A proof-of-concept neural brain implant providing speech(arstechnica.com)
arstechnica.com
A proof-of-concept neural brain implant providing speech
https://arstechnica.com/science/2025/06/a-neural-brain-implant-provides-near-instantaneous-speech/
73 comments
Curious how much intentionality is required from the user to produce sounds. It would be unfortunate if this device just started firing off speech for what would otherwise be thoughts one would not say out loud. I suppose that depends on the mechanism required to activate the neurons to which the device is connected.
Even without this device there's been some consideration to the thought that the conscious brain is merely an observer since it appears to activate after the unconscious brain takes actions. You just go along with what the unconscious mind did in actions and speech and you convince yourself you meant to do that after the fact.
https://pmc.ncbi.nlm.nih.gov/articles/PMC3746176/
So here it could indeed just fire off speech and you know what? We'd probably convince ourselves that we absolutely meant to do that. In fact it could be a very interesting experiment (with willing participants). Mess with the inputs the device receives so it's not really the person activating it, let it do it's thing and see if they notice when they do/don't have control of it.
https://pmc.ncbi.nlm.nih.gov/articles/PMC3746176/
So here it could indeed just fire off speech and you know what? We'd probably convince ourselves that we absolutely meant to do that. In fact it could be a very interesting experiment (with willing participants). Mess with the inputs the device receives so it's not really the person activating it, let it do it's thing and see if they notice when they do/don't have control of it.
I find this simultaneously fascinating and disturbing; once you have someone using this, they become a hitherto-impossible human-AI hybrid, where their mind is now a fusion between the two, completely unnoticed to the user.
Funny, I was just pondering this last night - how I often realize this phenomenon immediately after it occurs, and feel helpless when I realize it. It's like two minds, one that is moving my hands to pile mangoes into my face, as the other is telling me to save some mangoes for later (they're very hard to stop eating). But on a more serious note, now that I realize I realize it, I realize it a lot.
I once got too enthusiastic about eating mangoes and my face learned that the skin contains urushiol like poison ivy.
Even though the conscious brain doesn't always directly control motor skills, I wouldn't call it "merely an observer". It's your consciousness that decides your goals and beliefs; the rest of the body "learns" those things and is then preconditioned to react accordingly.
But they are separate systems. Harsh experiences (traumas) can teach your body some bad lessons in such a way that not even your conscious mind can overcome. In these situations, you can't "think your way out" of these traumatic consequences in, say, a talk therapy session; you need to deal with the body directly/somatically to recover instead.
But they are separate systems. Harsh experiences (traumas) can teach your body some bad lessons in such a way that not even your conscious mind can overcome. In these situations, you can't "think your way out" of these traumatic consequences in, say, a talk therapy session; you need to deal with the body directly/somatically to recover instead.
> It's your consciousness that decides your goals and beliefs; the rest of the body "learns" those things and is then preconditioned to react accordingly.
Quotation needed.
Quotation needed.
It’s intentional and requires quite a lot of focus.
The original paper [0] mentions electrodes are placed over Broca’s area (speech production, translates words to mouth movements) and motor area (adjusts the mouth movements). It’s attempted speech, not thoughts.
There is a lot of fear in mainstream media and populace of devices decoding thoughts, but that is a significantly harder problem, at this moment on the level of sci-fi of Civilisation Type II on Kardashev scale. There is a reason why the electrodes are not over Wernicke’s area instead (language comprehension and production).
0: https://doi.org/10.1038/s41586-023-06377-x
The original paper [0] mentions electrodes are placed over Broca’s area (speech production, translates words to mouth movements) and motor area (adjusts the mouth movements). It’s attempted speech, not thoughts.
There is a lot of fear in mainstream media and populace of devices decoding thoughts, but that is a significantly harder problem, at this moment on the level of sci-fi of Civilisation Type II on Kardashev scale. There is a reason why the electrodes are not over Wernicke’s area instead (language comprehension and production).
0: https://doi.org/10.1038/s41586-023-06377-x
Without having RTFA, I'd guess/predict that it will be possible to learn to only do this intentionally, much like we can think about raising our arm without actually raising it.
I hope it's something better than "Hey Siri, say..."
Otherwise, yeah, that would be a new sort of hell where you had no private inner monologue
Otherwise, yeah, that would be a new sort of hell where you had no private inner monologue
The brain is incredibly adaptive, I guarantee eventually it would learn to avoid firing the neurons the device is probing when you don't want your inner monologue spoken aloud as long as there's a feedback loop where you experience negative emotions when something you didn't want spoken aloud was broadcast.
