TED suggestions

[Mesozoic Mister Nigel]

What you're saying is a lot like "If I can take really, really good detailed video of the insides of a computer while it's working, I'll be able to reproduce the data on the computer".

Yes. You do realise that what you just said there would actually be possible right now if anyone could be bothered to do it?

Uhhhh nope. Unless you are using a completely different definition of "video".

No matter how high the resolution achieved with fMRI, it is not and will never be capable of taking an image of the brain that a personality and memories could be reconstructed from. I actually do think that at some point a technology may be developed that CAN do that, but it's not fMRI.

Sorry, yeah, getting ahead of myself  I wasn't thinking of video, I was thinking of some kind of electron microscope Of course, fMRI will only give part of the picture but I agree, something will eventually be developed, there's loads of theories flying about, about what form this new tech will be, whether it's some kind of direct, nanotech, neuron by neuron bridge or maybe an amalgamation of different scanners or maybe some other method that hasn't even been invented yet.

What the clip I posted shows is a milestone on that roadmap. I think a fairly significant one. What's significant here, though, isn't the scanner. It's the software that analyses the data. What it's doing, right now is effectively simulating part of the function of the visual cortex, without simulating the visual cortex itself. It's applying an algorithm that creates the same results from given input. It appears that a large part of that algorithm was not written by a human. The machine "learned" how to do it, humans merely gave it the parameters that told it how to learn.

Literally all it is doing is going "there is increased oxygen consumption in these places when person looks at A". Therefore when there is increased bloodflow in this pattern, display something that looks like A". The more times subjects look at A, the better of an average for A can be reproduced.

It's easy to get super excited about material presented in TED talks because it is often presented as if it is new and groundbreaking and exciting, because that is the purpose of TED talks; to get the general public excited about science. Unfortunately, in most cases this impression is going to be misleading, because scientists with new, groundbreaking, and exciting research and ideas are not presenting it at TED, they are presenting at it national and international science conferences, and it may eventually make its way to TED. In addition, in the last couple of years not only has TED been sadly diluted past all recognition with the utter crap that is TEDx, but they have taken to inviting guest speakers to present on research that is wildly outside of the scope of their own field of knowledge, which has resulted in the circulation of a good deal of speculative pseudoscience being presented as if it were spoken by an expert in the field.

I used to love TED. Not so much anymore.

[P3nT4gR4m]

Literally all it is doing is going "there is increased oxygen consumption in these places when person looks at A". Therefore when there is increased bloodflow in this pattern, display something that looks like A". The more times subjects look at A, the better of an average for A can be reproduced.

I think you might be getting the wrong end of the stick here. Here's the guy who actually did the experiments explaining it a bit better

The subject was not looking at an image over and over, rather a whole bunch of completely different videos were shown (tens of thousands of images), allowing the machine to spot which neurons were responding to which videos. and cross tabulate this with which pixels the video frames had in common, allowing it to build a picture of which neurons respond to which areas of the image. The second part of the experiment then showed new videos to the subjects which the machine interpreted according to the ruleset it established in the first part.

[Mesozoic Mister Nigel]

I didn't say they looked at the same image over and over again, and you just restated what I said.

But if you're that convinced that I just don't get it, go right on ahead and believe that. 

[P3nT4gR4m]

I didn't say they looked at the same image over and over again, and you just restated what I said.

But if you're that convinced that I just don't get it, go right on ahead and believe that. 

I'm convinced you're not seeing the same thing I am. You start out with "Literally all it is doing"

Literally all it's doing is reading neural hardware activity and decoding it, producing a reproduction (albeit a very rudimentary one) of the input signal. The impression I'm getting is like if I was showing you an old mainframe room 20 or 30 years ago and you were like "Yes but all it's doing is adding and subtracting numbers" and at the same time, I'm seeing how it's going to end up with youtube on a cellphone.

[Mesozoic Mister Nigel]

I didn't say they looked at the same image over and over again, and you just restated what I said.

But if you're that convinced that I just don't get it, go right on ahead and believe that. 

I'm convinced you're not seeing the same thing I am. You start out with "Literally all it is doing"

Literally all it's doing is reading neural hardware activity and decoding it, producing a reproduction (albeit a very rudimentary one) of the input signal. The impression I'm getting is like if I was showing you an old mainframe room 20 or 30 years ago and you were like "Yes but all it's doing is adding and subtracting numbers" and at the same time, I'm seeing how it's going to end up with youtube on a cellphone.

