIatMeta @Nature I have no idea what this is but it has a gif with a brain so I will hope it is something good

@AIatMeta @Nature This is the only thing I can think of that could beat the devex of dictation

@AIatMeta @Nature So Meta can read mind in few words