DMC-FX7
ƒ/4.2
12.3 mm
1/8
200

Uploading from the Almaden Institute on Cognitive Computing…

I’m sitting next to Jeff Hawkins who just posted a white paper (PDF) on the Hierarchical Temporal Memory framework. Hawkins writes: “The brain does not ‘compute’ the answers to problems; it retrieves the answers from memory… The entire cortex is a memory system. It isn’t a computer at all.”

(aside: I blogged a bit about the developmental trajectory of electronics recapitulating the evolutionary history of the brain. Specifically, both are saturating with a memory-centric architecture.)

In this photo, you see Henry Markram from EPFL showing complex neuronal visualizations and movies from the BlueBrain project. This is a single neuron with a color-coded map of synaptic connections to other neurons.

Some notes:

No brain wiring is the same, even in clones. Structural topology and functional spike train variation is immense. There are over 300 types of neurons in neocortex that are structurally and electrically different. And they each have ~200 ion channels from a pool of 20-40 variations. Looking at possible neuronal connections, it’s an all:all structure that is in place. But only 10% of the connections will grow a synapse.

Nobelist Gerald Edelman: “Every perception is an act of creation. Every memory is an act of imagination.”

He also referenced Posit Science founder and UCSF Prof. Merzenich’s research on rapid recruitment of neurons, a form of neural plasticity. If you cut the nerve to a thumb, the cortical sensory map changes within 1 hour.

“We are at a remarkable juncture… We are at the brink of understanding what had only been in domain of philosophers.”

29 responses to “Upload”

  1. angry shock Avatar
    angry shock

    “We are at a remarkable juncture… We are at the brink of understanding what had only been in domain of philosophers.”

    I love this phrase. Needless to add: and the facts which give origin to it.

    He couldn´t have been more precise, neither more humble, being a scientist himself.

    Reply
  2. benjiman Avatar
    benjiman

    This immediately reminded me of a very old book I read as a kid.

    The Binary Brain.

    I was so eager for this future to arrive and here we are on it’s doorstep.

    Reply
  3. angry shock Avatar
    angry shock

    I´ve been reading about the advance done in relation to understanding the mechanisms of hypnosis, and how this is helping neuroscientists understand the whole of the cognitive act, the mysteries of the brain. (article in spanish)

    They say that the key that these studies show is that what weighs in our perceptions, hence our actions and reactions (thought or acts) given a certain something happening "outside" ourselves, is not the raw sensory data the brain receives and processes first hand, but actually what it elaborates from this, mixing this data with our memories to make it consistent to our current model of the world, to our previous experience.

    The final result of this digestion is finally delivered to our conciousness (process called "feedback"). But we think that this is the actual "objective" data retrieved from our senses (called "feedforward").

    It´s impressive how much more synapsis are commited to the "feedback" (top-down) part of the process, rather than to the "feedforward" phase (bottom-up).

    So this called top-down process is what determines what are we going to feel or think about something we perceive, not the perception itself, neither of course the thing we are perceiving. We never reach it.

    It´s all a construction of the brain.

    Plato knew it. 😉

    Reply
  4. angry shock Avatar
    angry shock

    oh, and say Hi to Jeff!!!

    Reply
  5. mixed ink Avatar
    mixed ink

    on saturday morning, after a hard champagne party with friends, I fell violently on the head : it was a stupid bike accident. I mean it. the bike has no brain, it is not a horse…
    It was for me the special occasion to check-up my neural consistency, to verify any cortex, any answer to questions in my very own depth.
    the neurons connexions are well working perfectly. i think.
    I would like to thank you for understanding how well it works, and sharing it with us here.
    you make my night.
    peace -_-

    Reply
  6. Todd Huffman Avatar
    Todd Huffman

    "Every perception is an active of creation…"

    This is a fantastic point, every sensation has the ability to alter every subsequent sensation.

    I am very familiar with Markram’s work, he is one of the more insightful and exciting neuroscientists of our day. Are there videos of the talk? I am interested in what his future plans are for the Blue Brain project, and would like to hear it in his own words.

    Reply
  7. Mikey720 Avatar
    Mikey720

    Ohhhhhh, so the blue thingies are attached to the purple thingies…

    Well that just makes sense now! :p

    But on some rudimentary level, I’m understanding what your saying, and its all very very interesting! Thanks for sharing!

    And one more thing, soooo, does this mean that one day, following this framework, we should be able to create a completely artificial brain that functions in the same manner as a human brain?

    Reply
  8. angry shock Avatar
    angry shock

    On your note about brain wiring not being the same, even in clones, I found this here which I believe it´s very interesting -all the website actually-:

    "When you consider the central importance of the shape of the neuron in its function, you begin to wonder what determines the architecture of each cell. In part, this is built into the genetic code, but think of the enormous amount of information would have to be stored genetically were genes to specify exactly the hookup of each and every cell. This would be impossible. Connectivity and shape needs to unfold and develop within the lifetime of the organism and is part of a cascade of events that unfolds in the embryo. Later, this modifies further as the effects of learning and experience are reflected in the formation of new synapses and alter the cell shape, perhaps by creating new axon and dendrite sprouts. The exact positioning and structure of supporting cells like blood vessels and glia have their own effect in determining the final structure or design of the neuron. Undoubtedly there is a constant feedback loop of interaction between neural transmission and structure that determines it. The structure determines function which determines structure in a sort of continuous dialog."

