Jump to content

Welcome to FutureTimeline.forum
Register now to gain access to all of our features. Once registered and logged in, you will be able to create topics, post replies to existing threads, give reputation to your fellow members, get your own private messenger, post status updates, manage your profile and so much more. If you already have an account, login here - otherwise create an account for free today!
Photo

Mary Lou Jepsen physics demo of Openwater BCI tech at DLD 2018 New York


  • Please log in to reply
31 replies to this topic

#21
moderate_ai

moderate_ai

    Member

  • Members
  • PipPip
  • 11 posts

 

I'm sorry if I'm slow but I don't understand how OpenWater's device can "write" neural activity. I understand how her device works when it comes to "reading" neural activity. It's still unclear to me.

 

Does it require genetically modified cells? so that they can be light-sensitive (optogenetics)?

 

I don't believe it requires genetically modified cells. As far as I know if you heat up the brain cell membrane (with infrared and/or ultrasonic) it will fire, no GM required. I have no idea as to what resolution they expect to hit.



#22
engram

engram

    New Member

  • Members
  • Pip
  • 9 posts
 I don't believe it requires genetically modified cells. As far as I know if you heat up the brain cell membrane (with infrared and/or ultrasonic) it will fire, no GM required. I have no idea as to what resolution they expect to hit

 

 

Thanks for the reply.

That's sounds simple enough. Hope I can get my hands on the developer kits.

I heard her say it can read and write voxel for voxel. But didn't really state as to how many neurons it can simultaneously read and write. Would be interesting to know.



#23
engram

engram

    New Member

  • Members
  • Pip
  • 9 posts



I'm sorry if I'm slow but I don't understand how OpenWater's device can "write" neural activity. I understand how her device works when it comes to "reading" neural activity. It's still unclear to me.

Does it require genetically modified cells? so that they can be light-sensitive (optogenetics)?

I don't believe it requires genetically modified cells. As far as I know if you heat up the brain cell membrane (with infrared and/or ultrasonic) it will fire, no GM required. I have no idea as to what resolution they expect to hit.

Apparently, OpenWater is having some roadblocks when it comes to their products being used on the head/skin.

Infrared light is being absorbed by black hair or black skin. Meanwhile, focused ultrasound may necessitate head shaving.

#24
starspawn0

starspawn0

    Member

  • Members
  • PipPipPipPipPipPipPipPip
  • 1,842 posts

Where did you hear that?  That shouldn't cause them problems.  In fact, you could put a plate around the head that scrambles the light, and still make it work -- so long as it doesn't absorb too much of the light.

 

A little bird asked me what I thought Openwater's main challenge is, and based on early patent filings, I said the computational bottleneck is the main problem.  Basically, the computational load was too great to do much within the "speckle decorrelation" window of 10 microseconds. But new patent filings show they have changed the method somewhat.   Based on my cursory glance at it, I'd say the main problem with the new approach is the weakness of the light and signal you use (it's a bit complicated to explain the problem).  This might explain why they need such a pure laser -- but that's just speculation on my part.

 

The real breakthroughs will come from methods that can circumvent the speckle decorrelation window used by holographic approaches.  

 

Addendum:  I will add that if you are thinking the reason Openwater is pursuing medical applications first is a sign that they are having trouble getting BCIs to work, then you are misreading the situation.  I could be wrong, but it looks to me like they are thinking:  that's where the money is!  Sadly, I think they will discover how wrong that is.  MRIs, for example, are expensive in the U.S. and certain other countries not because the tech is so expensive to build and maintain -- but because of all the other stuff, e.g. contract negotiations (between large hospitals or insurance companies and medical supply companies).  I very strongly suspect that Openwater is going to find out the hard way how difficult it is to penetrate the medical supply industry.  



#25
engram

engram

    New Member

  • Members
  • Pip
  • 9 posts

Where did you hear that? That shouldn't cause them problems. In fact, you could put a plate around the head that scrambles the light, and still make it work -- so long as it doesn't absorb too much of the light.

