Understanding Savant Syndrome and Neural Intelligence (2013) | ||||||||||||||||||||||||
A
recent Popular
Science article discusses the phenomenon of acquired savant syndrome.
Essentially, it discusses the stories of normal everyday people who
happen to suffer acute head trauma and then instantaneously acquire
astounding musical, creative, or mathematical talents never before
exhibited. Think Rainman
(1988), starring Dustin Hoffman who plays a highly austistic savant
individual with amazing mathematical abilities.
The autistic character in Rainmain was congenitally autistic
savant; acquired savant syndrome can technically happen to anybody.
Implication: should a hard-pressed student need to pass a very
difficult upcoming math exam or music recital, there is no need to study
or practice hard – just have a partner strike specific head regions. This
is technically neurologically plausible.
Working with Alzheimer’s dementia patients, one can easily trace
back patient histories to find previously inhibited, distrustful, and
taciturn individuals rapidly growing more outgoing, extroverted, and
perhaps happier seeming as the disorder progresses and vice versa.
According to several neuroscientists as mentioned in the article
(e.g. Bruce Miller, UCSF), dementia patients and savants share
similarities in abnormal activity in their left anterior temporal lobes.
Wong
& Gallate (2012) show
that this region appears to process general semantic information biased
with social interactive interpretations.
Figure 1. Temporal lobe location. The front, or face, is towards the left. The anterior region of the temporal lobe is the left-most portion. While
nobody is seriously advocating self-inflicted or assisted-inflicted
permanent lesions of precise brain regions – e.g. designer brain lesions
– the technology already does exist for temporary lesion-like effects
(i.e. Transcranial Magnetic Stimulation). For research purposes, scientists may use low levels of
electric currents and magnetic fields to disrupt targeted brain regions
and simulate lesions. Generating
temporary lesion-like effects on the left anterior temporal lobe could
technically induce temporary acquired savant syndrome.
While
applying such research technologies for obvious commercial purposes (e.g.
“Transcranial Cram Study Centers”) would be anathema to any sane
neuroscience researcher, serendipitously exploiting head-trauma accidents
is done wherever possible. Phineas
Gage as an early instance – who in 1848 survived having his own
tamping iron penetrating through his brain – apparently began to make
public appearances afterwards to capitalize on his fame.
The fact that there could conceivably be a market for this sort of
brain and behavioral alteration is a social commentary on our culture. What
does this say about the brain? Logically, losing a portion of the brain and surviving
indicates that not all brain regions are immediately critical to bodily
survival. Logically, losing a
portion of the brain and qualitatively acquiring some novel or hidden and
latent skill indicates brain either brain regions compete for activity
(i.e. left anterior temporal lobe does not do music or math but
consistently out-competes the brain region that does) or some regions are
inhibitors of others (e.g. left anterior temporal lobe actively shuts down
the music and math brain region). Applying
the anthropic
principle and evolution, if the music and math region of the brain –
and ensuing music and math abilities – were critically important to
bodily survival, then there should be no left anterior temporal lobe.
It would have evolved away. So
why do we have one? One
can only surmise that since the left anterior temporal lobe is involved
with social interpretation and interaction – as implied by Wong &
Gallate (2012) and by the Alzheimer’s patient anecdote – then social
interpretation and interaction may be more probabilistically critical to
bodily survival than the math and music regions.
Therefore, one must be extremely careful not just in technically
disabling the precise anterior temporal lobe for that boost of music and
math, but also in wisely deciding whether the ability to do so under
voluntary control is desirable. To
place this under a more objective model, there are certain text and naming
matching mechanical algorithms existing for intent detection.
A simple one uses word presence and word absence differencing
kernels. For example, in a
two-word vocabulary consisting of “Uncle” and “Bob” one must
distinguish between an Uncle Bob and a Cousin Bob. In normal English parlance, we would call Uncle Bob, “Uncle
Bob” or “Uncle.” We
would call cousin Bob simply “Bob.”
Below are the word vectors for Uncle Bob and Cousin Bob,
respectively.
In
a given test phrase, similarity points are added for every word present in
common. In a given test
phrase, similarity points are added for every word absent in common.
In machine learning parlance, this is akin to including a
feature’s complement. In
neuroscience parlance, this is akin to an inhibitory connection.
So
if we heard the phrase, “Bob,” it matches Uncle Bob once (for the
“Bob”) and misses Uncle Bob once (since it is missing “Uncle”).
It matches Cousin Bob twice (once for “Bob” and once for the
absence of “Uncle”). Hearing
“Bob,” this algorithm detects the intent to call Cousin Bob rather
than Uncle Bob. If we altered
the algorithm to ignore the inhibition – say to correctly fix another
instance where the inhibition was non-optimal – then “Bob” would
equally match both Uncle Bob and Cousin Bob.
It is a matter of how likely the inhibition is helpful vs. how
likely it is harmful in correctly identifying intent.
See below.
In
this table, we compare “Bob” vs. Uncle Bob and Cousin Bob.
There are four word columns: two for the presence of Uncle and Bob,
and two for the absence of Uncle and Bob.
If a word is present, it gets a “1” under the presence column
and a “0” in the absence vector and vice versa.
The sum across any row is always 2.
Comparing
“Bob” vs. Uncle Bob and Cousin Bob, we can see that it perfectly
matches Cousin Bob but poorly matches Uncle Bob.
This is, by the way, the core matching algorithm for a Fuzzy ART or
ARTMAP neural network.
In
conclusion, if a hard-pressed student wishes to pass a math or music
course for which he is woefully unprepared, he could conceivably choose to
visit a future “Transcranial Cram Study Center.” If the hard-pressed student ever wishes to succeed in perhaps
anything else is life, up to and including getting a job, getting a
girl-friend, or getting married, or starting a business, then he might be
best advised to look elsewhere for answers.
|