    (bolding is mine)

    Reply
  9. jurvetson Avatar
    jurvetson

    alieness: Exactly…. The massive number of inter-neuron connections in an adult brain could not be simply encoded in our DNA, even if the entire DNA sequence was dedicated to this one task. There are on the order of 100 trillion synaptic connections between 60 billion neurons in your brain.

    This incredibly complex system is not ‘installed’ like Microsoft Office from your DNA. It is grown, first along chemical gradients in a fetus, with widespread connectivity sprouting from ‘static storms’ of positive electro-chemical feedback, and then through the pruning of many underused connections through continuous usage-based feedback.

    At the age of 2 – 3 years old, humans hit their peak with a quadrillion synaptic connections, and twice the energy burn of an adult brain. By 4 years old the number of synapses has dropped about 10x.

    Much of the power in bio-processing comes from reentrant mapping and the use of feedback in the electrical, physical and chemical domains.

    It’s a remarkable system, and quite the inspiration for this confernence. Represented in compressed digital form, and burned on a CD, the human genome is smaller than Microsoft Office.

    oddwick: not sure if they will publish the videos…. I took a couple videos which I might post on revver if they turn out ok.

    Reply
  10. Ian David Blüm Avatar
    Ian David Blüm

    This is exactly why I gave up a biotechnology career to pursue a degree in Neuroscience.

    Reply
  11. TroyWorks Avatar
    TroyWorks

    >"Every perception is an active of creation…"

    better start being careful on what one perceives.

    with all that variation, makes me increasingly dubious that neural scanning into silicon/software equivalent for uploading is viable, at least with any verifyability….I know I won’t be sigining up first.!

    I was just watching a video off google at google with the future of computing, where the speaker was touting FPGA’s over conventional processors, and I noticed as well the higher the speed of serial pipelined processors the greater the memory complexity to deal with the variations in information created (the input byproducts) during processing.

    Holographic/quantum memory retrieval I imagine is a step further down this path of processing more an act of memory retrieval/interference than clever bit flipping.

    Reply
  12. Vanita Avatar
    Vanita

    From the view out the window here, it looks like we are coming in for a night landing at DFW. It’s a hub you know. ‘-}

    Reply
  13. jurvetson Avatar
    jurvetson

    vanita: you are looking very dualist today. Yeah, air traffic control is the superego.

    Resonant homologies and fractal developmental patterns again… You remind me of the picture comparing a city plan and brain inside this book:

    TroyWorks: very interesting… was it memory topology complexity (e.g., multi-port or multi-level memories) or memory hierarchy complexity (L1,2,3 cache)? I was wondering about multi-port topologies when talking with Hawkins the first time around… Interconnnect is the physical implementation challenge…. to achieve the fine-grained details of dynamic synaptic fan-out.

    Reply
  14. angry shock Avatar
    angry shock

    There is no other album in all flickr like yours, Jurv.

    How much greatness condensed in one place. Not only your art through your photos and your posts themselves with all their insightful and mind triggering contents… but the most amazing comments and contributions you arise with them.

    This is not a photo album, this is a modern version of Plato´s Academy.

    Had to say it.

    Reply
  15. jurvetson Avatar
    jurvetson

    gulp… thanks….. it’s a great community….

    Oddwick: per your request, I just posted a video on Revver. The poor video quality is because of my camera. It was remarkably sharp on screen.

    More on early brain wiring: Toby Berger, University of Virginia, just shared an interesting summary: From age -1/2 to +2, we we form millions of new synapses per second… day and night.

    Troyworks (con’t): Joaquin Fuster just listed some of the modeling constraints for human memory (and indirectly hints at the needs for a biomimetic memory architecture):

    • Architecture (network)
    – Relational code
    – Hierarchical and heterarchical
    – Plasticity

    • Dynamics
    – Content-addressable
    – Stochastic
    – Reentrant
    – Hierarchical and heterarchical
    – Parallel
    – Serial processing element for conscious attention working memory
    – Categorical (degeneracy)
    (a) in perception
    (b) in action (same output from different inputs)

    “The idiosyncricies of our memory is a direct result of the combinatorial power of 10-30B neurons in our neocortex.”

