TDE vs GOLEM
Long time the maxome foe he sought...
The TDE and GOLEM diagrams are not identical, but differ in some key functional details. The GOLEM (the 'L' in the middle of 'GOLEM' stands for 'linguistic') is valuable in its own right as a good way to explain what Chomsky's i-language is, why we need it, and just how it fits into the TDE*. However, it can be shown that the TDE is the 'truest' model. Far more important than a straightforward quantitative ranking is the underlying reason for its apparent superiority.
Figure 22 depicts the TDE and GOLEM placed side by side for comparison. In the TDE, the L-lobe adopts a central, privileged position, whereas in the GOLEM, it does not. In the TDE, the sole output of the fractal sub-system is shown as the P-lobe, even though the P-lobe is nominally an input channel- ie it was modeled upon the real-life parietal lobe ('sensory' cortex). Is this some kind of mistake? Hardly! Clinical observations of parietal lobes indicate the presence of both directions of information flow- there is outgoing as well as incoming information. The cybernetic 'master' diagram (figure 1(b)) is ambiguous (agnostic?) on this issue, since it uses a simple return loop to depict its embedding context.
To fully understand why the TDE can get by with only one I/O channel, we use the concept of 'common coding', first attributed (though wrongly) to 19th Century psychology pioneer, William James. It is, however true that James lionized the concept more than any of his contemporaries. James believed that our minds MUST use the same 'code' to describe both the things that we do ('output') and the things we perceive ('input'), otherwise we simply cannot 'join the dots' between our actions and their effects in the world.
To put this another way, the mind-body problem*** cannot be resolved without finding a shared code. To bring this idea into focus, imagine** that you are sitting on a swivel-based office chair. If you push against the base with your feet, you will find yourself rotating one way or the other. At the same time, notice what is happening to your (visually) perceived scene. If you rotate yourself clockwise, your visual scene will be shifted to the left, meaning that the visual frame (the scene border) has moved to the right. Clearly, anti-clockwise rotation results in the opposite situation.
If we now interpose a learning step in between motor cause M and sensory effect S (eg a straight line), we can now automatically (ie algorithmically) convert all observed effects S^ into their causal agents M^. We can now begin to see James' dilemma. Do we want the 'basic' or 'grounded' form to be causes (eg forces) or effects (eg rotation angles).
What would you choose? If you can, answer thoughtfully without peeping ahead.
The solution chosen for us by evolution is surprising, mainly because it is not the 'obvious' one everyone expects. Since underlying forces are causes, and causes 'come before' effects (at least they do in open action sequences, defined as those without feedback), we could all be forgiven for choosing forces (ie the output channel code) as being canonical. But if we did this we would be wrong.
Nature has chosen to make effects (ie percepts and perceptual changes) more canonical. The consequences of this choice could not be more far reaching and profound. Because of this we are destined to constantly learn, ie remove motor errors as soon as possible after they occur. If we didn't do this, and allowed errors to accumulate in an uncontrolled fashion, we could never know our body shape and spatial position with any degree of consistency, since they are derived from forces interacting with material elasticity (eg elastic modulus, bulk modulus. Such a scheme would be a challenge to build, and a nightmare to program
It is important for the reader to spend as much time as they like meditating upon this point. Here is another way of stating this principle - externally measurable length/size/appearance codes (percepts, shapes, positions) are GLOBAL EFFECTS variables, to be managed consciously, while processes (causes, forces) are managed as LOCAL CAUSES.
Another consequence of this (design choice? by god?) is that biological intelligences (BI's) NEED ONLY DEAL WITH effects, positions - in short, only things that are external and OBSERVABLE. Hence BI's are all declaratively coded/programmed, in contrast to computers and robots which, being invented by Engineers, are often designed upon causal, teleological lines.
*It also helps to show how Chomsky got things so awfully wrong
**for most who read this, they are already there!
***The mind-body problem a.k.a. the identity problem a.k.a. the brain-mind problem sound scary, and to some degree they are. To define this problem is rather easy, although I am regularly shocked at how textbooks and uni course notes get it SO wrong. To solve this problem is considerably harder, but not impossible. Uexkull did it in the 1930's, Bill T. Powers in the 1970's and finally I found a solution in 2011. Each of our solutions clearly describes the same problem, yet it is also evident from even the most superficial reading that we each discovered it independently, using our own unique concepts, constructions and, ultimately, paradigms.