OT: Honey, I Cracked the Kurzweil!

[Tom Williams]

Apparently it also causes Z-axis compression in the skull. Or maybe the squished skull causes ADH -- oh look, a squirrel!

[MurMan]

T, I'd like to respond with a single thought, but there's no central, coherent theme. Just attacks and negative comments ...

 

You must have pretty low self esteem there, considering the "baiting" and "low on respect" remarks, ...

The baiting I do is purely for entertainment value. Please feel free to ignore it.

 

The "respect" remark was deadly serious. How can I respect someone that claims to know science, yet doesn't know how to dig into a topic? You criticized the overview article because it didn't contain the proof needed to support the research claims. Guess what: Overview articles never do. They are usually written by a journalist that doesn't understand the science and simply serve to create interest.

 

Someone truly interested in the research would have seen the Abstract at the bottom of the article and the link to download the paper. You didn't do that, but continue to criticize even after I pointed out the link to the paper. This is why I don't respect you.

 

My background in science is sufficient to take wee at the presented paper in more than a few dimensions, but I won't, ...

No comment about the amount of science required for one to take wee at a paper.

 

But, the fact that you won't read it supports my earlier comment about running and hiding. Look, T. Most of us here have made statements that were either wrong or couldn't be supported. I know I have. When this happens, you have to face up to them and admit that you got it wrong. Admitting that you're wrong would build respect.

 

And that is the subject *you* seem to put forward, "isn't that terribly interesting"? Why? You fail to give a content-wise answer, just like the article is mathematically horrifying (and lacks the needed "proof" images).

You want to know why I find this topic interesting? It's because deriving our thoughts from an EEG has so much potential for good and bad effects on society. The fact that several others are posting about it here is proof that they also find it interesting. If you don't, then please move on.

[Theo Verelst]

You are not a decent person about the obvious mathematical remarks I made. If you're ok with scientific charlatanry and do not require the solid foundation of the hard sciences, that shame on you when you enter mathematical and EE fields I know well enough to ignore you indeed.

 

 

[MurMan]

... I know well enough to ignore you indeed.

Finally!

[donblanco]

The interesting part (to me) was that the brain uses different areas of neurons to calculate recognition. Some focus on head shape, others on eyebrows, eyes, nose and mouth. That's why an EEG can work to begin with. Scientists have determined minute differences in brain activity in very specific locations depending on whether you are looking at a person with blue eyes vs green, blonde hair vs dark, etc. I've seen several before/after photos and the fact that we can even come close to reproducing a face based on brain activity astounds me!

 

Yes, there's definitely some creepy possibilities that come to mind, mainly fueled by sci-fi movies. But I'm just impressed by the tech.

 

Now if they ever manage to do something similar using a *memory* of an image, now *that* would get pretty sticky quickly!

[Reezekeys]

This thread is gonna have me taking some wee very soon!

[Desert Rat]

But, don't you agree that AI, which wasn't available 40 years ago, might take us down a a damned exciting path?

 

The AI we have today isn't magic. It's just a good pattern matching technology that uses training instead of programming. These crude experiments shows that EEG signals contain enough information from the brain's visual processing to identify specific images. The amount of information and its encoding will take a lot of time and work to figure out. But we now know that visual information is in an EEG. This is huge. What other brain activity will we discover?

 

Then we've got the implantable brain sensor tech that's being developed that will provide much richer signals.

 

At some point, there will be practical applications to music.

 

 

Sorry for the delay in my response. I do agree with you, but after watching this video,

[video:youtube]https://youtu.be/d_LPb624goQ I get the creeps with AI technology as well.

[DulceLabs.com]

As long as Joe the Quadriplegic gets to choose when to transmit...

 

Thoughts and prayers for Joe. I had no idea.

[jazzpiano88]

There is no "reconstruction" going on here.

 

Theo is absolutely correct that this is a EEG Signal Matching/Correlation demonstration resulting from a blind ML optimizer that has nothing to do with reconstructing an actual image from an EEG signal.

 

Very frustrating that lay people draw "life influencing" conclusions about the technology based on "Popular Mechanics" type hype.

[MurMan]

Theo is absolutely correct that this is a EEG Signal Matching/Correlation demonstration resulting from a blind ML optimizer that has nothing to do with reconstructing an actual image from an EEG signal.

Are you saying that there is no image signal within the EEG? Or to put it another way, you refute the conclusion of this research?

