Microsoft Word Kurzweil, Ray The Singularity Is Near doc

Computing with Light. 
Another approach to SIMD computing is to use multiple beams of laser light in which 
information is encoded in each stream of photons. Optical components can then be used to perform logical and 
arithmetic functions on the encoded information streams. For example, a system developed by Lenslet, a small Israeli 
company, uses 256 lasers and can perform eight trillion calculations per second by performing the same calculation on 
each of the 256 streams of data.
31
The system can be used for applications such as performing data compression on 
256 video channels. 
SIMD technologies such as DNA computers and optical computers will have important specialized roles to play in 
the future of computation. The replication of certain aspects of the functionality of the human brain, such as processing 
sensory data, can use SIMD architectures. For other brain regions, such as those dealing with learning and reasoning, 
general-purpose computing with its "multiple instruction multiple data" (MIMD) architectures will be required. For 
high-performance MIMD computing, we will need to apply the three-dimensional molecular-computing paradigms 
described above. 
Quantum Computing. 
Quantum computing is an even more radical form of SIMD parallel processing, but one that is 
in a much earlier stage of development compared to the other new technologies we have discussed. A quantum 
computer contains a series of qubits, which essentially are zero and one at the same time. The qubit is based on the 
fundamental ambiguity inherent in quantum mechanics. In a quantum computer, the qubits are represented by a 
quantum property of particles—for example, the spin state of individual electrons. When the qubits are in an 
"entangled" state, each one is simultaneously in both states. In a process called "quantum decoherence" the ambiguity 
of each qubit is resolved, leaving an unambiguous sequence of ones and zeroes. If the quantum computer is set up in 
the right way, that decohered sequence will represent the solution to a problem. Essentially, only the correct sequence 
survives the process of decoherence. 
As with the DNA computer described above, a key to successful quantum computing is a careful statement of the 
problem, including a precise way to test possible answers. The quantum computer effectively tests every possible 
combination
of values for the qubits. So a quantum computer with one thousand qubits would test 2
1,000
(a number 
approximately equal to one followed by 301 zeroes) potential solutions simultaneously. 
A thousand-bit quantum computer would vastly outperform any conceivable DNA computer, or for that matter 
any conceivable nonquantum computer. There are two limitations to the process, however. The first is that, like the 
DNA and optical computers discussed above, only a special set of problems is amenable to being presented to a 
quantum computer. In essence, we need to I be able to test each possible answer in a simple way. 

---

The classic example of a practical use for quantum computing is in factoring very large numbers (finding which 
smaller numbers, when multiplied together, result in the large number). Factoring numbers with more than 512 bits is 
currently not achievable on a digital computer, even a massively parallel one.
32
Interesting classes of problems 
amenable to quantum computing include breaking encryption codes (which rely on factoring large numbers). The 
other problem is that the computational power of a quantum computer depends on the number of entangled qubits, and 
the state of the art is currently limited to around ten bits. A ten-bit quantum computer is not very useful, since 2
10
is 
only 1,024. In a conventional computer, it is a straightforward process to combine memory bits and logic gates. We 
cannot, however, create a twenty-qubit quantum computer simply by combining two ten-qubit machines. All of the 
qubits have to be quantum-entangled together, and that has proved to be challenging. 
A key question is: how difficult is it to add each additional qubit? The computational power of a quantum 
computer grows exponentially with each added qubit, but if it turns out that adding each additional qubit makes the 
engineering task exponentially more difficult, we will not be gaining any leverage. (That is, the computational power 
of a quantum computer will be only linearly proportional to the engineering difficulty.) In general, proposed methods 
for adding qubits make the resulting systems significantly more delicate and susceptible to premature decoherence. 
There are proposals to increase significantly the number of qubits, although these have not yet been proved in 
practice. For example, Stephan Gulde and his colleagues at the University of Innsbruck have built a quantum computer 
using a single atom of calcium that has the potential to simultaneously encode dozens of qubits—possibly up to one 
hundred—using different quantum properties within the atom.
33
The ultimate role of quantum computing remains 
unresolved. But even if a quantum computer with hundreds of entangled qubits proves feasible, it will remain a 
special-purpose device, although one with remarkable capabilities that cannot be emulated in any other way. 
When I suggested in