Very excellent point.
One simple possibility I could imagine to mitigate this would be having an additional device/button one had to physically press in order for the thoughts to be vocalized. I think that would be an extremely simple but clear indication of intentionality.
I often wonder how such teams build their devices - I assume it requires quite a few pieces of equipment that can't be bought at a nearby shop. Are such devices ordered from some manufacturer or are they built in-house somehow?
There are specialized companies selling components, what you cannot get you manufacture yourself. Assembling the device for prototypes like this is usually done in house.
Established labs often have a specialised section that cultivates all the little tricks how to do these things.
Knowledge is transferred by hiring postdocs that have the skills you need or by sending a phd student over to be trained.
As a scientist, if you do this for a while you end up with insane skills, but there is no place for them on the job market.
Everybody else is living 15 years in the past from your perspective.
Established labs often have a specialised section that cultivates all the little tricks how to do these things.
Knowledge is transferred by hiring postdocs that have the skills you need or by sending a phd student over to be trained.
As a scientist, if you do this for a while you end up with insane skills, but there is no place for them on the job market.
Everybody else is living 15 years in the past from your perspective.
University researches in general? The boss always knows a small company that can make one, and staples a bidding notice to campus notice board for formality as the device is getting made and delivered. Isn't that how they do anywhere it might be?
The main challenge appears to be the neural-computer interface - the electrodes. As the article states, there are several startups in this space all bottlenecked by the same constraint, and accurately translating neural impulses into digital (Or even analog) signals is the key to unlocking a whole arena of transhuman development.
Most such startups are scaling up the number of electrodes interfacing with the neurons to overcome this bottleneck, but I wonder if an unconventional approach could overcome the limit more gracefully. I may be a dreamer, but a high fidelity synthetic neural fiber is the holy grail here. I do remember reading people partially healed of paralysis due to spinal injury, because of electrical conduits that bridged the injured neural gap.
Most such startups are scaling up the number of electrodes interfacing with the neurons to overcome this bottleneck, but I wonder if an unconventional approach could overcome the limit more gracefully. I may be a dreamer, but a high fidelity synthetic neural fiber is the holy grail here. I do remember reading people partially healed of paralysis due to spinal injury, because of electrical conduits that bridged the injured neural gap.
> In this second test, the word error rate was 43.75 percent, meaning participants identified a bit more than half of the recorded words correctly.
> [...]
> “We’re not at the point where it could be used in open-ended conversations. I think of this as a proof of concept,” [Sergey Stavisky, a neuroscientist at UC Davis and a senior author of the study] says.
The ability to produce sound without a use of a dictionary sounds awesome. It is an interesting result, a proof of concept as the author of the study says, but the title is editorialized at best and effectively clickbait at worst, because most readers will assume that "near instantaneous speech" means "clear intelligible speech and ability to communicate".
> [...]
> “We’re not at the point where it could be used in open-ended conversations. I think of this as a proof of concept,” [Sergey Stavisky, a neuroscientist at UC Davis and a senior author of the study] says.
The ability to produce sound without a use of a dictionary sounds awesome. It is an interesting result, a proof of concept as the author of the study says, but the title is editorialized at best and effectively clickbait at worst, because most readers will assume that "near instantaneous speech" means "clear intelligible speech and ability to communicate".
Ok, I've taken "near instantaneous" out of the title and put "proof of concept" in there, which is a phrase used by one of the researchers in the article.
That's the plot of movies Upgrade (2018) and Rakka (2017).
Telepathy is on its way. Next step they just skip the conversion of brain signals to words and just directly send the signals to another brain. But I think some conversion/translation would still be necessary.
The year is 2100. The brain of the Eurasian president got hacked by the Antartic Federation. While humans have hard coded a moral code since birth, there are illegally born babies that do not undergo brain modification treatment. Moreover, the South Pacific Whale Society has no moral code. We should’ve never implanted this stuff into whales. The world will never be the same again.
Ghost in the Shellfish.
I'll see myself out.
I'll see myself out.
>The year is 2100. The brain of the Eurasian president got hacked by the Antartic Federation.
They'd stil have a president? They would probably already have a dictator that controls everybody through a mind-reading police state...
They'd stil have a president? They would probably already have a dictator that controls everybody through a mind-reading police state...
Or they are a hive mind, like Alastair Reynolds' Conjoiners.
Lots of dictatorships use democratic language. See Putin, the President of the Russian Federation. Or the Democratic People's Republic of Korea.
