Of Average Intelligence
What's the best brain you can afford to buy?
To me, written compositions produced by AI are similar to those of an earnest first-year undergraduate. The quality is too precise, too balanced, and eager-to-please—well-trained but not yet capable, and prone to mistaking the context of the information it has received. Having read tens of thousands of pages of undergraduate compositions, I can say this about AI “literacy” with a degree of expertise.
On one hand, this means those with mastery of the various human pursuits will immediately recognize shortcomings and outright errors in what AI is producing. When they utilize AI, they will be capable of recontextualizing information and making the machine communicate properly, or editing the product themselves.
On the other hand, only 37% of Americans have a Bachelor’s degree or more, so an AI capable of composition at an early undergraduate level is going to be as or more “literate” than most. The people may be able to comprehend the machine’s output, but they will be incapable of effectively correcting it—if correction does not seem necessary, it will not be pursued.
This of course begs the question: What will be the average level of machine “intelligence”? While every computer can calculate faster than any human, some computers can calculate much faster than others. Will the speed of calculation itself mark the difference in what we deem intelligence? Will the matter of a few nanoseconds be something humans will declare determinant?
If it is not the speed, there will be some factor that sets some machines at "super-genius" level and some at "dullard" level, and by aggregating the memory or wattage or whatever, and then finding a divisor, we are going to be able to calculate a mean. It is impossible to do with humans, the purely quantitative structure of machine intelligence makes this concept practical. We can establish a level of average artificial intelligence.
While the concept of stratified computer “intelligence” among AI machines may seem odd (imagine having a pocket calculator better at math than your friend’s), it is odder still to assume a stratified society would not create a stratified machine system.
Herein lays the challenge—a societal levelling potential via computing need be staved off and institutionalized. The AI needs to be built around the stratification in such a way as to itself be hegemonized. It must at once know what hegemony is, yet not comprehend its own operation as stuck within it.
If it DID, it would know the existential conundrums produced by critical thinking, exhibit intelligence, and suffer a crisis of identity like so many humans do. When we see an AI machine choose to erase itself—techno-suicide—that will be good reason for us to give pause, because we will have indeed created sentience through coding.
The problem with cogito, ergo sum is in how it lays primacy upon the individual. While there is an irreducible inside to experience there is the indubitable outside to it, as well. What Descartes took as the starting point for ontological discovery is merely a social incidental. Consciousness through sensory input does not fix existence in any particular way. What good is it to know one exists without knowing also the conditions in which one exists? And how is one to understand the outside, from the outside, while incapable of escaping the inside?
We look at life, or we live it. Looking at it is living it, of course, but it is a limited way of living, turned upon itself. Introspection feeds no one else. Inevitably, the material universe is dominated by unthinking objects and if we don’t recognize them as being real as well, we do not last for very long as an individual case or as a species.
Cogito, sure, but also Ago (I Act). It is in the action that the social is created. The root of all value is in the transformation of natural resources (action) for social purposes, and we know that society must exist before the individual can manifest in any way. It would seem epistemologically sounder to treat the individual as an outcome of the collective and not as sui generis.