I can understand where you're coming from, but maybe it would help to understand that I start my morning, every single day, catching up on the latest neuroscience news. fMRI is one of the most overhyped, if not THE most overhyped, technology in neuroscience. So what I see is a guy who is tripping out and announcing that flying cars are next, because he just saw a combustion engine for the first time.

So maybe you're the great visionary in this scenario, but I still think you're attributing potential to this particular technology that just isn't there.

[P3nT4gR4m]

I didn't say they looked at the same image over and over again, and you just restated what I said.

But if you're that convinced that I just don't get it, go right on ahead and believe that. 

I'm convinced you're not seeing the same thing I am. You start out with "Literally all it is doing"

Literally all it's doing is reading neural hardware activity and decoding it, producing a reproduction (albeit a very rudimentary one) of the input signal. The impression I'm getting is like if I was showing you an old mainframe room 20 or 30 years ago and you were like "Yes but all it's doing is adding and subtracting numbers" and at the same time, I'm seeing how it's going to end up with youtube on a cellphone.

I can understand where you're coming from, but maybe it would help to understand that I start my morning, every single day, catching up on the latest neuroscience news. fMRI is one of the most overhyped, if not THE most overhyped, technology in neuroscience. So what I see is a guy who is tripping out and announcing that flying cars are next, because he just saw a combustion engine for the first time.

So maybe you're the great visionary in this scenario, but I still think you're attributing potential to this particular technology that just isn't there.

Cool! I think I can lay this to rest by telling you that fMRI means absolutely nothing to me. Just like the old valves in the original computers didn't mean anything and silicon means nothing, other than what they can accomplish. The best predictions on the planet don't have fMRI reaching anything like the kind the of scanning resolution required to capture all the atoms and state vectors required to grab a complete snapshot but technological advancement is always a series of incremental steps, punctuated by quantum leaps. I see this experiment as one of those quantum leaps, right out of leftfield. Whilst tons of approaches are plodding along at the rate of Moore's law, trying to simulate nematode worms and the like, growing neurons in petri dishes and  fuckton of other shit that you no doubt are more switched onto than me, this maniac has just gone, "fuck it, lets forget about understanding the shit and just take pictures of it and let a supercomputer figure out how to map the connections."

And it worked!

As I said when I posted the clip - what excites me isn't the latest fancy kind of microscope du jour - it's the big-data approach to the problem. That's the one part of this equation that does stray into my area of professional expertise. I've been following neuroscience progress, as an interested layman, for a few years and everything I've come across seems to involve learning how it all works from the ground up and, again as a layman, hooked into TED talks and with no stomach for dry scientific papers, it's been kind of plod plod plod, with the odd mildly interesting goalpost ticked off along the way.

Now, with this experiment, the way is paved to just grab the next awesome scanner and hook it into the next awesome gigacore processor. Wash, rinse, repeat. Neuroscience just got scalable. Let's face it, that's when you know your science is all grown up - when you develop it to the point where you hand it over to us engineers

Seriously, tho, I'm not dissing whatever the hell else is going on in the field and I've no doubt that something else will come along that either complements this approach or renders it obsolete but, until then, I can see a roadmap from here to the goal of saving my backup, which is, honestly, about my only interest in this whole field. Obviously if you guys can cure my brain cancer, or alzheimers or senility in the meantime, that'd also be cool but my eye is on the prize.

[Mesozoic Mister Nigel]

Except it's totally not out of left field, at all. I think that not knowing what fMRI is probably heavily contributes to your interpretation of what's actually being done. It's not new; this is stuff that neuroscientists all over the world have been working on for years, improving incrementally. Did you hear what her "big idea" was? Where she talks about the bottleneck in this technology, and what she thinks the solution is? The whole point of her talk is that she thinks the increase in fMRI resolution can be done with a better type of magnet. That's it, that's the whole spiel. The resolution limitation isn't in the computing, it's in the level of detail in the brain scans. She has an idea for fancier magnets.

What this technology does is makes a picture by reading activity in the brain. If you can't read the activity with a high level of precision, you can't make a very clear image, no matter how great your computational processing. Current fMRI technology doesn't read activity with a high enough level of precision to allow computers to reproduce very clear images. Either we need a new kind of brain imaging technology, or we need, as she says, better magnets.