A little bird asked me what I thought Openwater's main challenge is, and based on early patent filings, I said the computational bottleneck is the main problem. Basically, the computational load was too great to do much within the "speckle decorrelation" window of 10 microseconds. But new patent filings show they have changed the method somewhat. Based on my cursory glance at it, I'd say the main problem with the new approach is the weakness of the light and signal you use (it's a bit complicated to explain the problem). This might explain why they need such a pure laser -- but that's just speculation on my part.

The real breakthroughs will come from methods that can circumvent the speckle decorrelation window used by holographic approaches.

Addendum: I will add that if you are thinking the reason Openwater is pursuing medical applications first is a sign that they are having trouble getting BCIs to work, then you are misreading the situation. I could be wrong, but it looks to me like they are thinking: that's where the money is! Sadly, I think they will discover how wrong that is. MRIs, for example, are expensive in the U.S. and certain other countries not because the tech is so expensive to build and maintain -- but because of all the other stuff, e.g. contract negotiations (between large hospitals or insurance companies and medical supply companies). I very strongly suspect that Openwater is going to find out the hard way how difficult it is to penetrate the medical supply industry.

Mary Lou Jepsen on the podcast titled "Toward Practical Telepathy". She said black hair absorbs infrared light and couldn't penetrate compared to white hair/blond hair. But said her product works great if you have a shaved head. She said they're testing for black hair but still some roadblocks. Check the podcast out
Her product would definitely be hard to scale if this necessitates head shaving.

Addendum: "if you are thinking the reason Openwater is pursuing medical applications first is a sign that they are having trouble getting BCIs to work"

No i do not think that. It's pretty much open knowledge that they are doing medical applications first. It's in their website. Their BCI product comes later between 2022-2025.

They're doing medical applications first because it's easier (with the exception of the head). Any other organ, they can do and scan with high reso. In fact, they scanned a kidney and it has better resolution than an MRI. It's in their website. Their main medical application will be for the body. That's the first one. They definitely still have alot of time on how to deal with black hair and not having to shave the head.

#26
Kynareth

Kynareth

    Member

  • Members
  • PipPipPipPip
  • 185 posts

I heard about it and read about it but I don't believe in this, seems to good to be true. Remember Theranos? I think Openwater is like Theranos.



#27
engram

engram

    New Member

  • Members
  • Pip
  • 9 posts

I heard about it and read about it but I don't believe in this, seems to good to be true. Remember Theranos? I think Openwater is like Theranos.


I have some speculations. But I won't believe it till it is confirmed. Speculations like Mary Lou might actually just eyeing to sell her company to either Facebook, Google, etc. and we might never see the light of day of her product ever again.

Elizabeth Holmes, the founder of Theranos, had one year of college. Mary Loe Jepsen is legit. Graduated from Brown. Got her Masters at MIT then returned to Brown and got her Ph.D.

OpenWater is not the only one working on a non-invasive bidirectional brain-computer interface. DARPA is funding universities to figure it out.

I think in this decade we'll have some breakthroughs. But it'll be at least in the mid to late 2030s for it to be mainstream.

#28
starspawn0

starspawn0

    Member

  • Members
  • PipPipPipPipPipPipPipPip
  • 1,842 posts

Where in her Long Now talk does she mention dark hair and/or skin poses challenges?:

 

https://www.youtube....h?v=enFgn2sqoGw

 

I wanted to hear the context behind it.  

 

There is another potential business model here, that doesn't require scaling-up:  she could set up a lab that neuroscientists, academics, hobbyists, etc. use to generate data to help with their research.  

 

I asked a neuroscientist recently whether this is a viable approach, and he said it's a little backwards from how they normally think -- usually people give them the money.  But he said he supposes they could write a grant proposal requesting funds to go towards data-collection with a third-party company.  

 

So, the first company to build a viable BCI (high spatial and temporal resolution) might be able to tap into that government funding source.  

