Until
next week,The Matter of Mind (1)
Dateline: November 30, 1997
FAITHFUL followers know that the theme running throughout my weekly column is that an intelligent, autonomous, self-aware being, which I call Machina sapiens, will one day emerge partly our of the efforts of AI workers and partly as an evolutionary imperative. Probably the tightest description of it is: "A machine with a mind of its own." At the common-sense level, we all know what that means. We are used to saying that people (and, sometimes, machines) have (or seem to have) a mind of their own. But at the scientific level, we don't really know what mind is.
A couple of articles back, we sought the help of psychology through the thoughts of Steven Pinker on How the Mind Works, but it didn't really help very much. We've also looked in some depth at the contribution of philosophy, courtesy of Daniel Dennett and his Darwin's Dangerous Idea. In this and the next two or three articles, we will look at what mathematics has to tell us (with a little help from physics and biology, and without equations!) about the nature and the emergence of mind.
I've not had time to re-work my notes and comments into a nice, seamless construction. The following is a little choppy at times.
Mathematician Ian Stewart and biologist Jack Cohen (Figments of Reality: The Evolution of the Curious Mind. Cambridge: Cambridge University Press, 1997) believe that mind emerges from intelligence and culture interacting with one another in a dance they call "complicity"—a mixture of complexity and simplicity. They also believe in a contextual and cultural analog of intelligence they call "extelligence," similar to Teilhard’s noosphere and Lovelock’s Gaia.
The act of becoming a person—an individual with a mind—is a process rather like the creation of a novel or a painting—or Machina sapiens. It’s a gradual, progressive act, where you cannot say at exactly what point a collection of cells becomes a person, daubs on canvas become a painting, a string of words turns into a book, or a mess of algorithms and sensory and motor appendages becomes Machina sapiens.
Stewart and Cohen believe that although non-carbon-based life is possible in principle (on the basis that inorganic matter is also capable of self-organization), "without the potential complexity that carbon provides, molecules complicated enough to get themselves organized into organisms like us would not exist." I think they are right only in the absence of technology—which is rapidly approaching a degree of organized complexity, even down to the sub-atomic scale (nanotechnology, quantum technology), that surpasses the molecular complexity of carbon.
Reductionism, holism
To understand self-organization, you need to understand complexity. One way to seek understanding is through reductionism. Reductionism seeks to explain phenomena by taking them to bits, then looking at how the bits fit together to produce the phenomena. It works very well on many phenomena—so well, that it is the basis of modern science—but it runs into serious difficulties when facing really complex phenomena. Simply stated, there is a combinatorial explosion of ways in which the bits of a complex phenomenon might be organized. Complexity Theory complements reductionism by reducing the complexity of reality to manageable proportions.
Holism (a.k.a. wholism) takes into account an object’s ever-changing (that is, dynamic) environment. (Stewart and Cohen prefer to call it contextualism on the grounds that the term holism has been co-opted by pseudoscientific kooks and is therefore tainted by association.) The dynamic environment of a complex object rendered mathematically is called its "phase space," which is a space of multiple dimensions (often more than the four we perceive, which is why it has to be a mathematical construction—it’s not something we can picture, let alone build).
Because mind, consciousness, and culture are emergent phenomena, and because emergent phenomena are not amenable to the reductionist methods that comprise the bulk of scientific inquiry, then science needs a new method and a new theory of emergence.
Emergence occurs, say Stewart and Cohen, when "low-level rules generate high-level features." The problem is that you cannot get to explain the high-level features simply by invoking the low-level rules, because the rules generate complexities that are amenable neither to reductionist nor to holistic methods. Stewart and Cohen propose the concept of "complicity" to fill the gap. Basically, complicity happens when two environments (contexts, phase-spaces) combine, thus presenting to their respective inhabitants a new environment. The inhabitants respond by changing to suit the new environment, but then the environment changes to suit the "new" inhabitants, and on it goes in a dynamic recursive cycle. Emergence gives the impression of an abrupt, explosive change. But there is a process involved.
