AI: Future of Humanity
Sphere of Interest
Stanford Brain Lecture Notes
The RX Project:
CV Biblio (1985)
Index of Essays
Robotics, and AI
Health & Biotech
Earth Wisdom: Universe
Be Saved by Bob!!!
(And Other Balms )
Are Fats Killers
The Mystery of CONSCIOUSNESS
Who, What, When?
Consciousness & Brain
Near Death Experiences: In the Desert With Pim Van Lommel
Is the UNIVERSE
Fine-Tuned for Life?
Neuron Videos Say
Forget Realistic AI
EUV 2014 - Future of Moore's Law
BAM: Brain Activity Map of Spikes
What is Watson?
AI Overlord or Tool?
SETI: Search for Extraterrestrial Intelligence
KEPLER Seeks Earth-like Worlds
STEVE PINKER in the Amazon: photos
Billion Year Plan:
CONSCIOUSNESS as Global Resonance
CT Scan: Yes!
Book Review: TRANSCEND
Create a Mind
Does Drug X
Works in Monkeys!
BAM: Brain Activity Map
Every Spike from Every Neuron
A recent research proposal called BAM for Brain Activity Map Project generated
much excitement. (The BAM proposal, published in Neuron in June 2012 is online,
and an earlier draft with far greater detail is also online.)
(Addendum: 18 Feb 2013: I started drafting this story in Nov, 2012.
Today it was headline news when it was made public that THIS is
the very proposal that President Obama alluded to in his recent
State of the Union address. See John Markoff's NY Times piece.
NIH is drafting a 3 billion dollar, 10 year proposal to fund this project.
Also see this 25 Feb 2013 NY Times follow-up by Markoff.)
The project was proposed by several leading scientists:
Paul Alivisatos (Director of Lawrence Berkeley Lab and head of nanotechnology),
Miyoung Chun (VP of Science for the Kavli Foundation), George Church
Prof. of Genetics at Harvard), Ralph Greenspan (UCSD Prof. and Assoc Director
of the Kavli Institute for Brain and Mind), Michael Roukes (founding Director
of Caltech's Kavli Nanoscience Institute),
and Rafael Yuste (HHMI Investigator
and co-Director of Kavli Institute for Brain Science at Columbia).
Note that all the authors are affiliated with or are consultants to the Kavli Foundation,
founded by inventor entrepreneur Fred Kavli. Kavli is one of a number of
billionaire philanthropists with an intense interest in accelerating neuroscience research.
This group includes Microsoft founder Paul Allen, David Sainsbury of
the Gatsby Foundation, investor James Simons, and Google founder Larry Page.
Common to all these efforts is the creation of cutting edge tools that combine
neuroscience, computer science, and bio/nanotechnology (imaging and
manipulation of matter at the molecular scale).
The essence of the BAM proposal is to create the technology over the coming decade
to be able to record every spike from every neuron in the brain of a behaving organism.
While this notion seems insanely ambitious, coming from a group of top investigators,
the paper deserves scrutiny. At minimum it shows what might be achieved in the future
by the combination of nanotechnology and neuroscience.
In 2013, as I write this, two European Flagship projects have just received funding for
one billion euro each (1.3 billion dollars each). The Human Brain Project is
an outgrowth of the Blue Brain Project, directed by Prof. Henry Markram
which seeks to create a detailed simulation of the human brain.
The Graphene Flagship,
based in Sweden, will explore uses of graphene for,
among others, creation of nanotech-based supercomputers. The potential synergy
between these projects is a source of great optimism.
The goal of the BAM Project is to elaborate the functional connectome
of a live organism: that is, not only the static (axo-dendritic) connections
but how they function in real-time as thinking and action unfold. Several efforts
to elaborate an anatomic micro and macro connectome have received wide publicity -
by eg Sebastian Seung,
Stephen Smith, Randal Koene, Ken Hayworth, and
Anders Sandberg. This proposal trumps even those ambitious projects
in sheer bravado. Science fiction or science fact? Let's see.
Not much is known with absolute certainty in the brain, but this is.
The brain communicates long distance information with (and only with)
neural spikes. This is true whether you're a fruit fly or a human.
Neural spikes are amazing. Hold a 1½ volt flashlight battery in your hand.
Just 1/15 of that voltage would be 100 millivolts, about the size of a neural spike
or action potential.
In its resting state a neuron maintains that huge charge separation across a cell membrane
that is just a few nanometers thick. It's a fire cracker waiting to explode.
(credit: Dieter Jaeger, Emory University)
When a neuron is charged and spikes a huge amount of energy is expended,
so evolution has made sure that neurons only spike when they need to.
Even so, the brain incurs a huge metabolic cost generating spikes. It burns
20% of your total energy. If you consume/ expend 100 calories per hour
(= 2400 calories per day), your brain's spikes are using 20% or 20 calories per hour
(no matter what you're doing). (By 1 calorie, I mean 1 kilocalorie = 1.16 watts per hour.)
So, neural spikes are as precious as gold. If you want to understand the brain,
follow the money: the neural spikes. They tell you what is happening, where, and when.