 

Dark hair and/or shaving heads wouldn't be a problem, as the goal isn't to scale to every individual.  Also, this would protect against authoritarian governments getting their hands on the tech -- though, they (or rather, companies in their home country) will reproduce it eventually, anyways.



#29
starspawn0

starspawn0

    Member

  • Members
  • PipPipPipPipPipPipPipPip
  • 1,842 posts

I actually know more than I'm letting on -- I'm wondering if @engram does, too.  This posting of engram makes me wonder; I don't think it's exactly the same approach (read-only) as the one I'm thinking of, and requires optogenetic manipulations (I don't believe those "other methods" do), 

 

https://www.futureti...-york/?p=268882

 

Could @engram be the "little bird" I've hinted at above (or someone who works for the little bird)?  If so, that would explain how they found me and my postings.  The references to my postings on Reddit were pretty well-buried; but the ones on futuretimeline are probably more widely read.  It could be that gwern's mention of my posting drew their attention to me -- I'm not sure.

 

I'll just say that people who are overly skeptical about the near-future of powerful, non-invasive BCIs are in for a rude awakening.  In fact, it probably won't just confirm what they actually hoped (about which they feign skepticism, to head-off disappointment), but will work so well it may cause them to fear.



#30
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 21,883 posts
  • LocationNew Orleans, LA

I'll just say that people who are overly skeptical about the near-future of powerful, non-invasive BCIs are in for a rude awakening.  In fact, it probably won't just confirm what they actually hoped (about which they feign skepticism, to head-off disappointment), but will work so well it may cause them to fear.

Eh, I admit I doubt it myself. As much as I want it to be true, I can't see advanced BCIs really taking off this decade in a real way. At least not this early.


And remember my friend, future events such as these will affect you in the future.


#31
engram

engram

    New Member

  • Members
  • Pip
  • 9 posts

I actually know more than I'm letting on -- I'm wondering if @engram does, too.  This posting of engram makes me wonder; I don't think it's exactly the same approach (read-only) as the one I'm thinking of, and requires optogenetic manipulations (I don't believe those "other methods" do), 
 
https://www.futureti...-york/?p=268882
 
Could @engram be the "little bird" I've hinted at above (or someone who works for the little bird)?  If so, that would explain how they found me and my postings.  The references to my postings on Reddit were pretty well-buried; but the ones on futuretimeline are probably more widely read.  It could be that gwern's mention of my posting drew their attention to me -- I'm not sure.
 
I'll just say that people who are overly skeptical about the near-future of powerful, non-invasive BCIs are in for a rude awakening.  In fact, it probably won't just confirm what they actually hoped (about which they feign skepticism, to head-off disappointment), but will work so well it may cause them to fear.


Nah. I ain't the "little bird" you know. I'm new. I've had questions myself about the product a few months ago and binge watch every mary lou jepsen video/podcasts i could find.

Toward Practical Telepathy, timestamp that video on 1:05:55

They're still testing for black hair/black skin.

OpenWater's product works great if it's scanning only for body. But if it's for the head, it may be a bit difficult to read and even write neural activity of the brain.

True and real BCI's are probably only invasive ones. Disappointed when I figured that one out.

neuralink (dbs), optogenetics, and in not so distant future: neuralnanorobots, are probably the only legit read/write BCIs.

#32
starspawn0

starspawn0

    Member

  • Members
  • PipPipPipPipPipPipPipPip
  • 1,842 posts

Thanks for the reference to that point in the video where she discusses dark hair.  She says they are testing that, but doesn't say that it prevents the thing from working at all.  I seriously doubt that it would prevent it from working.  
 
