Microsoft Word Kurzweil, Ray The Singularity Is Near doc

meaningful, non-random, but unpredictable
information 
needed to characterize a system or process. 
In Gell-Mann's concept, the AlC of a million-bit random string would be about a million bits long. So I am adding 
to Gell-Mann's AlC concept the idea of replacing each random string with a simple instruction to "put random bits" 
here. 
However, even this is not sufficient. Another issue is raised by strings of arbitrary data, such as names and phone 
numbers in a phone book, or periodic measurements of radiation levels or temperature. Such data is not random, and 
data-compression methods will only succeed in reducing it to a small degree. Yet it does not represent complexity as 
that term is generally understood. It is just data. So we need another simple instruction to "put arbitrary data sequence" 
here. 

---

To summarize my proposed measure of the complexity of a set of information, we first consider its AlC as Gell-
Mann has defined it. We then replace each random string with a simple instruction to insert a random string. We then 
do the same for arbitrary data strings. Now we have a measure of complexity that reasonably matches our intuition. 
It is a fair observation that paradigm shifts in an evolutionary process such as biology—and its continuation 
through technology—each represent an increase in complexity, as I have defined it above. For example, the evolution 
of DNA allowed for more complex organisms, whose biological information processes could be controlled by the 
DNA molecule's flexible data storage. The Cambrian explosion provided a stable set of animal body plans (in DNA), 
so that the evolutionary process could concentrate on more complex cerebral development. In technology, the 
invention of the computer provided a means for human civilization to store and manipulate ever more complex sets of 
information. The extensive interconnectedness of the Internet provides for even greater complexity. 
"Increasing complexity" on its own is not, however, the ultimate goal or end-product of these evolutionary 
processes. Evolution results in better answers, not necessarily more complicated ones. Sometimes a superior solution 
is a simpler one. So let's consider another concept: order. Order is not the same as the opposite of disorder. If disorder 
represents a random sequence of events, the opposite of disorder should be "not randomness." Information is a 
sequence of data that is meaningful in a process, such as the DNA code of an organism or the bits in a computer 
program. "Noise," on the other hand, is a random sequence. Noise is inherently unpredictable but carries no 
information. Information, however, is also unpredictable. If we can predict future data from past data, that future data 
stops being information. Thus, neither information nor noise can be compressed (and restored to exactly the same 
sequence). We might consider a predictably alternating pattern (such as 0101010...) to be orderly, but it carries no 
information beyond the first couple of bits. 
Thus, orderliness does not constitute order, because order requires information. 
Order is information that fits a 
purpose. The measure of order is the measure of how well the information fits the purpose.
In the evolution of 
lifeforms, the purpose is to survive. In an evolutionary algorithm (a computer program that simulates evolution to 
solve a problem) applied to, say, designing a jet engine, the purpose is to optimize engine performance, efficiency, and 
possibly other criteria.
7
Measuring order is more difficult than measuring complexity. There are proposed measures of 
complexity, as I discussed above. For order, we need a measure of "success" that would be tailored to each situation. 
When we create evolutionary algorithms, the programmer needs to provide such a success measure (called the "utility 
function"). In the evolutionary process of technology development, we could assign a measure of economic success. 
Simply having more information does not necessarily result in a better fit. Sometimes, a deeper order—a better fit 
to a purpose—is achieved through simplification rather than further increases in complexity. For example, a new 
theory that ties together apparently disparate ideas into one broader, more coherent theory reduces complexity but 
nonetheless may increase the "order for a purpose." (In this case, the purpose is to accurately model observed 
phenomena.) Indeed, achieving simpler theories is a driving force in science. (As Einstein said, "Make everything as 
simple as possible, but no simpler.") 
An important example of this concept is one that represented a key step in the evolution of hominids: the shift in 
the thumb's pivot point, which allowed more precise manipulation of the environment."
8
Primates such as chimpanzees 
can grasp but they cannot manipulate objects with either a "power grip," or sufficient fine-motor coordination to write 
or to shape objects. A change in the thumb's pivot point did not significantly increase the complexity of the animal but 
nonetheless did represent an increase in order, enabling, among other things, the development of technology. 
Evolution has shown, however, that the general trend toward greater order does typically result in greater complexity.
9
Thus improving a solution to a problem—which usually increases but sometimes decreases complexity—
increases order. Now we are left with the issue of defining the problem. Indeed, the key to an evolutionary algorithm 
(and to biological and technological evolution in general) is exactly this: defining the problem (which includes the 
utility function). In biological evolution the overall problem has always been to survive. In particular ecological niches 
this overriding challenge translates into more specific objectives, such as the ability of certain species to survive in 
extreme environments or to camouflage themselves from predators. As biological evolution moved toward humanoids, 
the objective itself evolved to the ability to outthink adversaries and to manipulate the environment accordingly. 

---

It may appear that this aspect of the law of accelerating returns contradicts the second law of thermodynamics, 
which implies that entropy (randomness in a closed system) cannot decrease and, therefore, generally increases.
10
However, the law of accelerating returns pertains to evolution, which is not a closed system. It takes place amid great 
chaos and indeed depends on the disorder in its midst, from which it draws its options for diversity. And from these 
options, an evolutionary process continually prunes its choices to create ever greater order. Even a crisis, such as the 
periodic large asteroids that have crashed into the Earth, although increasing chaos temporarily, end up increasing—
deepening—the order created by biological evolution. 
To summarize, evolution increases order, which mayor may not increase complexity (but usually does). A primary 
reason that evolution—of life-forms or of technology—speeds up is that it builds on its own increasing order, with 
ever more sophisticated means of recording and manipulating information. Innovations created by evolution encourage 
and enable faster evolution. In the case of the evolution of life-forms, the most notable early example is DNA, which 
provides a recorded and protected transcription of life's design from which to launch further experiments. In the case 
of the evolution of technology, ever-improving human methods of recording information have fostered yet further 
advances in technology. The first computers were designed on paper and assembled by hand. Today, they are designed 
on computer workstations, with the computers themselves working out many details of the next generation's design, 
and are then produced in fully automated factories with only limited human intervention. 
The evolutionary process of technology improves capacities in an exponential fashion. Innovators seek to improve 
capabilities by multiples. Innovation is multiplicative, not additive. Technology, like any evolutionary process, builds 
on itself. This aspect will continue to accelerate when the technology itself takes full control of its own progression in 
Epoch Five.
11 
We can summarize the principles of the law of accelerating returns as follows: 
•
Evolution applies positive feedback: the more capable methods resulting from one stage of evolutionary progress 
are used to create the next stage. As described in the previous chapter, each epoch of evolution has progressed 
more rapidly by building on the products of the previous stage. Evolution works through indirection: evolution 
created humans, humans created technology, humans are now working with increasingly advanced technology to 
create new generations of technology. By the time of the Singularity, there won't be a distinction between 
humans and technology.