 

To conclude, our investigation targets the neural dynamics of face processing as reflected by EEG patterns. Our findings shed new light on the time course of facial identity processing while providing a way to extract and to assess the underlying visual information. Last, from a methodological standpoint, our results establish the feasibility of EEG-based image reconstruction and, more generally, they confirm the rich informational content of spatiotemporal EEG patterns.

[Doerfler]

There is no "reconstruction" going on here.

 

Praise the Lord. I was confused for awhile.

[J. Dan]

I think maybe you're getting mixed up by the EEG image vs the "reconstructed" image. It is correlation and the image being built (the reconstructed image, not the EEG) is put together by data that is fed in and correlated to specific EEG image data.

 

I have a friend that has several Engineering degrees and has spent a great deal of his life doing research mainly around medical imaging - specifically stuff like this. I posted the link to the article on his facebook page and here were his comments:

 

There have been a few studies like this recently. The actual results are a lot less impressive that the media headlines. And they are made possible by a lot of additional information and constraints in the experimental design--for example, the computer program that reconstructs the face is "told" that you are looking at a face, and is "told" that there are a certain set of features and/or faces it can choose from (though I don't know details on the study the article refers too--just browsed it quickly). I'm not bad-mouthing the studies though. They usually learn some interests aspects of how the brain processing visual (face, in this case) information. But headlines stating that are much less exciting.

 

TLDR: Don't worry, that stuff doesn't really work and no computers will be reading our minds any time soon. Probably not in our lifetimes.

 

Unless we have 100s-1000s of electrodes implanted *inside* our skulls. IMO.

 

[MurMan]

I have a friend that has several Engineering degrees and has spent a great deal of his life doing research mainly around medical imaging - specifically stuff like this.

Could you ask your friend the same question: Are you saying that there is no image signal within the EEG?

[J. Dan]

That's what he just said in the quote above.

[Bif_]

 

 

Hershey may have a legal problem with the design.

[MurMan]

That's what he just said in the quote above.

Dan, your friend reaffirms my favorite law of physics: For every PhD, there's an equal and opposite PhD.

 

I have to admit that I don't know enough about this field to make a strong judgement. Until someone proves that there is no image signal in an EEG, I'll stay in the camp that believes that it exists even though the S/N may be too low for it to be of value.

 

[Threadslayer]

For every PhD, there's an equal and opposite PhD.

 

Love it. I'll have to remind myself not to use this line at work. I'm surrounded by them.

[jazzpiano88]

Are you saying that there is no image signal within the EEG? ...

Correct! At the risk of failing to convince you of Dan's source which you haven't refuted other than to blindly disparage all researchers...

 

If you read the article, they are producing images based upon features identified from the EEG that can be correlated to the trained input response. If the input were monkey images, they would be showing monkeys neutral or happy in their paper.

 

Do you really think they "reconstructed" those images shown at the end of the article based on a raw EEG? If so, why wouldn't they have produced new images unlike the trained images to demonstrate that the "reconstructed" image is contained in a EEG just like NTSC?

 

Notice, for example, how they didn't test the "reconstructed image" of a "Truck" (or any other arbitrary image) after training on Males Neutral and Happy :)

 

I only crawl out of the woodwork in my area of expertise because you chose to bait and bully Theo in the position of ignorance. I think you are the first person at KC to elicit even a mild negative response from Theo in response to a personal attack based upon his technical musings.

 

As an aside..

This is one of the big issues with AI that even the chief evil-doer Eric Schmidt admits. It is a correlative technology that will never take the place of human judgement. The error rate is way too high until some other new technology comes along : Eric Schmidt Prediction

Mr. Schmidt of Alphabet told his German audience that AI will not be reliable enough for the foreseeable future to exist in any capacity besides advisory roles.

These technologies [AI] have serious errors in them, and they should not be used with life-critical decisions, Mr. Schmidt said, Defense News reported. So I would not want to be in an airplane where the computer was making all the general intelligence decisions about flying it. The technology is just not reliable enough ― there are too many errors in its use. It is advisory, it makes you smarter and so forth, but I wouldnt put it in charge of command and control.

 

 

[MurMan]

Correct! At the risk of failing to convince you of Dan's source which you haven't refuted other than to disparage all researchers...

Lighten up. I didn't disparage all researchers. I used a well-known saying (that some find humorous) to point out that researchers often don't agree. You are expressing disagreement with this particular research, right?

 

Do you really think they built those images shown at the end of the article based on a raw EEG? If so, why wouldn't they have produced new images unlike the trained images to prove the actual image is encoded in EEG just like NTSC?

