The Essential Ingredient for General Intelligence
An essential ingredient to achieve general intelligence is that it must be grown into this world and not manufactured into this world. This requirement captures the unavoidable difference between the subjective and the objective. A general intelligence learns about this world, it isn’t programmed by another agent.
There are fundamental limits to instruction that a teacher cannot express or a student can’t digest. A general intelligence must learn for itself how to learn. Just as they say that a zombie isn’t conscious, an intelligence that cannot learn is not a general intelligence. General intelligence exists in a spectrum of forms that we can recognize in living things. Just as there is a spectrum of kinds of consciousness, there’s a spectrum of kinds of general intelligence.
The kind of general intelligence that we find in humans is a unique kind that originates from our intrinsic motivations to share experiences. It is the kind that demands that we conjure up explanations of our experiences. We must be able to express to another mind what we think. This is why the learning algorithms that we find in deep learning are insufficient in achieving general intelligence. Although it is true that all general intelligence must involve curve fitting algorithms.
Making the leap from mindless curve-fitting-based learning into a more purposeful abstraction-based learning requires competence in language. All technologies are based on the recombination of complex things, it is language that makes this possible in the virtual realm. But how do we learn how to understand and manipulate language? We do so in a way that surprisingly is the same as curve-fitting! Just as we intuitively understand the affordances that we find in physical reality, we learn the affordances present in language.
The complexity in our thought patterns are a consequence of the complexity of the kinds of languages that we have learned. We think because we have language and think differently because we are fluent in different languages.
When I speak of language, I do not confine myself to just the verbal form. Drawing is a language. People who claim not to know how to draw are people who have never learned the language of drawing. Yet people assume that drawing is an innate talent. We are never surprised to know that we don’t know a foreign language. People who do not know how to draw simply have not immersed themselves enough in the language of drawing. Yet, most people assume drawing is not a foreign language. It is foreign if you’ve never done it.
It’s fascinating that the study of paintings throughout the centuries reveals a progression of our understanding of what we see. It was only in the 1400s that perspective drawing appeared in art. How could humans not see what is obvious to us today? Of course, humans have always seen perspective. What was missing was a way to express perspective in a 2D plane. Humans had to wait for the Italian Renaissance to have that invented!
Language however is a pervasive feature of the biological world. It is not an exclusive technology that is confined to humans. The entire mechanism of life, DNA, and the cell that interprets its coding is a language system. The creativity and innovation that we find in life are a consequence of different interpretations of DNA. Life is robust because of DNA. Life has variety because DNA is interpreted based on context.
This begs the question, what kind of algorithm is interpreting DNA? It’s the same kind of general intelligence that also interprets the world.
It’s turtles all the way down. It’s this kind of inversion that is often referred to as strange. Damasio writes about it in his book ‘The Strange Order of Things’. Living things are collective things. Complex multicellular organisms involve individual cells that must be capable of coordination of the entire organism’s body. This involves not just its movement but also how it comes into being.
The open question is how does this coordination come about? One could make a strong argument that a language has evolved between the cells of an organism that makes possible its efficient coordination.
The architectural distinction between a living organism and a man-made machine is that the former is massively decentralized. When an organism is exposed to an infection or is wounded, decentralized processes are triggered to fight the infection or to repair the wound.
This kind of robustness is simply unavailable in current human technology. The most advanced kind of technology will look to us as being biological. In fact, current medical practice is effective only because human bodies repair themselves. Human technologies are incapable of repairing themselves. They are not designed that way. That’s because they were manufactured based on specification. They were not grown into this world.
