Microsoft Word Kurzweil, Ray The Singularity Is Near doc

Bigger or Smaller.
Given that the computational capacity of our solar system is in the range of 10
70
to 10
80
cps, we 
will reach these limits early in the twenty-second century, according to my projections. The history of computation 
tells us that the power of computation expands both inward and outward. Over the last several decades we have been 
able to place twice as many computational elements (transistors) on each integrated circuit chip about every two years, 
which represents inward growth (toward greater densities of computation per kilogram of matter). But we are also 
expanding outward, in that the number of chips is expanding (currently) at a rate of about 8.3 percent per year.
78
It is 
reasonable to expect both types of growth to continue, and for the outward growth rate to increase significantly once 
we approach the limits of inward growth (with three-dimensional circuits). 
Moreover, once we bump up against the limits of matter and energy in our solar system to support the expansion 
of computation, we will have no choice but to expand outward as the primary form of growth. We discussed earlier the 
speculation that finer scales of computation might be feasible—on the scale of subatomic particles. Such pico- or 
femtotechnology would permit continued growth of computation by continued shrinking of feature sizes. Even if this 
is feasible, however, there are likely to be major technical challenges in mastering subnanoscale computation, so the 
pressure to expand outward will remain. 
Expanding Beyond the Solar System.
Once we do expand our intelligence beyond the solar system, at what rate will 
this take place? The expansion will not start out at the maximum speed; it will quickly achieve a speed within a 
vanishingly small change from the maximum speed (speed of light or greater). Some critics have objected to this 
notion, insisting that it would be very difficult to send people (or advanced organisms from any other ETI civilization) 
and equipment at near the speed of light without crushing them. Of course, we could avoid this problem by 
accelerating slowly, but another problem would be collisions with interstellar material. But again, this objection 
entirely misses the point of the nature of intelligence at this stage of development. Early ideas about the spread of ETI 
through the galaxy and universe were based on the migration and colonization patterns from our human history and 
basically involved sending settlements of humans (or, in the case of other ETI civilizations, intelligent organisms) to 
other star systems. This would allow them to multiply through normal biological reproduction and then continue to 
spread in like manner from there. 
But as we have seen, by late in this century nonbiological intelligence on the Earth will be many trillions of times 
more powerful than biological intelligence, so sending biological humans on such a mission would not make sense. 
The same would be true for any other ETI civilization. This is not simply a matter of biological humans sending 
robotic probes. Human civilization by that time will be nonbiological for all practical purposes. 
These nonbiological sentries would not need to be very large and in fact would primarily comprise information. It 
is true, however, that just sending information would not be sufficient, for some material-based device that can have a 
physical impact on other star and planetary systems must be present. However, it would be sufficient for the probes to 
be self-replicating nanobots (note that a nanobot has nanoscale features but that the overall size of a nanobot is 
measured in microns).
79
We could send swarms of many trillions of them, with some of these "seeds" taking root in 
another planetary system and then replicating by finding the appropriate materials, such as carbon and other needed 
elements, and building copies of themselves. 

---

Once established, the nanobot colony could obtain the additional information it needs to optimize its intelligence 
from pure information transmissions that involve only energy, not matter, and that are sent at the speed of light. Unlike 
large organisms such as humans, these nanobots, being extremely small, could travel at close to the speed of light. 
Another scenario would be to dispense with the information transmissions and embed the information needed in the 
nanobots' own memory. That's an engineering decision we can leave to these future superengineers. 
The software files could be spread out among billions of devices. Once one or a few of them get a "foothold" by 
self-replicating at a destination, the now much larger system could gather up the nanobots traveling in the vicinity so 
that from that time on, the bulk of the nanobots sent in that direction do not simply fly by. In this way, the now 
established colony can gather up the information, as well as the distributed computational resources, it needs to 
optimize its intelligence.