The examples they gave are EEG readers which have terrible resolution. There will need to be muchhigher resolution devices that can read more neural activity before we have anything close to mainstream BMI devices.
Part of me wonders if one could use bio-nano-sensors[1] to amplify the electrochemical signals from the brain(or body) so they can be picked up easier or transformed into some other kind of language to be processed better/more efficiently (through an EEG reader or receiver of some kind).
Wish I hadn't ran out of time for this ... https://vimeo.com/mindhead. You can easily get to "on/off" with equipment like this. The rest depends on training protocols and the software design needed to implement them.
This is pretty cool, though probably a bit obtrusive for consumer applications. Also it doesn't look like the signal strength is that strong or varying (from what i can see in the video) which would make it pretty hard to try to fit equations that would match a specific actions induced from the brain.
I'm still skeptical of BCIs being useful, to healthy people, for, at least, another twenty years. I can't think of a single thing I could control with my brain, that I couldn't do more accurately, and consistently, with my hands, voice, etc. But, I cannot tell if my perception is stunted by my current life experience, the quality of existing sensing technologies, or an actual lack of application.
My approach is not so much your interaction to control a machine, more so how machines interact with your brain and body to guide your decision making. So think about BCIs and more broadly human machine interfaces for things like persistent heuristic mapping for memory augmentation (think along the lines of "tagging" events to neural activity - this might help with focused tDCS during recall) and in-stream blood toxin monitoring.
One simple example where a good BCI could beat hands and voice every time: window switching. No more task bar, no more start menu, no more alt-tab or window previews or virtual desktops. Just think "firefox" and you get firefox.
If you want an even more hyperbolic what-if, to give you an idea of what we hope is so much as possible, a really could BCI could do away with the concept of windows entirely. Your eyes can only see one thing at a time and the computer can move information to your eyes faster than your eyes can move information to the computer, so just present the user with information as they think about wanting to observe it.
I'm highly skeptical. I don't think "I am going to move my fingers around to press and hold the command button and press the tab button one or more times until I get to firefox". My brain already autopilots desires into physical output.
EEG is a joke, and it's never explained how exactly BCI is going to work. If I switch to Firefox when I think it, then I would have annoyingly switched over multiple times when writing this comment. If it's something more complicated then that, then what is the generic metaphor my mind learns to app-switch that's simpler than the current one my brain performs using my fingers?
I believe BCI absolutely can and will be the future of computing interfaces, but there's no way that simply "mapping" the brain is going to let us turn desires into actions more efficiently via digital telepathy than our existing high-res interfaces with our bodies - there's a lot of groundwork that still needs to be covered.
To put it another way, if you could build software to accurately describe "what information a user wants to observe", you'd likely win a combo Nobel prize in neuroscience, philosophy and AI.
I'm not disagreeing with that. As I said, "even more hyperbolic what-if". Call it the same kind of science fiction that people were writing about smartphones in the 60's: a kind of general direction that far overshoots reality in many ways but which is hopefully not entirely wrong.
As for intentionality, your brain already handles that for you. I mean, I'm almost certain that you can imagine walking out the front door of your home without actually doing it. Same thing applies.
I don't know the exact details of the mechanisms, but I have an overview. One technique is to have the user associate some rare signal, like thinking the word "chair" repeatedly, with an action, and let the user learn to use that signal quickly in the same way that you've learned to use command-tab. This is mostly done by tracking changes in blood oxygenation using infrared sensors or fMRI. Another technique is to tap into the user's motor cortex completely blindly, show the user the output, and let neuroplasticity[1] take care of the rest; this works astonishingly well, with monkeys being able to "quickly learn to voluntarily control the firing rates of individual and multiple neurons" and directly control robotic arms [2].
In the academic field 'how exactly' a BCI is going to work is very well defined and there are only a limited number of basic types.
It's the representation of BCI in the media which is a joke. The actual science behind EEG based BCI is not that complex that the average person couldn't understand them. The disconnect between reality and peoples conception is similar to the way AI is represented, compared to what it is.