Where the hell did that conclusion come from? Of course I don't think that. I'm simply asking if there's some component of an image signal in an EEG that the SVM might be reacting to. It was an honest question. I don't know. You could have told me that the answer was "No" without injecting the sarcastic NTSC comment.

 

obtw, re-read the thread and you'll see that Theo introduced the analog TV signal analogy.

 

I only crawl out of the woodwork in my area of expertise because you chose to bait and bully Theo in the position of ignorance.

My history with Theo goes well beyond this thread. He has a long history of injecting himself into threads and making negative statements that he refuses to support. I didn't bait him here; he entered on his own volition. But I did challenge him to read the paper and respond. And what happened? He focused on the silly video in the overview article then went dark. Just like he normally does.

 

As for my ignorance, I don't claim to be an expert here. Just a "lay person" trying to figure it out.

 

If this is your field, why not make a contribution and shed some light on the topic? At the very least, provide some references that might help us out. Remember, this is a keyboard forum.

 

[Doerfler]

Remember, this is a keyboard forum.

 

Yes. Yes it is.

[jazzpiano88]

My history with Theo goes well beyond this thread.

I was just coming to his defense since his comments on the technology were spot on (and also confirmed by Dan's source).

 

You said you were baiting Theo purely for entertainment value and were "deadly serious" about lacking respect for him based on the matter, so I naturally had to push back. If you have a historical beef, why not take it offline rather than drag all of us into it?

[MurMan]

... why not take it offline rather than drag all of us into it?

There once was a time when I would patiently asked Theo to explain what he was saying. No longer. Many have tried over the years to work things out with Theo both online & offline and have failed. Lots of PMs from very well-intended forum members expressing their frustration in trying to work with him off-line. I'm a "lay person", not a psychologist.

 

If you think PMs will work with Theo, could you ask him to please stop dragging all of us through his multiple Kurzweil "correction" threads? He's put up at least six so far this year.

 

Don't worry, JazzPiano88. I have no desire to keep up this line of dialog. FWIW, I was sincere when I asked you to make a contribution to this topic.

 

[MotiDave]

The baiting I do is purely for entertainment value. Please feel free to ignore it.

My new sig. Mur - you get me.

[Ed A.]

Scanners...

 

 

[MurMan]

My new sig. Mur - you get me.

Nice to know that some still have a SoH.

[MotiDave]

My new sig. Mur - you get me.

Nice to know that some still have a SoH.

 

Its a San Diego thing, perhaps. Were not getting all ruffled about a few chips.

[jazzpiano88]

There once was a time when I would patiently asked Theo to explain what he was saying. No longer. Many have tried over the years to work things out with Theo both online & offline and have failed. Lots of PMs from very well-intended forum members expressing their frustration in trying to work with him off-line. I'm a "lay person", not a psychologist.

 

If you think PMs will work with Theo, could you ask him to please stop dragging all of us through his multiple Kurzweil "correction" threads? He's put up at least six so far this year.

Got it. You two have a history. I didn't know that, and I don't want to get into the middle of it other than to say that Theo was correct in his interpretation of the research for the purposes of this thread.

 

FWIW, I was sincere when I asked you to make a contribution to this topic.

I thought I had. I was trying to correct misconceptions that the researchers were actually reconstructing a facial signal contained within the EEG, as opposed to using a Machine Learning Algorithm (in this case a Support Vector Machine) to correlate to EEG attributes that had been trained on the same images used to draw their conclusions.

 

Here's a very simplified situation to make my point... Your dog always twitches his ears in a special way when the UPS truck pulls up to your house. Do you think his special ear twitch actually sees the truck, or do you think his specific ear twitch is correlated, and indicative, of his sensory response to a large truck pulling up to your house?

 

Just something to think about.

[Markay]

Here's a very simplified situation to make my point... Your dog always twitches his ears in a special way when the UPS truck pulls up to your house. Do you think his special ear twitch actually sees the truck, or do you think his specific ear twitch is correlated, and indicative, of his sensory response to a large truck pulling up to your house?

 

Just something to think about.

 

OK, I feel better now. So someone may be able to tell that I am visualizing a riff, but they will not be able tell what the notes are. I can safely put my tin foil hat up for sale on Amazon now in slightly used, near new condition.

[MurMan]

I thought I had. I was trying to correct misconceptions that the researchers were actually reconstructing a facial signal contained within the EEG, as opposed to using a Machine Learning Algorithm (in this case a Support Vector Machine) to correlate to EEG attributes that had been trained on the same images used to draw their conclusions.

Looking back on your answer, you did. Guess I was looking for something more complex than it needed to be. Your explanation is very clear. Thanks.