Musings on Hofstadter’s Notion of ANALOGY as the ESSENCE of Human Intelligence
Timothy Budd - Published on Jan 26, 2013
Causal dynamical triangulation (CDT) is a lattice model of quantum gravity. In two space-time dimensions (instead of the four we live in) it gives us an exactly solvable model of quantum geometry, and therefore a nice toy model to play around with. The weird sea creatures in this short movie are in fact randomly generated two dimensional universes. To be precise, they are random triangulations appearing in the disk amplitudes (cup functions) in generalized CDT (check http://arxiv.org/abs/1302.1763). They all consist of a number of triangles of the order of 10000 and a boundary length between 50 and 800. The coloured bands represent constant time slices, which in this case have constant distance to the "Big Bang" (depicted by a black ball). All images are produced with the help of Mathematica®. See my website for more details: http://www.nbi.dk/~budd/
Could a machine ever create this on its own?
Hofstadter suggests that the key to human intelligence is our ability to create and communicate analogous meanings of one thing for another using words. Machines that think like humans would, therefore, have to be able to analogize to communicate with humans about the primal stuff of reality.
What is this primal stuff anyway, this play dough of life that makes us human?And how are we able to manipulate it effortlessly to create new stuff from existing stuff (innovation)? Douglas Hofstadter argues that analogy is the key—the ability to recognize and manipulate the sameness of different kinds of experienced reality to create a new sameness that we name with the word “original.”
The problem with words is that we have to use words to define what words mean, and words are imprecise because we need words to describe what me mean by words and so on.
This is the self-referential problem Hofstadter explored in his book, Tangled Loops and more in his recent work. This problem is also at the heart of Wittgenstein’s rejection of Western philosophy for being a fool’s errand for its failure to progress beyond its development 2000 years ago.
Here is an interlude on the definition of key words used as tools to explore the concept of intelligence for human or machine.
According to the Cambridge Online Dictionary for American English:
ANALOGY – A comparison of the features or qualities of two different things to show their similarities. He was explaining that the mind has no form and is invisible, and that a useful analogy is of the mind being like the sky.
SIMILE – an expression including the words “like” or “as” to compare one thing with another. She was good at capturing a passing figure with an apt simile or comparison. [e.g., she caught his attention by smiling like a woman looking for love.]
Here is my take.
I use Thing here as a placeholder for any object or concept, and their action. So, the essence of experienced Reality consists of Things and their movement. I argue that Things are convenient delusions of absolute Reality as a necessary constraint, and also because all matter is ineffable (i.e., profoundly inaccessible) energy, and because the fabric of space-time is an ineffable entity too (your GPS needs Einstein’s relativity equations for it to work!). All of this ineffability makes the true nature of Things as objects and their motion inaccessible to us, even though our experienced reality—the model in our brain—works reliably well for our view of physical reality. If the true nature of Reality is inaccessible to us, then Truth must also be inaccessible to us, as Wittgenstein also claims, at least according to the failure of Western philosophical tenets.
How, then, can we know whether it is true or not that machines can think, or analogize, like humans? The dilemma is that we cannot know unless we develop and experience it. If we are successful, can we control it or will it control us? Who will be the better analogizer? Such is the power of analogizing.
The only thing that humans or “smart” machines are able to experience is Reality as presented through crude senses (what we call reality) by a pattern seeking-and-making “brain,” which creates a delusional sense of order that allows self-modification. This delusion, the promise of there being an absolute reality to know, keeps humans sane and wanting to know more. This is not a bad way to design a universe. It keeps everything going.
Somehow, we create new patterns of order out of existing patterns by analogy, which, according to Hofstadter, is the key to our intelligence. It is as if deep down inside of us we have innate knowledge of the primal Thingness of absolute reality, which we are constantly trying to represent better and better by analogizing. It is this ability, this drive, that gives us the trait of intelligence; no machine is yet close to having this capability. At present smart machines are big-data analytic pattern-finding machines based on crunching data, not meaning-finding machines. Google algorithms merely search out patterns by mimicking our search behavior where “meaning” resides and/or correlating data from human actions with big data analytics. Mimicking patterns and correlating data are NOT intelligent thinking, not analogizing.
However, for the sake of argument, let’s assume that self-modifying learning machines could analogize like humans? Is there a primal Thingness or Beingness that makes analogizing possible for machines as well as for humans? My guess is that no discrete state machine based on absolute “yes or no” logic could ever “contemplate” this question because it cannot allow for uncertainty. Perhaps a quantum computer with its “fuzzy logic” of simultaneous possible states of existence could contemplate this question.
Here is my science take on the power of analogy. All Things—of every kind and sub-kind seen as a NOUN or a VERB—seem to amount to exchanges of energy from one form to another. The potential energy of a rock on a ledge changing into kinetic (i.e., “moving”) energy as the rock falls to the ground is a familiar example. If the total energy in the universe is constant (cannot be created or destroyed), as it seems to be, then the totality of life and existence in the universe (the self-aware parts and not) simply amount to the act of participating in the incessant flow of what we call “energy.”
This can be seen as a galactic energy-exchange play of giving and getting, attracting and repulsing, creating and producing, selling and using—some by choice and some not—from the smallest to the largest scale.
Seeing this energy exchange through the physics laws of thermodynamics reminds us that energy and information are intimately related to each other through entropy (disorder); it takes energy to create and maintain patterns of order (e.g., life) over disorder (e.g., death).
So, this fundamental energy exchange activity can be seen as an exchange of patterns of information order to which we attach the word meaning, thereby closing the loop back to our experienced delusion of Reality defined by those meanings. Analogies seem to be fundamental carriers of general meanings, which allow us humans to connect, mix, extend, and assemble expressions of general meanings to construct our delusions of Reality. This makes the possibilities for human life seem virtually endless.
What possibilities for artificial life will self-evolving machines construct from the general meanings they create from analytic analogies for their delusions of Reality?
This play of human and artificial life takes place on the same four-dimensional stage of space-time Reality, which some physicists believe was created de novo by a quantum-driven thermodynamics process, and which also conveniently leads mathematically to Einstein’s space-time curvature equations for gravity. The everyday space-and-time fabric we experience as fixed is actually “flexible” (the near light-speed motion of mass contracts space and slows time), and was/is possibly created from the thermodynamic constraint on Reality known as The Conservation of Energy. [See Origins of Space-Time: http://www.nature.com/news/theoretical-physics-the-origins-of-space-and-time-1.13613 ]
The stage and the play are set for perhaps the next great paradigm shift drama in the evolution of life. Our little earth is humming with Information pattern exchange processes at all levels. There are only two primary actors, humans and machines; one, the antagonist and one the protagonist. The supporting characters are in place. The script is being written now.
Visionary technologist Bill Joy, who was instrumental in making the Internet use-able through fundamental software developments he created, has worried for 15 years that we may be nearing the tipping point of creating smart technology from a combination of GNR (genetics, nanotechnology, and robotics) technology that can rapidly self-evolve, and which we may not be able to control. His long article of rational alarm was posted in Wired Magazine in 2000, but has not received sustained national attention.
At the heart of this issue may be the philosophical question of whether a “machine” of biological and/or robotic type can analogize like a human, and create a different set of meanings and delusions than ours from the same primal stuff of the universe.