No, that's the Republic of Murica, where Trump The Third is a BDFL.
I think it would be deserved if whales started bombing our streets.
So Long, and Thanks for All the Krill
I think (futurology / science fiction) that they will make some kind of brain link, but there won't be any translations happening in between, just raw brain signals from one to the other, like an extra sensory input; there won't be any encoding or data that can be translated to speech or images, but the connected brains will be able to learn to comprehend and send the signals to / from each other and learn to communicate that way.
That sort of connection would be very susceptible to psychic attacks - I'm thinking of the telepaths in Babylon 5, being trained for offensive capabilities, as well as just plain old spam advertising. So while "defaulting to trust" is often considered societally useful, I believe that it would be better for everyone if cross-brain messages are sent in a format that can be analyzed (and entirely blocked) by a filter on the receiving side.
I still fail to see how that's possible since it is assumed every brain "encodes" data uniquely. Communication between computers is possible because we have agreed upon standards. If every computer encoded characters differently, no communication would be possible. Without agreed upon ports or agreed upon mechanism to agree upon ports one computer could not communicate with another. So how can brain-to-brain communication work given that encoding/communication "standards" are impossible since each brain is different?
For example, I see a tree and my brain generates a unique signal/encoding/storage representing the tree. Another person sees the tree and generates a unique signal/encoding/storage representing the tree. How would my brain communicate "tree" to his brain since both our "trees" are unique to our brains?
My brain device reads my brain signal "1010101" for tree. My friend's device reads brain signal "1011101" for tree. How could we possibly map 1010101 to 1011101. Or is the assumption that human brains have identical signals/encoding for each thought.
For example, I see a tree and my brain generates a unique signal/encoding/storage representing the tree. Another person sees the tree and generates a unique signal/encoding/storage representing the tree. How would my brain communicate "tree" to his brain since both our "trees" are unique to our brains?
My brain device reads my brain signal "1010101" for tree. My friend's device reads brain signal "1011101" for tree. How could we possibly map 1010101 to 1011101. Or is the assumption that human brains have identical signals/encoding for each thought.
I already learned to interpret touch, taste, smision etc. when I was just a baby. How hard can a new one be?
Did you even read my comment? I'm not talking about your brain's ability. I'm talking about how a device can interpret your brain signals and transfer it to another brain to have actual communication when both brains essentially have their own internal "language". How your brain "stores" the idea of tree is entirely different from how I "store" the idea of a tree. Different neurons, different location in the brain and different signals. Do you know computers, networks and communications work? It's all bound by artificial standards we agreed upon a priori.
The only way I see is by textual or auditory mechanism between people who speak the same language ( standards agreed upon a priori ). But that wouldn't be brain to brain. It would be brain to text/speech to eyes/ears to brain.
The only way I see is by textual or auditory mechanism between people who speak the same language ( standards agreed upon a priori ). But that wouldn't be brain to brain. It would be brain to text/speech to eyes/ears to brain.
I'm saying it's clearly possible to develop from scratch the framework with which to learn to interpret each individual remote machine's raw data stream.
Your intermediate protocol woes are a red herring.
Your intermediate protocol woes are a red herring.
There would probably be a Universal Common Embedding used as an intermediate representation between people's individual private neural representations. Likely the distant descendant of our open-source neural models.
And machines would of course also use the Universal Common Embedding to communicate, as man and machine meld into a seamless distributed whole.
It all seems a little bit too inevitable for my liking at this point.
And machines would of course also use the Universal Common Embedding to communicate, as man and machine meld into a seamless distributed whole.
It all seems a little bit too inevitable for my liking at this point.
We are so different, but I guess with a lot of training we could interpret each others thoughts. A first step would be to record your own thoughts and then replay them in order to see if you experience the same thing you did when the thoughts where recorded. It's possible that our brain is constantly re-configuring so that even your own recorded thoughts would make no sense.
This is sort of explored in the film Strange Days, which is probably not very well known generally but perhaps would have a large fan base on HN.
https://en.wikipedia.org/wiki/Strange_Days_(film)
Had no idea til I looked it up just now that James Cameron did the story, of Avatar, which shares a lot of tech influences with Strange Days. They could even be in the same cinematic universe, though many years apart.
https://en.wikipedia.org/wiki/Strange_Days_(film)
Had no idea til I looked it up just now that James Cameron did the story, of Avatar, which shares a lot of tech influences with Strange Days. They could even be in the same cinematic universe, though many years apart.
I think the main problem with that is that different people thing in different ways. I think in full sentences and 3D images, whereas other people might think without images at all. How do you translate that?