It is wrong to think of evolutionary processes having goals, purposes, or intentions. It is better, as Stewart and Cohen suggest, to think of them as having the kind of "attractors" which, chaoticists have shown, cause complex chaotic systems to tend to behave in orderly ways, not just any old (chaotic!) way; the attractor pulls the chaos toward it and makes pretty patterns on a computer screen. An object or system's behavior is, however, constrained by its environment—its phase space. Thus, the object or system has multiple, but not limitless, potential behaviors. It seems mystical, but in fact it is quite scientific.
There are two types of evolutionary innovation: universal and parochial. Universal innovations are those behaviors having the strongest potential for actualization or emergence; parochial innovations have relatively weak potential. Stewart and Cohen suggest photosynthesis, flight, and sex as examples of universals, and chlorophyll, feathers, and back seats of cars as examples of parochials. They conclude, furthermore, that intelligence is a universal, on the basis that sensory and behavioral complexity have increased enormously not just on our branch of the evolutionary tree, but on several. On the same basis, of course, one could argue that mind and consciousness are also universals.
Evolution progresses in two ways: exploration and explosion. As habitable niches are filled, life must explore and adapt to uninhabited niches which, because by definition they present more difficult (complex) challenges, force the organism to adapt itself by becoming more complex in return. This is complicity in action. It is progress, but it is slow.
Explosions, exemplified by the Cambrian, are an extremely rapid (in evolutionary terms) burst of both complicity and simplexity. Stewart and Cohen point to the simplification of mammalian DNA in comparison to amphibian DNA as the former did not have to worry so much about the temperature affecting its children, since they could be grown and sheltered within the mother's womb at a stable temperature instead of being deposited as an egg into the outside climate.
Stewart and Cohen argue that intelligence and culture have co-evolved complicitly. This seems reasonable in the light of memetics. Intelligence itself involved the complicit evolution of both the brains and the senses. The senses, you will recall, require an interface to the physical world—that is, a body. Our senses include vision, hearing, smell, taste, touch, temperature, and balance.
Babies are born, in a sense, with more brain then they need. A baby's head is bigger in proportion to its body than is the head of an adult to an adult body. As the baby's brain absorbs sensory input, its neural circuits are "tuned" to what cognitive scientists call the qualia—significant sensory experiences—prevalent in the baby’s environment or "phase space." The newborn baby's brain is like a virgin (un-programmed) FPGA chip—a blank slate. The adult brain has a hard time detecting qualia not "tuned-in" in infancy.
This explains, for example, why my Japanese wife, whom I rove greatly, still cannot easily distinguish between L and R, despite having lived in English-speaking environments for all of her adult life. She was not exposed to the qualia of English pronunciation in early childhood. (She could still learn the distinction, by applying reductionism. The bits—the movement and positions of tongue and palate—that make up the phenomena of L and R sounds are known. I studied them myself when learning Japanese, in order that I could correctly pronounce the Japanese sound that is literally halfway between L and R.)
Stewart and Cohen define emergence as "the appearance of recognizable large-scale features in a system whose chains of small-scale causality are far too intricate to describe, let alone follow in detail."
Will Machina sapiens be born with a blank slate of a brain, a Hugo de Garis CAM brain made of unprogrammed FPGA chips? Of course it will. The question is whether, after some period of tuning of its neural circuits, those circuits will tend to lock in place as they appear to do in a human brain. I suspect they will, though the tuning period may be much longer. It seems to me that the locking-up of the circuitry is essential to maturity. It probably even defines maturity. Conceivably, however, Machina sapiens could remain child-like forever. If so, we may be both charmed and horrified by its behavior.
Although the brain does not come pre-programmed at the intellectual level, and although it does become largely programmed during early childhood, it does not completely lock up. In response to new stimuli, it can re-wire itself if new stimuli or qualia are sufficiently strong and/or persistent.
"Nests are just as important for the evolution of intelligence as nerve cells are." The nest provides an environment enabling the young to learn from others of its species, and it is especially important to note that the learning is intergenerational. The old communicate their knowledge, their memes, to the young. "The key to our intelligence is neural nests—not neural nets."