    Reply
  16. greentea flute Avatar
    greentea flute

    re benjamin, at 14 cents US, "binary brain" is probably a bargain! my old book was frank rose’s "artificial intelligence," re vanita and Emergence and complex systems, Jung suggested that we are constantly recreating around us the structures inside of us, down even to the molecular level, thus seemingly abstract contemporary art;
    but speaking of philosophy, (philosophy being the field for determining what we can know and how we should respond to what we can know,) which used to be a field of knowledge that was one with science, today philosophy is struggling to keep up with all of this new knowledge, particularly quantum mechanics and brain – computer comparisons, ala artificial intellegence, we are getting way beyond Heidegger and Bertrand Russell here; all of this poses huge problems for philosophers, and educators, just note the difficulty people are still having with concepts such as evolution … so my question is how can we get this stuff out there into the mainstream, because it is exciting and enlivening, not just useful for building cognitive computers, but fascinating because of what it tells us about our world, our place in it, and how we perceive it … maybe Jeff can figure that out! thanks for posting!

    Reply
  17. Todd Huffman Avatar
    Todd Huffman

    Thanks Steve!

    Reply
  18. mares_romero Avatar
    mares_romero

    Hi Steve,

    Aircraft complexity is increasing exponentially, mainly due to the avionics. I’m designing the avionics system for a regional/corporate airliner to be in service in 2012 (preliminary design, architecture-level, undergrad project). I’m not expecting to find anything deeply transformative (although I think I might be on to something…) but I’m trying, and it’s definitely a good excuse to do some serious trend studies…

    Has anybody done any work in applying HTMs to complex system design/analysis?

    Thinking about Kasparov standing a chance vs. Deep Blue in chess makes me think pattern recognition might be better than classical analytical methods in say, studying the reliability of a complex system such as an aircraft… Or more excitingly, designing a complex system.

    I couldn’t agree more with Jeff Hawkins in that complexity is subjective. (‘complexity is a symptom of confusion, not a cause’)

    It’s truly amazing to have the chance to speak to you directly like this. Thanks for your insightful/inspiring blogs, and I hope you make a visit to Imperial College some time and enlighten my fellow engineers/scientists!

    Regards,
    Javier

    Reply
  19. mixed ink Avatar
    mixed ink

    2012? You sure ? Isn’t that the end of times ? -_-

    Reply
  20. angry shock Avatar
    angry shock

    LOL! shhhhh… keep it secret. 😉

    Reply
  21. mares_romero Avatar
    mares_romero

    Oops… I think Steve thought I was pitching at him, and ran away!
    =)

    Anyway, interesting post… as usual.

    Reply
  22. quinet Avatar
    quinet

    Jeff Hawkins approach has considerable implications for information design. I’ve been delivering seminars to corporations about how what we know about the brain affects how we should design information.

    It’s all about pattern recognition!

    – Thom

    Reply
  23. zhoufeimu Avatar
    zhoufeimu

    Great! emergence …..interesting!

    Reply
  24. jurvetson Avatar
    jurvetson

    Javier – just not sure I understood the question…

    Videos and PPTs are now available from the conference:
    http://www.almaden.ibm.com/institute/agenda.shtml

    Reply
  25. zany curve Avatar
    zany curve

    Great shot – so great in fact, I borrowed it!

    I was looking for something that spoke to my point, and found this shot I can pull it down if you like, and if you will let me use it, I can add a link back to whatever you would like, as well as credit however you would like. It is at http://www.neuralplasticity.org and on the "the answer:plasticity" page.

    on a couple of points others have raised:

    It is funny that the brain is obviously a pattern recognition mechanism bot on its external processing and in how it processes internally, and that it is taking us to learn how to consciously do pattern recognition to understand how the pattern recognition is occurring. Almost like a B movie using the whole time travel conundrums!

    as for avionics – I gave them up to to trouble shoot the ultimate flight director! I have a licensed A&P as a well as a Phd. I worked as a aircraft mechanic for 7 years. The skills and processes, and trouble shooting principle I learned there have not only helped me incredibly in neuroscience, they also shaped the way I look at and study this human flight director system! It is far different than the way most neuroscientists approach the brain!

    Reply
  26. jurvetson Avatar
    jurvetson

    sure. all of my photos are free to use with an attribution link back to the original.

    Reply
  27. Marc Binkley Avatar
    Marc Binkley

    Does anyone else see the similarities between the shape of neuropathways and that of social networks?

    Reply
  28. jurvetson Avatar
    jurvetson

    a common pattern in scale-free networks. These patterns arise when a network grows with preferential attachment of new members, and thus they describe most complex networks of interest (Internet, dating, power grids, genomes and proteomes, corporate Board seats, etc.).

    The fire-together wire-together rubric for neuronal pruning from usage-based feedback is a form of preferential de-attachment in cortical systems.

    (more on this from Albert-László Barabási’s book Linked).

    Reply
  29. test35965 Avatar
    test35965

    Sadly, the Google Video version of this fantastic presentation is gone (with GV itself). I’ve cleaned up & posted another copy on youtube: http://www.youtube.com/watch?v=9gFI7o69VJM

    "Visualization is a way of pre-screening the data…. and having fun at the same time" 😉

    Reply

Leave a Reply to test35965 Cancel reply

Your email address will not be published. Required fields are marked *