My dabblings with BCI were nothing like that. It was more like spend 10 minutes ensuring you're properly grounded, then try really really hard to relax, and then discover that the slight movement of your leg muscle is swamping the signal.
It was very hit and miss. I'm sure that given enough time and effort I could learn to control it more or less reliably, but it's a far cry from 'think firefox and get firefox'.
In fact, what I found interesting was that more physical measures, like blinking, or even the x coordinate of your eye gaze were picked up fairly well. I could imagine using those as triggers for something with the same equipment, but I doubt I would get as quick with a BCI as am I with a keyboard until they improve markedly.
When that day comes it's going to be awesome. I'm going to walk up to various people in the office and say something like "Hey, I was at the mall yesterday and midget donkey porn I thought I saw you".
> A direct neural connection could be more like 5-15ms [2]
That's not what EEG is. It's vastly more disappointing.
> The best physical reaction time is about 160ms [1]
Yes, but you (probably) have 10 fingers, and you can get the gaps between keypresses down to <100ms. That really helps with typing.
Suppose you type 200 wpm, average word length 5 characters, maybe 980-1000 characters/minute or about 15-18 characters/second. When typed at a steady pace, you can reliably do 63ms between keystrokes. This isn't reaction time.
With EEG, you get like 1 character every 4,800 milliseconds. (Although I admit I don't know if this is using dasher, or some other typing scheme, or what. Do you get just a binary character, or any UTF-8 character in those thousands of milliseconds?)
People with RSI would probably be happy to switch from keyboard and mouse to a BCI.
"I can't think of a single thing I could control with my brain, that I couldn't do more accurately, and consistently, with my hands, voice, etc."
Your brain is controlling your hands, why would being able to control a mouse cursor directly be less accurate or consistent? You're removing two layers of indirection (your hands and the mouse).
I've always imagined it being a link to other brains or computing devices. The new computing devices become part of who we are, it would give us the advantages of a computer inside our heads. Instant calculations and true photographic memory would be possible.
Plus it would eventually allow us to "transfer" our brains, which might be required to live forever. There has been lots of discussion on the requirements for a brain transfer, it would likely have to be gradual, and there is always the debate on if the new you is really you or just a clone.
How much of Brain Computer Interfaces is biology and how much of it is computing on a rough scale? Does anyone know of any good books to gain more knowledge in this field? I study biology as an undergrad and this sounds very interesting.
It's definitely more computing than biology, but it does depend slightly on the type of signal acquisition device. You have to know more about the underlying biological processes of the brain if you are recording from an implanted microelectrode array rather an an EEG cap.
Reading doctoral theses on this subject is also a great way to learn more. There are at least a few online that do a great job of explaining BCIs from the ground up.
Whole article behind paywall, and also above the fold of the business section of the Sunday April 28th SF Chronicle. Good times.
Also, as the maintainers of some reverse engineered EEG drivers, I'm in agreement with pretty much every other pessimist. They're toys. Fun toys, but toys. We're waaaaaaaay off from the cool control shit, especially at a mass marketable consumer level, but believe me, there's gonna be tons of cool research shit on the way there.
22 comments
[ 2.9 ms ] story [ 54.6 ms ] thread[1]https://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=...
If you want an even more hyperbolic what-if, to give you an idea of what we hope is so much as possible, a really could BCI could do away with the concept of windows entirely. Your eyes can only see one thing at a time and the computer can move information to your eyes faster than your eyes can move information to the computer, so just present the user with information as they think about wanting to observe it.
EEG is a joke, and it's never explained how exactly BCI is going to work. If I switch to Firefox when I think it, then I would have annoyingly switched over multiple times when writing this comment. If it's something more complicated then that, then what is the generic metaphor my mind learns to app-switch that's simpler than the current one my brain performs using my fingers?
I believe BCI absolutely can and will be the future of computing interfaces, but there's no way that simply "mapping" the brain is going to let us turn desires into actions more efficiently via digital telepathy than our existing high-res interfaces with our bodies - there's a lot of groundwork that still needs to be covered.
