Oct 29 2010
Digital Mind Reading
We are making progress, very slowly, in “reading” human brain activity. This is an exciting area of research, but reporting about it consistently falls prey to one typical formula of science reporting – lead with the most speculative and fantastical potential application of the technology.
In this case just about every news story I read about the latest study involving reading brain activity leads with the idea that this technology can be used for either mind reading or dream recording. It would be better for such far future extrapolation to be buried deep within the article, after some more balanced discussion about where the science actually is.
I know this is a frequent quibble of mine, but it does affect the public perception of science. The public is constantly hearing about how one tiny advance in our understanding of some aspect of biology might lead to a cure for some horrible disease. Wasn’t the genome project supposed to lead to countless cures and medical advances? Where are all the invisibility cloaks we were promised, and why can’t I watch my own dreams like a high definition movie?
After a few years of these bold promises not coming to fruition, the media begins to write stories about how the original advances were overhyped and maybe now they are not going to provide us with anything. It is an iatrogenic news cycle of the media.
Admittedly, naive scientists play into this phenomenon as well. It’s easy to get goaded into saying irresponsible things, especially when your research is getting media attention.
It is with all this in mind that I read the latest news reports about a dream-recording technology. Sure – one day we may have Matrix-type technology where we can translate brain activity into a virtual reality, and vice-versa. But the current research is very far away from this.
What we have now is another research group conducting research that basically looks for neuronal correlates of specific ideas and images. We already know that often an idea can be recorded in a single neuron in the brain. This is an important concept, and certainly lends itself to brain mapping. If the same neuron fires every time I think about Wink Martindale, that certainly creates the potential for interpreting the meaning of brain activity.
What Dr. Cerf and colleagues have done is test subjects that already have electrodes implanted deep within their brain for the purpose of studying their epilepsy. His research team was able to record the firing of specific neurons and correlate them to specific concepts or people. They demonstrated that this firing was consistent.
This is great, although nothing really new. There are also many limitations. This level of precision requires implanted electrodes. This is not a minor hurdle but a major limitation of any practical application. Sure – one day we may all have supercomputers implanted in our skulls with millions of connections to our gray matter. Or we may develop scanning technology with single-neuron precision, but not anytime soon.
Also – the proof of concept here is that we can map individuals to find out which of their neurons correlates with certain concepts. An individual has to be mapped before they can later be read. You cannot apply any generic map to a person. This is probably not a significant limitation – but means that any software that reads neurological activity will have to go through a calibration period where the subject will be exposed to standardized stimuli while their brain is recorded.
The question is – how elaborate and time-consuming will such a calibration phase need to be before any useful complexity is achieved? How many millions of concepts are stored in our brain, and how useful will mapping a few thousand of them be? I suspect that mere advances in raw computing power will greatly help such tasks, so this might ultimately not be a limiting factor.
However, there remains an important proof of concept limitation. So far researchers have mapped visual information and individual items or people. But what about a complex narrative – the complex interplay of language, vision, sound, abstract concepts, and emotions with people and things? Reading or creating a full virtual experience may be orders of magnitude more complex than reading individual components.
My concern is that this complexity will follow the rules of chaos theory, and we will no more be able to map full experience than we can predict the weather a month from now. This goes beyond high-fidelity recording. We may be able to record someone’s mental activity and play it back for them – so you could re-experience anything recorded in the past, maybe even your own dreams. This process would involve recording and playback only – not interpretation. The same recording may be gibberish to any other person, and could not be played on a video for anyone to watch.
My question is about the ultimate limits of interpreting brain activity. I’m not sure what the answer is. I follow the research to see if we are making any progress in this regard. We are – but it is still so preliminary we cannot yet answer this question.
14 Responses to “Digital Mind Reading”
Leave a Reply
You must be logged in to post a comment.
Recording dreams is a cool idea, but I can’t think of a practical application.
However mapping enough of the brain to make using artificial limbs practically inseparable from real limbs – that is worth mapping. And I don’t suppose it would be quite as involved since its fairly targeted and not trying to encompass the entirety of the human experience.
@SARA
Yes, please. If I can’t get my flying car, can I *at least* have some extra limbs and arbitrary-wavelength vision? Kthnxby.
Really I think if we are able to have high-fidelity neuron-level recording of brain state that implies such a high level of technology that its currently impossible to say what that sort of world would be like.
…and needless to say its inappropriate for newspaper to speculate on something so far flung. Newspapers should leave that sort of thing to folks like Vernor Vinge.
I know Dr.N’s expertise is neurology and the parts of this post dealing with the brain research seem extraordinarily well thought out — I particularly liked the point of complexity following the rules of chaos theory– is that original (cause it is really good!!) .
(odd thought–If the brain is changing all the time (plasticity),then couldn’t we run into the problem that a recording I made a few days (weeks or years…) ago wouldn’t work for me now, much less anyone else?)
However, I am concerned that the comments about the news media and exaggerations of results and research by scientists might be naive.
http://blogs.telegraph.co.uk/news/jamesdelingpole/100058265/us-physics-professor-global-warming-is-the-greatest-and-most-successful-pseudoscientific-fraud-i-ha ve-seen-in-my-long-life/
I really am concerned that Lewis is correct here. If so, we need to be more vigilant than ever about claims.
Did you just decide this blog somehow supports global warning denialism? Um, congratulations?
Steven Novella:
“My concern is that this complexity will follow the rules of chaos theory, and we will no more be able to map full experience than we can predict the weather a month from now.”
Sonic:
“I particularly liked the point of complexity following the rules of chaos theory…
us-physics-professor-global-warming-is-the-greatest-and-most-successful-pseudoscientific-fraud-i-have-seen-in-my-long-life”
BillyJoe to Sonic:
What is the difference between weather and climate?
