Page 139 - Biomimetics : Biologically Inspired Technologies

Bar-Cohen : Biomimetics: Biologically Inspired Technologies DK3163_c003 Final Proof page 125 21.9.2005 11:41pm

Mechanization of Cognition 125

I and Calculus II). Total rebuilding of a lexicon typically only occurs in the event of trauma (e.g.,
stroke), where the entire information input environment to the lexicon has dramatically changed.
Total rebuilding takes weeks and requires lots of practice with the new symbols. This is why
recovery of function after a stroke takes so long and why intensive physical and mental therapy
based upon practice and use is so important. Aspects of childhood development are being recap-
itulated on an abbreviated schedule.
Lexicons also slowly incorporate replenishment neurons into existing symbol representations
that are used. As with forgetting of knowledge; long-disused symbols eventually have their sets of
representing neurons redeployed (see below) or eroded beyond functionality. A person who spoke
French when he was a child, but who has not used French at all for 40 years, will likely have many
of the French word representation symbols eroded beyond recovery.
The only instance of deliberate fast cognitive knowledge erasure in human cortex is redeploy-
ment, where a source symbol in a lexicon, which used to be linked to a particular set of target
symbols in other lexicons, suddenly has an entirely new ensemble of links to new target symbols
arise for it, and these new links persist (and the old ones are disused). For example, when we move
to a new home, it may be necessary to learn that the alarm clock is now on the left side of the bed,
not the right. What happens in this instance is that the sets of transponder neurons representing the
involved source symbol have a finite limit to the number of highly strengthened synapses that they
can have at any time (this probably has to do with a total individual cellular limit on synthesis of
certain consumable biochemicals — the critical ones of which are produced only in the neuron’s
soma and dendrites, where the ribosomes reside). (Note: The ultimate limit to knowledge storage
capacity is not synapses; it is the number of strengthened synapses that each transponder neuron can
support at one time. There are probably people [e.g., perhaps the author] who have spent their entire
lives studying and who reached this capacity limit long ago.) As the transponder neuron synapses
implementing the many new links are learned and strengthened, many of the old, now unused, links
must be immediately sacrificed (their synapses shrivel to the unstrengthened state). Within a few
weeks, we instinctively reach left. The old knowledge has been effectively erased. The synapses of
many of the old knowledge links have shriveled (but not all of them; some remnant knowledge links
often remain — which you can experience by revisiting one of your old haunts and trying to carry
out formerly familiar patterns; like skipping down stairs at a childhood residence). Fragments of
your former knowledge will still be there.
Redeployment is a critical cognitive capability that allows us to adapt to environmental change
quickly. It is also hypothesized to be the only mechanism of deliberate forgetting in cognition.
Finally, it is important to note that any global theory of human cerebral cortex and thalamus is
bound to be vastly oversimplified. For example, it is well known (Paxinos and Mai, 2004) that
different areas of cortex have some Layers dramatically attenuated (e.g., Layer IV in certain areas
of frontal cortex). Others have Layers that are dramatically elaborated (e.g., in primary visual
cortex, Layer IV becomes tripartite). These local modifications almost certainly must have signifi-
cant meaning for the nuances of function. However, the theory proposes that these are all relatively
small variations of the same overall grand theme.
The central notion of the theory: that cognition, that greatest engine of animal ennoblement, is
universally mechanized by one information processing operation (confabulation) employing a
single form of knowledge (antecedent support), with each singular conclusion reached launching
an associated set of action commands; seems to me to now be secure. The concreteness and
specificity of this theory guarantees that it is testable.

ACKNOWLEDGEMENTS

Thanks to Fair Isaac Corporation for long-term research support and to Kate Mark for help with
the manuscript. Thanks to Robert F. Means, Syrus Nemat-Nasser, and Luke Barrington of my