To put it another way, if you could build software to accurately describe "what information a user wants to observe", you'd likely win a combo Nobel prize in neuroscience, philosophy and AI.
As for intentionality, your brain already handles that for you. I mean, I'm almost certain that you can imagine walking out the front door of your home without actually doing it. Same thing applies.
I don't know the exact details of the mechanisms, but I have an overview. One technique is to have the user associate some rare signal, like thinking the word "chair" repeatedly, with an action, and let the user learn to use that signal quickly in the same way that you've learned to use command-tab. This is mostly done by tracking changes in blood oxygenation using infrared sensors or fMRI. Another technique is to tap into the user's motor cortex completely blindly, show the user the output, and let neuroplasticity[1] take care of the rest; this works astonishingly well, with monkeys being able to "quickly learn to voluntarily control the firing rates of individual and multiple neurons" and directly control robotic arms [2].
1: http://en.wikipedia.org/wiki/Neuroplasticity 2: http://en.wikipedia.org/wiki/Brain–computer_interface#Early_...
It's the representation of BCI in the media which is a joke. The actual science behind EEG based BCI is not that complex that the average person couldn't understand them. The disconnect between reality and peoples conception is similar to the way AI is represented, compared to what it is.
My dabblings with BCI were nothing like that. It was more like spend 10 minutes ensuring you're properly grounded, then try really really hard to relax, and then discover that the slight movement of your leg muscle is swamping the signal.
It was very hit and miss. I'm sure that given enough time and effort I could learn to control it more or less reliably, but it's a far cry from 'think firefox and get firefox'.
In fact, what I found interesting was that more physical measures, like blinking, or even the x coordinate of your eye gaze were picked up fairly well. I could imagine using those as triggers for something with the same equipment, but I doubt I would get as quick with a BCI as am I with a keyboard until they improve markedly.
A direct neural connection could be more like 5-15ms [2]
An order of magnitude difference in response time has huge ramifications for all sorts of applications.
[1] http://en.wikipedia.org/wiki/Reaction_time [2] http://en.wikipedia.org/wiki/Neuron
That's not what EEG is. It's vastly more disappointing.
> The best physical reaction time is about 160ms [1]
Yes, but you (probably) have 10 fingers, and you can get the gaps between keypresses down to <100ms. That really helps with typing.
Suppose you type 200 wpm, average word length 5 characters, maybe 980-1000 characters/minute or about 15-18 characters/second. When typed at a steady pace, you can reliably do 63ms between keystrokes. This isn't reaction time.
With EEG, you get like 1 character every 4,800 milliseconds. (Although I admit I don't know if this is using dasher, or some other typing scheme, or what. Do you get just a binary character, or any UTF-8 character in those thousands of milliseconds?)
"I can't think of a single thing I could control with my brain, that I couldn't do more accurately, and consistently, with my hands, voice, etc."
Your brain is controlling your hands, why would being able to control a mouse cursor directly be less accurate or consistent? You're removing two layers of indirection (your hands and the mouse).
Plus it would eventually allow us to "transfer" our brains, which might be required to live forever. There has been lots of discussion on the requirements for a brain transfer, it would likely have to be gradual, and there is always the debate on if the new you is really you or just a clone.
It was a biofeedback bowling game, I'm not even sure what decade it was released.
Here is a decent resource, but specific to one BCI research platform: http://www.amazon.com/Practical-Guide-Brain-Computer-Interfa...
Reading doctoral theses on this subject is also a great way to learn more. There are at least a few online that do a great job of explaining BCIs from the ground up.
http://blog.sfgate.com/techchron/2013/04/28/brain-hacking-re...
Whole article behind paywall, and also above the fold of the business section of the Sunday April 28th SF Chronicle. Good times.
Also, as the maintainers of some reverse engineered EEG drivers, I'm in agreement with pretty much every other pessimist. They're toys. Fun toys, but toys. We're waaaaaaaay off from the cool control shit, especially at a mass marketable consumer level, but believe me, there's gonna be tons of cool research shit on the way there.