Harold Lewis is Emeritus Professor of Physics at the University of California:
“Anyone who has the faintest doubt that this is so should force himself to read the ClimateGate documents, which lay it bare. (Montford’s book organizes the facts very well.)”
ClimateGate?
Montford??
Please!
Both have been thoroughly debunked and anyone who has the slightest doubt about that should be thoroughly ashamed of his ignorance on the matter.
Signed, sealed, delivered.
Please open the package and read.
Stand aside professor.
You’ve slipped out of your depth and out of your mind with your crap flowing out behind you.
Dr. Cerf is hoping that he can read minds of patients in coma. He assumes there are images popping up in coma that he can trace to actual images. What if he turns on the equipment and finds that all he can get is static on his equipment monitor? I’ve seen coma and drug-induced coma patients. How would erratic brain pressure figure with Dr. Cerf’s expected images?
It is extremely likely that the operation of the brain is chaotic in a mathematical sense.
Essentially all physiological processes that have been looked at carefully enough do turn out to be chaotic, including heart rate variability and mitochondrial control. To get chaos, all you need are a sufficient number of non-linear coupled parameters. It is the non-linearity and the coupling of fluid flow, heat transfer, mass transfer and thermodynamics that makes the weather chaotic. Physiology has many more degrees of freedom than the weather does. It would be extremely surprising if the brain was not chaotic in a mathematical sense.
From an engineering standpoint, you want the brain to exhibit chaos, so that you can go from one state to another state with differential effort. It turns out that all natural neural networks self-regulate in the near critical percolation threshold, a true mathematical critical point, where the properties of the network change exponentially with connectivity and actually diverge (become infinitely sensitive to change) at the critical percolation threshold.
This feature of neural networks is quite important in understanding brain behavior under extreme stress. The neurotransmitter that sets functional connectivity is nitric oxide, and under stress the basal NO level becomes lower. That lower NO level can be observed as reduced BOLD fMRI signal (neurogenic NO causes the vasodilatation, with a lower basal NO, the neurogenic NO doesn’t produce as much BOLD fMRI signal because it is the sum of NO from all sources (neurogenic plus basal) that causes the NO effect.
The critical percolation threshold is a point, and is only metastable, so a neural network cannot operate at that exact point very easily or for very long. With distance from the critical point but from below, the functionality drops as exponentially as it increases with closeness from above. In other words the functionality of the network drops precipitously below the critical point. This is what causes the “melt down” observed in autistic individuals under stress. Stress lowers the basal NO level, the connectivity in the brain drops, sensitivity increases until the critical percolation threshold is passed when functionality then collapses. This is what happens in acute psychosis too, in bipolar disorder and in the euphoria of near death metabolic stress. It is probably also what happens in coma and unconsciousness. Drop enough below the percolation threshold and the brain cannot exhibit collective properties.
You can observe an acute decline in basal NO levels in the brain by the effect on water diffusion via MRI tensor imaging. It is called “diffusion”, but any movement of hydrogens amounts to the same thing. A major source of water movement is the transport of cargo down axons. Each neuron has axons that extend long distances from the cell body. In the case of motor neurons the cell body is in the spine and the axons may extend as much as a meter (in tall people) from it. The DNA is in the cell nucleus in the cell body, so that is where protein synthesis is, and where mitochondria biogenesis must be (99% of the proteins in mitochondria are coded by nuclear DNA). Those mitochondria are transported out axons to the tippy end to supply each axon with ATP. Carried out, and then carried back when the mitochondria get “tired” after a month or so and must be recycled. The active movement of cargo out the axons and then back to the cell body consumes a lot of ATP, and is one of the first things shut down when there is not enough. NO is the diffusible signal that regulated the ATP level in all of the neurons to be the appropriate level so that the whole brain works together “in sync”.
This water “diffusion” is anisotropic and is associated with axons and does drop precipitously during ischemic events. There are other components to it also because the “diffusion” drop seen in white matter hyperintensities persists even after death (when ATP and axonal movement goes to zero).
Brain properties can change very rapidly. When a woman gives birth and bonds to her infant, that change happens extremely rapidly (a few hours) and persists long term. Memories are also very rapid changes to brain properties. These changes are non-linear and are coupled (one memory or experience builds on another). Maybe some simple properties like sensory perception will stay pretty constant, but probably not. Searching of visual scenes improves with practice. There must be remodeling of the neural pathways involved in visual processing.
I am fascinated by the fact that almost any post can be turned into an argument on global warming (just as Daedalus can turn any post into one about NO
)
My comment was not about global warming. My comment was about the effect that money and bold claims have had on the science reporting.
The gentleman I linked to, Lewis, is of the opinion that the fact that trillions of dollars are on the line has had an impact on the science and the claims that have been made about what is known regarding the future climate of the planet.
Certainly given his experience, I am willing to consider the proposition. Having considered it–
I think he is correct about that.
His next opinion (that AGW in particular is a pseudoscientific fraud) is interesting, but is not completely applicable to the point I was trying to make.
BTW- I don’t know all the answers about the future of the climate of this planet. I should be more aware that is a touchy subject.
Don’t talk politics, religion, money or vaccines or AGW. (new rules for dinner conversation)
Not any post, only ones that at their core are based on physiology that does involve nitric oxide.
Happy Halloween. I’m dressing up as a CAM practitioner.
lazy too
All natural neural networks self-regulate at the near percolation threshold. That is a true critical point, where the properties of the network change exponentially with connectivity. Again, that is the idea place to self-regulate at because it takes differential effort to get to another place. A single neuron firing can change the whole state of the brain. Can’t get much more non-linear than that.
Posted by: daedalus2u | March 1, 2010 5:27 PM