It is very likely that this device works by perceiving and interpreting brain waves. Actually, from the article:
> “We recorded neural activities from single neurons, which is the highest resolution of information we can get from our brain,” Wairagkar says. The signal registered by the electrodes was then sent to an AI algorithm called a neural decoder that deciphered those signals and extracted speech features such as pitch or voicing.
> “We recorded neural activities from single neurons, which is the highest resolution of information we can get from our brain,” Wairagkar says. The signal registered by the electrodes was then sent to an AI algorithm called a neural decoder that deciphered those signals and extracted speech features such as pitch or voicing.
If statements
for how the brain chip chooses to function
for how the brain chip chooses to function
Next step: techno-slavery
We’ll probably call it something less on the nose, more euphemistic, almost clinical, medical even.
Nerve stapling, perhaps?
https://tvtropes.org/pmwiki/pmwiki.php/Fridge/SidMeiersAlpha...
George Carlin - Euphemisms - https://www.youtube.com/watch?v=isMm2vF4uFs
Nerve stapling, perhaps?
https://tvtropes.org/pmwiki/pmwiki.php/Fridge/SidMeiersAlpha...
George Carlin - Euphemisms - https://www.youtube.com/watch?v=isMm2vF4uFs
Don't forget the injection of ads for the basic subscription plan, Black Mirror S7/E1 vibes.
Telepathy, or maybe memory and experience sharing media machine, or maybe humanity id unification device, or maybe flesh robot actualizer, or maybe a looping torture horror show mask. The possibilities are endless!
I look forward to learning that your favourite fruit is AAAARRGGHHHH
Maybe he was dictating?
Not without ads being injected. But it's a small price to pay for such amazing capabilities
Like the trisolarans
would you need an antenna for your brain implant to tx/rx those signals? what shape would that antenna need to be? would it need to be a rigid antenna that would have to stick up out of one's head? would it be a curly pigtail style or straight semi-rigid yet flexible so it could bend when passing through doors like the CB antennas of yesteryear? Could it be very flexible like a cable so you could run it down your back and under your clothing?
would your antenna be susceptible to crosstalk, and would that interference come across as new voices in your head? in fact, i wonder what the signalling protocol would be so that the message is only be received by the intended recipient, something like wifi? to that, would someone be able to tune into the spectrum around them and see the metadata of these telepathic signals to see who was talking to whom, when, and for how long. obviously, i'm assuming the actual signals will be encrypted so that these conversations will be private. will the NSA be able to pick up these signals from their satellites and be able to listen in? or will the gov't force backdoors into these communications?
i know this might sound conspiratorial, but all of these are valid concerns being dealt with now, and only sound conspiratorial if you've have your head in the sand. these are also questions that startups tend to ignore. look at the IoT market that never considered any potential security issues, and now we have massive bot farms. i know i wouldn't want to be used as a bot because my neural implant manufacture never considered what a hacker might do once they gained access
would your antenna be susceptible to crosstalk, and would that interference come across as new voices in your head? in fact, i wonder what the signalling protocol would be so that the message is only be received by the intended recipient, something like wifi? to that, would someone be able to tune into the spectrum around them and see the metadata of these telepathic signals to see who was talking to whom, when, and for how long. obviously, i'm assuming the actual signals will be encrypted so that these conversations will be private. will the NSA be able to pick up these signals from their satellites and be able to listen in? or will the gov't force backdoors into these communications?
i know this might sound conspiratorial, but all of these are valid concerns being dealt with now, and only sound conspiratorial if you've have your head in the sand. these are also questions that startups tend to ignore. look at the IoT market that never considered any potential security issues, and now we have massive bot farms. i know i wouldn't want to be used as a bot because my neural implant manufacture never considered what a hacker might do once they gained access
I think you might find that the Teletubbies have already solved this problem, in a variety of different configurations, including the rigid and curly antenna you mention above. The benefits of the circular and triangular configurations are as yet unclear to me, but I'm sure the Voice Trumpets and Sun Baby know what they're doing.
It's already here. And likely you are deliberately trying to deny that. Here's a nice little rabbit hole for you: https://pastebin.com/raw/8TQyPKUF
the scary thing isn't the sci-fi in this stuff, it's that the tech already exists to do the polygraph scam again
You could skip a lot of stuff by training a transformer on some kind of neural embeddings. You probably get effective FTL communication and limited immortality that way.
How did you arrive at FTL? At best we'll have comms at radio transmission speed, right?