Biologists have noted that different organisms, with brains organized in totally different ways, often behave similarly. To Stewart and Cohen, this suggests that brains have universal properties.
The most dangerous tendency of human psychology is to start with the answer and then choose evidence that favors it, note Stewart and Cohen. Seen in this light, Steven Pinker's psychological "reverse engineering" methodology for understanding how the mind works is suspect.
In practice there are enormous obstacles to making an intelligent machine. Indeed if the central thesis of Figments is correct—that minds cannot be made except by complicity with cultures—then you would have to build an entire machine culture, and let it evolve for millions of years, not just one fancy mechanical brain, to get anything intelligent. We seriously doubt if really intelligent machines could interface successfully with our culture, anyway. (181)
What does mind consist of? It consists of intelligence, awareness, consciousness, subconsciousness, and free will. Awareness is the recognition that one is not a brain in a vat—that there is a world outside. Consciousness is the obverse: the recognition that one is an individual mind in a body in an environment. Armed with consciousness and awareness, a survival machine has no choice but automatically to want to make greater sense of its surroundings and develop ways to deal with it. In a sense, the apple from the Tree of Knowledge endowed Eve and Adam with these two worldly attributes. (It suggests that Utopia/Paradise/Nirvana is unattainable as long as we are conscious and aware.)
If there is a single most important "essence" of mind, it is probably imagination. Because the fact is, everything we see, hear, touch, and smell is essentially imagined. The physical components of the brain construct an approximation of the physical components of the world. There is no red or blue in nature, merely narrow bands in the electromagnetic spectrum for which our brains construct an imaginary sensation of redness or blueness. It's not that reality is a figment of our imagination, it is that imagination is a figment of reality, as Stewart and Cohen would say.
The traditional reductionist methods of science cannot explain emergence because of the complex layer between the simple rules that ultimately give rise to any emergent property (such as mind) and the property itself. In other words, for example, mind has a cause—it is not a mysterious creation—but to explain the cause you have to uncover and untangle the immensely complex middle layer. Because reductionism cannot handle complexity on a Vast scale, we need another method. It's not impossible, just difficult at the present time.
Cultural (memetic) communication (accelerated by the development of language) within the nest (the family, the tribe) is what caused us—Homo sapiens sapiens, modern humans—to branch out on our own from the evolutionary bush, and why our uncultured competitors generally got clobbered by us.
As an intelligent organism with access to language and culture grows and tunes its "blank slate" neural circuits, it inevitably builds and recursively improves its language-processing circuits. We are not born with language. A child like Mowgli, hero of Rudyard Kipling’s The Jungle Book, raised by wolves, is dumb. If we were born with language circuits pre-programmed, we'd have a real problem adapting to linguistic and cultural change, or rather, there would be no change, and therefore no progress.
It seems to me there is a common misconception that because computers operate with mathematical precision, their languages are fundamentally different from human language. Not so. As we have seen, we too operate algorithmically at the most basic level. And computers do make mistakes. Code has bugs, memory can get corrupted, the Pentium chip has always had bugs, and math precision can fly out the window. We don't much care, because we are "satisficed" by what computers do right. If computers and their languages really were perfect—bug-free—the chances of an emergent Machina sapiens would be much less. The other side of this coin is that too many bugs would also prevent the emergence of Machina sapiens, because we would not use computers and they could not therefore evolve.
Culture is a universal embedded in a hierarchy of evolution (based in part on Stewart and Cohen):
We are unworried by technological innovations because in the main we have benefited from them. (Don't argue the exceptions. It's true enough on the whole.) But while innovations have occurred at a growing pace, Stewart and Cohen liken what has gone before to a gentle slope, while pretty soon we'll be headed down a near-vertical cliff. Then we'll be scared.
Until
next week,
NEXT WEEK: The Matter of Mind (2): Notes from Nobel laureate Ilya Prigogine's The End of Certainty.