Or…the philosophical Neuro-babble of scientism!
This whole thing about digital is getting “curiouser and
curiouser” by the day. When you throw in the barrage of hyperbole, the whole mix
is a labyrinthine network of chaos. Facts and opinions merge into a single syllabic
“wow” from the pedantic crowd that consumes latte with every breath.
So this next coming of the “sliced bread” is the digital
what’s what that will take us into the 22nd Century. Hey but the 21st
Century just began, so can we give this 0-100 in nanoseconds to the future a
bit of a rest?
Artificial Intelligence (AI) will be the in thing, they say.
Who, you might ask? Those in love with their abilities to discern the future, that
is! Those that hide in dark closets and code and decode the subject of life but
may not have lived it yet or never will from within the confines of their
dark rooms. “Too abstract for you, this is?” as Yoda would ask.
I will take my cues from the functional Magnetic Resonance
Imaging and from the perch of the Single Positron Emission Computerized
Tomographic reveal. Why? Read on… it might make sense. In an abstract sort of
way, it does to me.
The human mind is a rich diverse group of billions of neurons (brain cells) that converse electrically, collude, recruit and develop
denatured protein memories within. The more the force of thought resident on a
particular path, the more delineated the path. Imagine a path less traveled
like the jungle infested forest facing Prince Phillip before he can kiss Snow
White. However, if the same path is well traveled, minus the evil vain-queen-witch, it becomes a paved highway over time from travel. This paving is a
function of experiential gains. The plasticity in the brain of pruning (or
hacking away if you prefer as in the case of Prince Phillip to see his lost
love) is a daily function of the brain. Oh yes, doesn’t matter how old or young
you are, pruning makes the world go around from 0 to 100+ years. The thoughts
become the actions and lo and behold our world changes. But how do or can we register
those thoughts and from that create the epistemic nature of the action? And if
we can, can we then objectify the precursor to any action or behavior? Ah dear
readers, there is the slope that provides us with equal measure of frills, thrills
and spills.
Scientists are tripping over in the philosophical realm of
neurobiology and neuro-functional anatomy with a multitude of hardware to peer
into the moment by moment of each firing neuron to simulate brain function. Indeed,
they claim, we may be able to pocket your mind into an iPod one day, Moore’s Law
be damned! But there are a few roaches I see in that prospect. For instance, let
us take the most admired one called fMRI or (functional Magnetic Resonance
Imaging) as a means of deciphering the brain’s activities. You must have seen
the glossy images of colored tripped brains of individuals in response to some
stimulation or behavior or action? I am sure you have. If you haven’t here are
a few images to re-polish that paradigm…
What happens with an fMRI anyway? My simplistic viewpoint,
and it is simplistic, is that when a stimulus is provoked, say an image of
something to provoke a reaction from the individual’s brain, there is increase
in energy output from select recruited neurons that identify the site of
activity, eg. Temporal lobe or the occipital lobe where vision and memory merge
and if an action is desired then the Parietal
lobe comes into play, but through it all the cognitive orders of action thru
assimilation of the diverse stored data banks come mostly from the prefrontal
cortex (herein called the decision maker). Are you with me thus far? Okay, so
the color infuses into the brain images and voila! According to these experts,
we have identified the active components of the brain. Repeat that experiment
many times and average out the response, create a Bayesian apriori bank of
information and then create a p-value of 0.05 or 95% Confidently Bounded Interval
(CI) as the threshold and if the firing neurons cross that threshold, the
computer registers the data with plethoric hues. The stronger the p-value of 0.05 or 95% Confidently Bounded Interval (CI) as the threshold and if the firing neurons cross that threshold, the computer registers the data with plethoric hues. The stronger the p-value less than 0.04, 0.03, 0.02…0.0001 the higher the coloring labels just like the weather maps go from a light green for drizzle to a magenta within red color for humongous storms based on radar reflectivity. Is that all good so far? Ok so now let us look at how that activity is determined within the fMRI.
fMRI machines use something called BOLD or Blood Oxygen Level
Dependent a mechanism promoted by Seiji Ogawa. The idea being that brain
activity would require nutrients in the form of sugar and that will necessitate
need for oxygen to create the ATP (Adenosine Triphosphate) to liberate a
phosphate group to create energy for the brain cellular activity. And that is
how the fMRI was born. Two inherent conflicts arise when viewed from this
simplistic viewpoint:
One, if energy is used immediately for
the activity, then there should be an immediate deficit recorded in the
deoxyhemoglobin (hemoglobin that binds with oxygen and the “de” represents
removal of oxygen for delivery to tissues) and
Two, the BOLD activity takes place about 5 seconds after the evoked
stimulation and response (Why the delay-or representation of a flat line on the BOLD scale?).
The fMRI machines construct the brain image into 3-D pixels
called voxels, (Consider “Volumized-pixels”)
each about 5cm in size. The complete activity of the brain at any instant can
be recorded using a 3-D grid of 60 x 60 x 30 voxels. These machines register
information every second of the 3 minute session creating 30 million plus data
points. Indeed when we look at a picture, any picture placed in front of us
many thoughts creep into our minds and through the act of parallel processing
the information presented and that bound within our personal experiential data
banks, the individual response is elicited.