However, most of her talk is basically a visionary Google display developer talking about other people's brain research, and toward the end she just starts wandering all over the place. Yeah, a brain scan device that lets us see brain activity at a neuronal level would be basically the holy grail of neuroscience, sure. Kind of a non-sequitur, but totally true.

Of course, as this research progresses it is pretty inevitable that both brain imaging and computing is going to continue to improve; we're already looking forward to some much more powerful fMRI equipment (I don't know if she just wasn't aware of this when she gave her talk, or if she didn't talk about it due to lack of time): http://spectrum.ieee.org/biomedical/imaging/the-worlds-most-powerful-mri-takes-shape.

Even so, at the end of the day the kind of technology she's talking about is purely a way to convey visual imagery. It's not a way to snapshot a personality. It's really cool, but not personality-reconstruction or memory-storage cool. It could, if it reached a certain point, record a playback of a memory, but it's still kind of the equivalent of pointing a video camera at a TV set, and not the equivalent of burning a CD, if that makes sense. 

Some of the other current sensory neuroscience research really interests me more, like this: http://nsnbc.me/2014/03/07/blind-people-can-see-bodies-sound-study/

To each her own, though. Bringing pictures out of the brain just don't excite me the way putting pictures into it do, although long-term, these technologies are probably complementary.

[Mesozoic Mister Nigel]

This line of research goes back further, but here are some articles documenting the progress.

http://www.scientificamerican.com/article/translating-images-from-brain-waves/

http://www.livescience.com/16190-movies-reconstructed-brain-activity.html

http://abcnews.go.com/Health/MindMoodNews/scientists-youtube-videos-mind/story?id=14573442

Gallant and his team are slowly and steadily chipping away at the visual translation, and other teams are working on words and sounds. Eventually it will all come together, but it's no renegade out-of-the-blue thing, it's an ongoing, long-term, and highly collaborative process.

As far as I know, nobody's doing smell. Maybe I should email Luca Turin about that.

[Mesozoic Mister Nigel]

PS.

Welcome to the future.

[Mesozoic Mister Nigel]

I would actually be more interested in technology that let us look at brain activity from a glial level, for a number of reasons, though.

[P3nT4gR4m]

This line of research goes back further, but here are some articles documenting the progress.

http://www.scientificamerican.com/article/translating-images-from-brain-waves/

http://www.livescience.com/16190-movies-reconstructed-brain-activity.html

http://abcnews.go.com/Health/MindMoodNews/scientists-youtube-videos-mind/story?id=14573442

Gallant and his team are slowly and steadily chipping away at the visual translation, and other teams are working on words and sounds. Eventually it will all come together, but it's no renegade out-of-the-blue thing, it's an ongoing, long-term, and highly collaborative process.

As far as I know, nobody's doing smell. Maybe I should email Luca Turin about that.

"We need really big computers," Gallant said."

So this particular experiment is a couple of years old. His computer is a lot bigger now. Anything interesting happened since?

[Mesozoic Mister Nigel]

This line of research goes back further, but here are some articles documenting the progress.

http://www.scientificamerican.com/article/translating-images-from-brain-waves/

http://www.livescience.com/16190-movies-reconstructed-brain-activity.html

http://abcnews.go.com/Health/MindMoodNews/scientists-youtube-videos-mind/story?id=14573442

Gallant and his team are slowly and steadily chipping away at the visual translation, and other teams are working on words and sounds. Eventually it will all come together, but it's no renegade out-of-the-blue thing, it's an ongoing, long-term, and highly collaborative process.

As far as I know, nobody's doing smell. Maybe I should email Luca Turin about that.

"We need really big computers," Gallant said."

So this particular experiment is a couple of years old. His computer is a lot bigger now. Anything interesting happened since?

"this experiment"



I give up. You're right, it's basically computer wizardry, old guy.

[P3nT4gR4m]

Ever wonder what the most evil billionaire on the planet is up to these days?

[POFP]

Andrew Stanton: The Clues to a Great Story

If you're interested in writing, story development, etc.

[JamesStrangefellow]

I saw Sam Jones (Flash Gordon) doing a Q & A the other day.

Ironically, he was in "Ted" the movie.

He should do this speaking Ted stuff.

The man has charisma, presence, some great stories, endless character and seemingly a real good heart.

It would also be somewhat discordian to have him there, but I think there would be much beauty in that.