Why are they talking about speech synthesis that isn't demonstrable yet?
Surely this is ALREADY a super-advanced hearing aid?
Scientists have reconstructed the words people hear by using a computer algorithm to decode electrical signals in the brain. Eggheads at the University of California in Berkeley fastened electrodes onto the bare brains of patients undergoing neurosurgery. The scientists were studying the superior temporal gyrus (ridge on the …
The sounds would have to come through working ears to get to the brain to be read. It wouldn't work on deaf folks.
If it does turn out that the same bits of the head meats are used for processing incoming speech and for internal dialogue then this will be pretty cool and incredibly scary*, otherwise it's just a complicated and error prone microphone....
* A polygraph that actually prints out your thoughts? What could possibly go wrong?
I wonder why the researchers didn't replay the electrical signals through the electrodes and ask the patients if they 'heard' the word... If that happened it would mean a hearing aid is closer than a speech synth given that they haven't located a brain circuit carrying imagined words. Also, imagining words is probably more difficult to detect than if a mute patient imagines trying to speak, in which case more neurons would surely fire...
I heard this story on the Radio 4 news this morning. You have to hear the audio recordings, where the computer reconstructs the words the 'patient' is hearing, from the brainwave patterns made by them hearing it, to fully appreciate how amazing this breakthrough appears to be. It sounds like the alien robot from all your favourite childhood sci-fi films speaking; distorted, modulating all over the place, but still eerily recogniseable.
When I heard that report this morning, I had that feeling you get occasionally, when a new piece of technology is revealed and leaves you feeling like you've witnessed the dawn of a new era. My mind boggles at the thought of where this stuff could be in a couple of decades, if it keeps advancing.
As I listened to the news article this morning, my initial "wow!" inevitably gave way to pondering how such 'mind-reading' technology would eventually be abused, once it equally inevitably falls into the hands of those lovely types, such as DARPA and GCHQ.
Even as I thus pondered, John "Humpty-Dumpty brain" Humphries adopted his best "but..." voice, leaving me thinking for a few seconds that he was actually going to ask an intelligent probing question along these lines, only to have hime revert to type, by enquiring what might happen if a "Dr. Strangelove villain" got hold of the technology. [Hence the "Dr. Strangelove" quote in the original article].
Great potential for good but will, no doubt be used for evil. T'was ever thus!
Speech reception, interpretation/comprehension and production are located in structurally and functionally distinct domains of the brain.
After strokes or other brain injury, you can commonly find people who can't understand speech, but can speak reasonably well and vice versa. Deaf people can often learn to speak etc.
These brain regions are interconnected. However there is no reason to assume that your brain assembles what you either 'think' or about to say into the same encoding as what you are listening to.
I've wondered for a long time why or rather when people would stop trying to stimulate the *sensors* of the brain and instead try to encode/decode the actual signals in the nerves.
Hopefully this is building a library to allow *arbitrary* speech to be interpreted without knowing what it is ahead of time.
Your words gave me a thought, how about they use a similar test to determine the type of "networking" used across the human nervous system. It would be interesting to know if the brain routes specific instructions to specific nerve clusters, or whether information like this is broadcast across the entire "network" with some sort of encoding to be recognised by the relevant target.
One wonders just how close our ideas of computer networking is to the inner workings of the brain.
"One wonders just how close our ideas of computer networking is to the inner workings of the brain."
A cursory study of brain structure versus versus computer structure will show they are a very *long* way apart.
As for peripherals AFAIK most things are direct wired a part of the brain. It's point to point. There is no *bus* structure. It's just the wiring is *very* fine (and under the right circumstances i fthe ends are sown together will re-connect. Not something that happens with most cut cable).
We don't actually know that. At least, not yet. Obviously the next phase of the research is to pick up imagined speech. The stage after that is to see whether asking someone a question prompts them to imagine enough of the right answer to detect when the words they actually choose to utter are different.
And I wouldn't worry too much about the neuro-surgery, either. There are *other* research groups looking at harvesting useful information from the interior of the brain with non-invasive detection.
They too are claiming that it might be a big step forward for people with disabilities. They are correct in this and the research is worth doing. However, society *ought* to bear in mind that such powerful techniques *will* be mis-used. At some point in the next century, science may deliver tools that the Inquisition of old would have killed for.
Not to put a downer on the conspiracy theorists , but while there is obvious potential for those people who know they cannot communicate any other way to "coherently vocalise" their internal thoughts in a way that may be detected and translated.
Is this how people really analyse their options and make plans ? It's not what I do .. given the choice betwwen chocolate or vanilla it's just a mashup of experience/availability/nothing in particular that guides my decision. Can't imagine what this machine would make of that. My wife is always telling she can't work out what I'm thinking and that's probably because most of the time I'm not , personal decisions are a mix of instinct and experience not sub-vocalised analysis.
What's the next step, eh?
Reading minds? You mean there'll be no future excuses for waterboarding at Guantanamo anymore? Hum...
...Stop instantly. Can't have system where torturers can no longer concoct excuses.
Imagine the chaos. Out of work and back to pulling wings off butterflies, methinks!
Isn't it adorable how people who research this sort of stuff don't instantly think of the masses of harmful and plain nasty applications that the military classes of every nation will put it to but instead think only of the .01% of the population in rich countries it might help? I suppose it helps them sleep at night.
You may have a point, but I don't think the guy who invented the wheel (I think it was Ugg Urrrr or something like that) really envisioned the M1 tank either. Pretty much everything invented or discovered can be used for evil as well as for good. It's up to society to decide how it's put to use.
The electrodes were not 'attached' to the patient's brain they were merely placed upon it. It is just about impossible to attach an electrode like that to the surface of the brain without damaging the very structure you are trying to record from.
BTW this means even if the govt forces you to have your head opened and electrodes placed on your brain you could frustrate it by thinking nefarious thoughts while flinging your head around violently (which will cause the unattached electrodes to move around). Or just pretend to sneeze will do it. Who would have thought possession of snuff would become a subversive act?
If you are worried about less invasive methods, there are reasons why they didn't do this years ago with an EEG net placed on the head. Think insulation, signal to noise ratio and lack of spacial resolution all combined. Not coming to a black helicopter near you soon.
Anyway, haven't all you paranoid people had your brain enveloped in tinfoil already? It's so much more convenient, and less ridicule inducing, than an old fashioned external foil beanie.