Basically, whatever structure you put in front of a laser beam to scramble up the light, and absorb it in certain places -- no matter how mangled the "exit pattern" (the light you get out on the other side), you can invert that process.  In other words, by recording the light coming out the other end, you can work out the pattern (basically, just record the light pattern, and project it back in the opposite direction), so that if you shine that pattern into the structure, out the other end will come a straight-line laser beam.  Jepsen demoed this a while back, on-stage (begins 5 minutes, 45 seconds in):
 
https://m.facebook.c...id=132302867112
 
She made a "brain laser" -- they figured out the pattern to project into a brain so that a straight-line beam comes out the other end.  
 
This is counterintuitive to most people, I think -- they don't imagine you can unscramble the light to such an exquisite degree through an object so complicated as a brain (or, rather, the light re-focuses to a beam as it passes through a brain)... but you can.  
 
Note, also, that the brain not only scatters, but also absorbs.  
 
Now, what if you put a skull and hair on that brain?  Do you really think it would make a difference?  The hair would absorb some more of the light, and so might the skull.  Furhtermore, there are so-called "snake photons" that can bounce around in the region between the brain and skull over a great length.  But all can be rectified in the end -- it all a deterministic process.
 
The main thing you have to worry about in order to pull off this feat is the "speckle decorrelation" time:  "speckle" refers to the little grainy interference patterns of laser light.  What happens with an object like a living brain is that blood moves around and changes the pattern -- very quickly.  And so, if you want to back-project light to make a brain laser with an actual living brain, you need to be able to bounce the light back quicker than the time it takes for the brain to change that pattern.  It turns out to be 10 microseconds!  So long as you operate within that window of time, you can do it -- no matter how tangly the hair.
 
....

Let me add another thing, unrelated to what you wrote, just to mention my point of view: 

As far as "reading minds" go, the skepticism on that is way overblown.  The main reason it is considered "hard", and why progress has been slow, is the fact that teams are operating with only very small amounts of data.  When you have only very little data, you have to be a lot more clever, and throw as much neuroscience knowledge at the problem as you can.  But if you an abundance of data, you can just train a machine learning model to figure everything out for you -- and it doesn't even have to be that complicated; a high school kid with access to Google Colab (to write and run Python programs) could do it.  
 
In fact, there was a recent Medium piece by the founder of Paradromics on the misplaced skepticism, centering it more on managing expectations and controlling hype:
 
https://medium.com/@...ge-855d37e1db01
 

With high-profile entrepreneurs such as Elon Musk and Mark Zuckerberg getting involved in the field of brain-computer interfaces, it is fair to say the technology is receiving its share of hype. Many scientists who have devoted their life’s work to creating meaningful and lasting advances in BCI are justifiably concerned about managing expectations to ensure sustained public support. It is in this context that several leaders in the field have expressed, in some form or another, the following reservation: We do not understand how the brain works; therefore, we are far away from advanced BCI applications. I believe this sentiment, though rooted in good intentions, is both reactionary and misguided.

First, though we don’t yet have an all-encompassing theory of mind and consciousness, neuroscientists know a lot about the brain, neurons, and neural coding. We know more today about how the brain works than Jenner knew in 1796 about how the immune system works. Critically, we know a lot about the localization of sensory and motor functions within the cerebral cortex and how brain activity in those areas reflect their respective functions.

More importantly, building useful BCI technology does not require a perfect understanding of the brain. Arguably, it doesn’t even require a good understanding of the brain. We know how to record brain signals to control prosthetic arms and computer cursors. We know how to feed tactile data into the brain for prosthetic feedback. We know that stimulating the visual cortex has enabled blind patients to “see” visual patterns. While it is true that BCI could be enhanced by additional neuroscience research, waiting until we have a complete understanding of the brain before building BCI-based therapies would be no less folly than if our predecessors had withheld the polio vaccine until the sequencing of the human genome was completed.


The esteemed neuroscientist Konrad Kording concurred, and went further and said that "understanding" may not even be necessary at all:

https://mobile.twitt...454463452315648
 

Matt Angle nicely makes the point that to have well-working BMI systems there is no reason to first understand the brain. Imho, it may not even really be helpful.






0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users