[FoRK] Super-Intelligent Humans Are Coming
Stephen D. Williams
sdw at lig.net
Sun Oct 26 22:32:52 PDT 2014
On 10/26/14, 11:27 AM, J. Andrew Rogers wrote:
>> On Oct 25, 2014, at 12:11 AM, Stephen D. Williams <sdw at lig.net> wrote:
>> Intelligence may have a rigorous definition, but the measure of intelligence sparsely covers the full ground truth.
>> It may be that EQ is simply an application of plain old intelligence, or it may be that it partially relies on specialized subsystems that may, like other parts of the brain, be more capable in some people compared to others. Given that we know about specialized units like mirror neurons, and that we have, according to some scientists explanation, a whole processing system for social and emotional processing that is exo-consciousness in operation, it may be that some are more naturally capable than others.
> Tangentially, we know almost nothing about mirror neurons, and certainly not what you are implying. Most things ascribed to them in the popular media is pseudo-science. Like EQ, it enables a pleasing Gladwell-ian narrative with a truthiness that real science fails to deliver.
Another example of specialized subsystems, one that we do seem to have very specific proof for, are grid cells:
> It is possible that some people are limited in their ability to acquire data for social and emotional reasoning in some fashion e.g. biological cues that operate outside higher sensory function. That’s not related to intelligence, it is a defective peripheral. It is far more likely that people have no inclination or have not trained in developing those domain skills, perhaps requiring relatively higher (but by no means difficult) cognitive effort on their part.
It has been widely believed by certain scientists that we have a "limbic brain" that is somewhat separate from our higher level
brain. Even if it is wrong, it seems explanatory. The processing ascribed to it is thought variously to handle all emotion,
tracking of "friendship balances", motivation, etc. It is said to be acutely aware of subtle signals in determining what others are
thinking and feeling, relaying only high level results to our more conscious brains. It seems plausible. In any case, most
emotional queue processing and reaction seems to be subconscious and automatic. We essentially keep score with friends and lovers
effortlessly compared to math. It is some kind of neural net weight adjusting system that maps to specific individuals in some
nuanced way. Perhaps it also has an analogy based reasoning system.
Some things very definitely happen outside our logic processing frontal lobes. Some of that is purely electronic preprocessing,
like our visual input layers that identify edges at certain angles. But it may be that we mostly use the same machinery to do most
things. On the other hand, we do have a fair amount of apparent specialization. I'm fairly surprised that we have something as
specific as grid cells in a particular location for such an important purpose. It hints at more specialization. Clearly we have
one or more generalization layers on top of these things too.
> Effective emotional and social processing is computationally trivial. The state-of-the-art for computers discerning and being able to manipulate human emotional state and social interaction, both human-to-machine and human-to-human, has progressed at an astonishing rate. Not only is there no ineffable magic behind human emotion and social processing, it is a pretty boring domain from a cognitive standpoint. Naturally humans take it quite seriously but it is neither difficult nor interesting for even not-very-intelligent machines. Any human could learn the nuances of the necessary theater and apply it without special equipment.
> As a more general observation, the “EQ for intelligence” argument is the old “qualia for intelligence” argument in different clothes.
It seems qualitatively different than the "qualia" navel gazing. Most of that seems like confusion and various types of false
logical arguments based on language rather than clear thinking. The fact that a JVM represents a program differently than Python or
a C++ compiler is mostly irrelevant to whether they are running the same algorithms, inputs, and outputs. Assume that every neural
net is different just as every forest is different. So what?
More information about the FoRK