Re:Reading minds with deep learning
by carapace on Friday November 30, 2018 @02:21PM
(#18571802)
@dmreedy Thank you. I really appreciate that.
@AstralStorm As you point out, GSR (or more generally EMG) isn't necessarily the best modality for this. But my point is that, whatever the comm channel's nature and characteristics, once you have connected a brain to a computer, it's useful to recognize that the more powerful and easier to program signal processor is the brain.
@h0p3 First, thanks. I am deeply weird and to me it's a compliment. ;-) Second, OMG your website is a custom TiddlyWiki? It looks fascinating. Email sent. Well met.
@trevyn I have not actually done this, no. To me it's trivial, and I don't want nor need to talk to a computer that intimately. (I'm one of those people who, despite using a computer daily for years, still "hunt and pecks". I actually can kinda touch-type, but my ratio of thinking to typing is such that there's not a great incentive to do better. Typing speed has never been the limiting factor in my life, so... Lazy.)
And with modern NN techniques and an HD camera you could just point the camera at your face and calibrate facial expressions, pulse rate, etc...
@mrfusion It just makes it much easier to self-train and calibrate. In crude terms, the conscious mind has a lot of belief structures and other homeostatic things going on that interfere with communicating with the motor cortex. In a trance most of that stuff is temporarily offline and you can "go in" and "ask" parts of your brain (presumable we all agree that the brain is where this stuff is happening, yes?) to do things like "work with the NN to calibrate unambiguous signals" etc..., and it will typically "go along with" this sort of hypnotic suggestion.
The thing is, it's not like this is something you could franchise and open stores in malls or something. There's no mass market for hypnosis, techno-assisted or otherwise.
@AstralStorm As you point out, GSR (or more generally EMG) isn't necessarily the best modality for this. But my point is that, whatever the comm channel's nature and characteristics, once you have connected a brain to a computer, it's useful to recognize that the more powerful and easier to program signal processor is the brain.
@h0p3 First, thanks. I am deeply weird and to me it's a compliment. ;-) Second, OMG your website is a custom TiddlyWiki? It looks fascinating. Email sent. Well met.
@trevyn I have not actually done this, no. To me it's trivial, and I don't want nor need to talk to a computer that intimately. (I'm one of those people who, despite using a computer daily for years, still "hunt and pecks". I actually can kinda touch-type, but my ratio of thinking to typing is such that there's not a great incentive to do better. Typing speed has never been the limiting factor in my life, so... Lazy.)
I expect you would be able to input text as fast as you could think it. You're not limited to eight bits. I would also explore methods like, e.g. "Dasher" http://www.inference.org.uk/dasher/ https://en.wikipedia.org/wiki/Dasher_(software)
And with modern NN techniques and an HD camera you could just point the camera at your face and calibrate facial expressions, pulse rate, etc...
@mrfusion It just makes it much easier to self-train and calibrate. In crude terms, the conscious mind has a lot of belief structures and other homeostatic things going on that interfere with communicating with the motor cortex. In a trance most of that stuff is temporarily offline and you can "go in" and "ask" parts of your brain (presumable we all agree that the brain is where this stuff is happening, yes?) to do things like "work with the NN to calibrate unambiguous signals" etc..., and it will typically "go along with" this sort of hypnotic suggestion.
The thing is, it's not like this is something you could franchise and open stores in malls or something. There's no mass market for hypnosis, techno-assisted or otherwise.
@kakarot Sure. Email sent.