[FoRK] Pick-on-Eugen-Day: Pocket-sized NMR

Jeff Bone jbone at place.org
Mon Jun 14 06:24:48 PDT 2010


On Jun 14, 2010, at 7:50 AM, Eugen Leitl wrote:

> On Mon, Jun 14, 2010 at 07:03:38AM -0500, Jeff Bone wrote:
> 
>> As I recall the assertion was about the tenability of getting ASCII  
>> streams out of the human brain with low-cost non-invasive gear anytime  
>> soon... no?
> 
> Useful ASCII streams with a practical setup, yes. The problem is
> scalp electrodes, which need fresh prepping and electrolyte gloop.
> Electrode pincushions or implanted electrodes (especially, picking 
> up potentials from the exposed brain surface is a signal bonanza). It might
> look like a small thing, but it's really the main issue blocking
> progress in noninvasive EEG.

Right, that was my recollection of the exchange.  But as demonstrated by the Tweet app, you don't even need expensive gear, the goo or direct contact for some amount of useful signal.  (For some definition of "useful";  and granted, the method there involved a user interface that is probably only really suited for the severely disabled --- not enough bandwidth to replace other general input methods for those who can use them.)

Have you seen the Emotiv demo they did w/ Marvin Minsky as the demonstration subject?  

  http://blog.makezine.com/archive/2010/04/tan_le_demos_mind-controller_on_mar.html

How about the guy that built a whole Rube Goldberg architecture (involving, among other things, Skype as its messaging backplane) to allow Emotiv control of a Rovio?

  http://www.robodance.com/mind-controlled-robot.php

Not that this is unique or new;  lots of folks have been doing very limited applications like this for a while:

  http://www.guardian.co.uk/science/2009/mar/31/mind-control-helmet-honda-asimo

  http://www.livescience.com/technology/061215_humanoid_robot.html

IMHO, my contention was and is that the limited bandwidth and expressivity necessary for a variety of useful control applications and for at least some specialized input facility is both tenable and available with today's technology, w/o the need for any substantial high-resolution EEG that would necessitate direct scalp contact and / or conductive gel;  enough proximity pickups and enough smart signal processing, coupled with perhaps some amount of training, should suffice.  

I'd say that's been amply demonstrated at this point, wouldn't you?  And the folks attempting to commercialize this stuff are going about it the right way:  aiming at games, first.  The interface is and likely will be insufficient to become the sole input method for any interestingly complex game, but it could easily become a must-have secondary control interface for certain tasks and actions within a variety of different kinds of games.  That should provide enough trickle revenue to push the technology close to the limits of its potential under the physical limits and constraints you're talking about.  (I guess the disagreement, really, is whether the technology can be useful and interesting and profitable under said constraints.  I think the signs point to yes at this time.  Far from any full even uni-directional BCI, but useful enough IMHO to be interesting.)

I've now got an Emotiv on backorder w/ SDK.  Previously I'd been holding off due to the Windows-only nature of their SDK;  just didn't want to futz with it.  But now interested in playing around with it.  I'm betting that w/in a couple of weeks of getting it, I'll be able to control at least the 5-6 lighting scene and whole-home macros presently defined in the home automation system via that interface.  Not that it'll be practical.  But mind-blowingly geeky-cool, IMHO.  Enough so to overcome my distaste for the SDK platform, anyway, having now seen a couple of inspiring demos. 

A Parrot AV drone, a static set of defined waypoints and beacons, an Emotiv, and some Vuzix would make for great entertainment, too... ;-)



jb




More information about the FoRK mailing list