Microsoft Word Kurzweil, Ray The Singularity Is Near doc

Computing with DNA. 
DNA is nature's own nanoengineered computer, and its ability to store information and 
conduct logical manipulations at the molecular level has already been exploited in specialized "DNA computers." A 
DNA computer is essentially a test tube filled with water containing trillions of DNA molecules, with each molecule 
acting as a computer. 
The goal of the computation is to solve a problem, with the solution expressed as a sequence of symbols. (For 
example, the sequence of symbols could represent a mathematical proof or just the digits of a number.) Here's how a 
DNA computer works. A small strand of DNA is created, using a unique code for each symbol. Each such strand is 

---

replicated trillions of times using a process called "polymerase chain reaction" (PCR). These pools of DNA are then 
put into a test tube. Because DNA has an affinity to link strands together, long strands form automatically, with 
sequences of the strands representing the different symbols, each of them a possible solution to the problem. Since 
there will be many trillions of such strands, there are multiple strands for each possible answer (that is, each possible 
sequence of symbols). 
The next step of the process is to test all of the strands simultaneously. This is done by using specially designed 
enzymes that destroy strands that do not meet certain criteria. The enzymes are applied to the test tube sequentially, 
and by designing a precise series of enzymes the procedure will eventually obliterate all the incorrect strands, leaving 
only the ones with the correct answer. (For a more complete description of the process, see this note:
26
) 
The key to the power of DNA computing is that it allows for testing each of the trillions of strands 
simultaneously. In 2003 Israeli scientists led by Ehud Shapiro at the Weizmann Institute of Science combined DNA 
with adenosine triphosphate (ATP), the natural fuel for biological systems such as the human body.
27
With this 
method, each of the DNA molecules was able to perform computations as well as provide its own energy. The 
Weizmann scientists demonstrated a configuration consisting of two spoonfuls of this liquid supercomputing system, 
which contained thirty million billion molecular computers and performed a total of 660 trillion calculations per 
second (6.6 
°
10
14
cps). The energy consumption of these computers is extremely low, only fifty millionths of a watt 
for all thirty million billion computers. 
There's a limitation, however, to DNA computing: each of the many trillions of computers has to perform the 
same operation at the same time (although on different data), so that the device is a "single instruction multiple data" 
(SIMD) architecture. While there are important classes of problems that are amenable to a SIMD system (for example, 
processing every pixel in an image for image enhancement or compression, and solving combinatorial-logic 
problems), it is not possible to program them for general-purpose algorithms, in which each computer is able to 
execute whatever operation is needed for its particular mission. (Note that the research projects at Purdue University 
and Duke University, described earlier, that use self-assembling DNA strands to create three-dimensional structures 
are different from the DNA computing described here. Those research projects have the potential to create arbitrary 
configurations that are not limited to SIMD computing.)