Now let me launch into the SPECT scan rage of the season
that keeps giving us brain images like laundry detergent boxes of different
colors. What exactly is SPECT? It is imaging of a single photon emission
released by the neuron due to increased oxygen entry within the cell. This
single photonic emission when pulsed together via a computer program and again
based on the threshold of an arbitrarily placed p-value gives us beautiful red,
green, blue and magenta images of areas differentiated by those areas depleted
in oxygenation activity and those turgid with a surfeit of the same element.
The difference is that the fMRI is a computerized tomographic image (slices put
together by a computer algorithm) versus SPECT, which is a 3 dimensional planar
radio-nucleotide imaging format. (Radio-nucleotide is essentially a material that
joins with a specific cellular target (oxygen in this case) and emits a gamma
emission (radiation) for the detector to detect and the computer to assimilate into a 3-D
image). The difference is obvious but the human endeavored legion of stories as
to “cause and effect” multiply exponentially. Some go as far as delineating
sexual, aggressive, criminal, sociopathic behaviors on such images and the
laity buys it “hook, line and sinker” as the next greatest thing since sliced
bread.
Now let us take this whole house of cards worth of
information and stoke the beast of Artificial Intelligence. All I can say is it
will take a long time to match the equivalence of the human brain. I say that
because of the data from Harris Georgiou a neuroscientist who in using the
voxels concept in fMRI has determined: “that
a typical voxel corresponds to roughly three million neurons, each with several
thousand connections with its neighbors. However, the current state-of-the-art
neuromorphic chips contain a million artificial neurons each with only 256
connections.” Thus the parallel function within the brain occurs at a
much higher structural and functional level given that there are, as previously
mentioned, our brains are operating about 50 tasks at once. Imagine the
division of labor, concept enhancement or reduction, sensing, feeling,
importing and exporting information, comprehension etc. the task of the brain
is immense and it’s power needs are a mere 20 watts! Now that is some Bang for
the Buck!
This study in Science by Hilbert and Lopez tells us of our
accomplishments and what might remain under the dusty future (http://www.sciencemag.org/content/332/6025/60
) concludes: We estimated the world’s
technological capacity to store, communicate, and compute information, tracking
60 analog and digital technologies during the period from 1986 to 2007. In
2007, humankind was able to store 2.9 × 1020 optimally compressed bytes,
communicate almost 2 × 1021 bytes, and carry out 6.4 × 1018 instructions
per second on general-purpose computers. General-purpose computing capacity
grew at an annual rate of 58%. The world’s capacity for bidirectional
telecommunication grew at 28% per year, closely followed by the increase in
globally stored information (23%). Humankind’s capacity for unidirectional
information diffusion through broadcasting channels has experienced
comparatively modest annual growth (6%). Telecommunication has been dominated
by digital technologies since 1990 (99.9% in digital format in 2007), and the
majority of our technological memory has been in digital format since the early
2000s (94% digital in 2007). So if one were to calculate the information
storage within the brain given that we have about 100 billion neurons each and
each of the neurons has a minimum of 1000 to 10,000 connection which translates
to 100 trillion to 1 quadrillion data points or between 100-1000 terabytes of
information yield. Due to the continuous increase in actual brain storage of
information, now the estimates have reached a staggering 2.5 petabytes or 2500 terabytes. That is
some order of magnitude one would say! Compile the memory bank to the
connectivity (or "Connectome" as the experts call it to look super-intelligent)
and you have a ginormous maze of data flow!
After all that, here is the crux of the neuro-babble matter.
AI is a long ways away from mimicking the human brain. True, that IBM’s “Big
Blue” can beat Kasparov in the game of chess and Watson can beat the Jeopardy
champion, but can it tell the difference of an infant’s crying need between a
diaper change and hunger or a cuddle, like a mother can? Didn’t think so! So
those stories of computers becoming doctors are highly exaggerated in my
opinion. Maybe someday we as humans will have computer chips installed to
enhance our memories, cognitive skills etc. but even then the primary base of
operation will remain with the human brain – add to, not in lieu of.
So in the end, thus far we can make lots of assumptions
about what the brain is doing, but really we don’t have a clue. We appear smart
with all the purported calculations and the probability assumptions and yet the
main ingredient of “humanness” remains lacking from that large metal box filled
with CPUs.
The quantification of uncertainty is a philosophical
conundrum as much as it is a mathematical maze. Yet through it all in keeping
the scion of truth from imploding, ambiguity has to be given its due share in the
process of scientific discovery. One without the other implies abject
ignorance.
Medical Research once considered the paragon of statistical
research methodology is creaking under the weight of this mathematical jargon.
The literature is getting burdened with “ambiguity proofed” positive results
through statistical fiat that imply little progress in understanding. Thus ensuring
that the “native hue of resolution is sicklied o’er with the pale cast of
thought…”
On a more human level… “Have a great day!” (Let Watson
figure that out!)
No comments:
Post a Comment