It's faster than light because your model is already there so no information has to be exchanged.
But isn't it the same sort of FTL and "immortality" that a creator gets from us listening to a downloaded podcast that they prerecorded?
To quote Woody Allen:
> “I don't want to achieve immortality through my work; I want to achieve immortality through not dying. I don't want to live on in the hearts of my countrymen; I want to live on in my apartment.”
To quote Woody Allen:
> “I don't want to achieve immortality through my work; I want to achieve immortality through not dying. I don't want to live on in the hearts of my countrymen; I want to live on in my apartment.”
Seems like a technicality tbh.
very cool concept - kind of like an offline inference model of your conscious. You could have periodic (over the wire) sync updates to your real self
Cant wait for a man-choker that executes "I'm having costlier rice, check if glucose spike is lower than usual rice". Yes, both devices are outside my body.
I don’t understand what this means. Did you use a translate app?
I think it's perfectly legible?
man-choker would just be a choker (you know, the accessory usually for women) with some tech on it, in this case to accept a command to check with another bio-device if the glucose spike of more expensive rice is better or worse than cheap rice.
man-choker would just be a choker (you know, the accessory usually for women) with some tech on it, in this case to accept a command to check with another bio-device if the glucose spike of more expensive rice is better or worse than cheap rice.
it's legible but a very confusing sentence
man-choker is not a word, and choker is a niche garment, why not "necklace" or just "wearable"?
Then it "executes" a question?
and then there's a reference to "both devices" - what devices?
You need to put together a lot of context clues and assumptions to get to: They are probably a diabetic with a glucose monitor and pump, and they want a smart device to analyse the data with natural language (but again, why a choker specifically? Wouldn't a smart watch or something make more sense?)
man-choker is not a word, and choker is a niche garment, why not "necklace" or just "wearable"?
Then it "executes" a question?
and then there's a reference to "both devices" - what devices?
You need to put together a lot of context clues and assumptions to get to: They are probably a diabetic with a glucose monitor and pump, and they want a smart device to analyse the data with natural language (but again, why a choker specifically? Wouldn't a smart watch or something make more sense?)
Yeah, I find it weirder for someone to claim that it's perfectly legible.
The CGM example was so out of place that I thought they meant to post in yesterday's CGM thread. And I spent a moment wondering why "costlier rice" was qualified. Maybe they are hoping their more expensive yuppy rice is absorbed more slowly?
And man-choker sounds like a word that Frank Herbert made up for Dune.
The sibling reply to mine is getting lost in the weeds: you can admit that something is written in a confusing way while also being able to understand what they meant or even appreciating it. And telling someone to "Get over it" sounds like you aren't tracking what the convo is about.
The CGM example was so out of place that I thought they meant to post in yesterday's CGM thread. And I spent a moment wondering why "costlier rice" was qualified. Maybe they are hoping their more expensive yuppy rice is absorbed more slowly?
And man-choker sounds like a word that Frank Herbert made up for Dune.
The sibling reply to mine is getting lost in the weeds: you can admit that something is written in a confusing way while also being able to understand what they meant or even appreciating it. And telling someone to "Get over it" sounds like you aren't tracking what the convo is about.
Apparently GP prefers to be more playful with their language.
> or just "wearable"?
wearable where? A choker is not some niche garment. It is a well understood accessory that maybe not everyone wears, but it's not because it is not well known. Describing it as a man-choker while a made up word, get over it that's how language evolves, is very descriptive. Calling it a choker means that it will be around the neck. Wearable could be a watch, and that's not what was meant. Using the word choker explicitly tells you where it was proposed to be worn.
> and then there's a reference to "both devices" - what devices?
one is the speech device, the other the glucose monitor. separate devices, but worn on the choker which is now becoming a tool belt. But I'm guessing you'd have a problem if it was described as a tool/utility belt worn around the neck like a choker???
wearable where? A choker is not some niche garment. It is a well understood accessory that maybe not everyone wears, but it's not because it is not well known. Describing it as a man-choker while a made up word, get over it that's how language evolves, is very descriptive. Calling it a choker means that it will be around the neck. Wearable could be a watch, and that's not what was meant. Using the word choker explicitly tells you where it was proposed to be worn.
> and then there's a reference to "both devices" - what devices?
one is the speech device, the other the glucose monitor. separate devices, but worn on the choker which is now becoming a tool belt. But I'm guessing you'd have a problem if it was described as a tool/utility belt worn around the neck like a choker???
Massachusetts' New England merch only names Irish citites. Is also legible but equally incoherent.