Now here's the central problem of brain science. The tools in 2013 are not yet
up to the task of following all the neural spikes. Functional MRI (fMRI) is
way too slow and too coarse to decipher neural spikes. fMRI has a
time resolution of several seconds - thousands of times slower than a neural spike,
which takes only about 1 millisecond (msec). Furthermore, each voxel in fMRI
is typically about 2 by 2 by 4 mm on a side or 16 mm3, intractably big
considering that each mm3 contains roughly 50,000 neurons, ie about
one million neurons per voxel. Understanding the brain by looking at
fMRI is like trying to understand human civilization by looking
from earth orbit at the heat or light given off by cities.
In contrast to fMRI, electrical recording is far faster, eg recording EEG or,
even better, recording from single units (neurons) as they spike yields activity
at the msec level. But EEG has even worse spatial resolution, about a cm, than does fMRI
(and, of course, single unit recording with electrodes is usually only done in lab animals).
Enter calcium imaging, as in this video from MIT Prof. Feng. Every time a neuron fires,
it glows green (from the GFP, green fluorescent protein, grown inside). The video shows
neurons firing in a brain slice kept alive in a petri dish.
Notice one big problem: the brain needs to be sliced to see the light given off
by each neuron firing. How do you see neurons firing beneath the surface of an intact brain?
The human neocortex alone is 2 to 4 mm thick, so how would you ever see
the spiking pattern of a large collection of neurons? (As a separate matter,
much of the action in brains takes place several cm deep inside subcortical regions
like the thalamus and basal ganglia, where the only hope of monitoring the action
is with either fMRI (too slow) or microelectrodes (too invasive).)
Enter 2 photon microscopy. Invented by Winifried Denk (who shared
a million dollar Kavli Prize in 2012), 2PM (two-photon microscopy) uses red-shifted light
from two simultaneous laser beams that intersect at the region of interest.
2PM greatly increases the depth of imaging possible in living tissues up to about 1 mm.
For deeper structures, we still need microelectrode arrays of various sorts,
since they are out of range of microscopes. Microelectrodes were traditionally either
single shank (one shaft) wires or glass pipettes drawn to very fine points. But in this area,
too, huge progress has been made. Researchers now use multi-shank arrays of
electrodes, as in the figure below. Some current arrays also incorporate optical fiber bundles.
(credit: Ed Boyden Lab, MIT)
Finally, to figure out how the brain accomplishes perception, cognition,
and behavior it's extremely informative to be able to turn key pieces of
the machinery on and off. Without this ability it's hard to claim that
a given neural circuit actually causes a given phenomenon
(rather than simply being a marker).
A key tool in reverse-engineering the brain is optogenetics
in which light-sensitive opsins from algae are genetically inserted into mammalian neurons.
Once modified, the neurons can be turned on or off literally at the flip of a light switch.
Entire circuits can be activated or silenced to make mice (eg) walk, turn, or sleep.
My son Sean (a molecular biologist) and I were first introduced to this technique
when we heard Stanford optogenetics pioneer Karl Deisseroth present his work in 2006.
(Here Karl presents optogenetics to a lay audience at Stanford 18 mins.)
MIT Prof. Ed Boyden, who began in Karl's lab, is another star of optogenetics.
(Here Ed presents optogenetics in an 8 min YouTube at SPIE and
here in a 19 min Allen Institute YouTube.)
The most challenging aspect of neuroscience is obtaining fast, real-time
monitoring of a broad swath of deep brain structures. Recall that fMRI is too slow
and too coarse. The version of the BAM Project that made it into Neuron just hints at
possible methods that are elaborated in the longer draft version. Also see this article on
cutting-edge neuroscience tools by Google researcher Tom Dean.
This is nanotech on steroids. 1) Nano-sized RFID tags that derive their power
(But please note: just because I'm enthusiastic about the overall
from the same externally applied fields used to read their data (on neuron location and
spiking history). Early versions would be implanted. Later versions might migrate
intravascularly . 2) Quantum dots and diamondoids that amplify voltage-sensitive dyes
to increase emissions used for monitoring activity.
proposal, doesn't mean my enthusiasm extends to all the proposed
techniques. They range from credible extrapolations of the present to those
that would seem to be straight out of b-grade sci-fi.)
It seems apparent that reverse-engineering the human brain
(and building new and improved versions) is an irrestible scientific attractor
that is drawing in the money, talent, and tools to make it happen.
The European Flagship Human Brain Project may create the computational
capability to simulate large, realistic neural networks. But to compare the model
with reality, a real-time, functional, brain-wide connectome must also be created.
Nanotech and neuroscience may be mature enough to justify funding this proposal.
(See this PBS News Hour interview of NIH's Director Francis Collins,
a BAM enthusiast. But note that reservations abound: "too expensive in these
austere times, too much focus on one highly speculative approach, difficult
to apply to humans, etc." I made a bet that BAM will yield crucial breakthoughs,
although my payoff date is in 2027.)
Addendum: March, 2013. Here's an article showing the state of the art
of whole brain functional imaging in zebrafish larvae.
Addendum: April, 2013. In early 2013 I had a lengthy correspondence with
my friend Stanford Prof. William Newsome about the pros and cons of this proposal.
Then, imagine our surprise when NIH Dir. Francis Collins tapped
(with the brilliant Prof. Cori Bargmann) to direct the rollout of the project!