Latest from ICANN: Quantum Computers + N85 Peering Forum
Latest from ICANN: Quantum Computers are "Interesting"… But Don't Lose your Head *An Interview with ICANN, Paul Hoffman * In a recent publication, written by ICANN (Internet Corporation for Assigned Names and Numbers), chief technology officer Paul Hoffman discussed the hot topic of Quantum Computing and the DNS. NANOG had the opportunity to talk to Hoffman to better understand the current status + future of quantum computing. Learn what questions our community should be asking + and how much discussion we should be giving this topic at this time. *READ NOW <https://www.nanog.org/stories/latest-from-icann-quantum-computers-are-interesting-but-dont-lose-your-head/>* Peering Forum: Now Accepting Applications Peering Coordination Forums are an incredible way to network with others in the industry! The 90-minute session will occur during the hybrid NANOG 85 conference, which will kick off in Montreal on Jun. 6. Please note that participants must be physically present to attend. NANOG 85 Peering Forum applications will remain open until we have received twenty applications or the deadline date of May 31. *SUBMIT NOW <https://www.nanog.org/events/nanog-85/peering-forum/>*
Nanog News wrote:
Latest from ICANN: Quantum Computers are "Interesting"… But Don't Lose your Head
But, quantum computers are mocked up by theoretical physicists, IT amateurs who don't know basics of computational and/or information theory at all, and, as such, just do not work better than classical ones. As for adiabatic quantum computers (those of D-wave), they can be configured to form various quantum circuits. Thus, a quantum circuit, the minimum energy state of which is a solution to some optimization problem, may be constructed. In addition, according to quantum mechanical adiabatic theorem, the state can be reached within polynomial time (w.r.t. problem size), *IF* (very big if) energy difference between the lowest and the second lowest energy states is not very small. However, it is obvious that errors to construct the circuit will badly affect the result. IIRC, D-wave quantum computer has 8bit ADC to control coupling strength between qubits, which is poorly accurate. Worse, there are known classical, not quantum, approximation algorithms to compute the lowest energy state with certain accuracy (w.r.t. energy) within polynomial time. As such, "if energy difference between the lowest and the second lowest energy states is not very small", the classical algorithms can find the correct answer and quantum computing is no better than the classical algorithms. It should be noted that, with really hard problems, the energy differences are very small (become exponentially smaller as problem size increases), which can not be solved by classical approximation nor quantum algorithms in polynomial time. Though it is possible to construct quantum circuits to enhance energy differences a lot, such circuits, like high Q resonators, require extreme accuracy, which is impossible to construct for large problem size, which means such attempts do not contribute scalability. Uselessness of quantum logic gate style quantum computers will be discussed in a separate mail. Masataka Ohta
As I wrote:
Nanog News wrote:
Latest from ICANN: Quantum Computers are "Interesting"… But Don't Lose your Head
Uselessness of quantum logic gate style quantum computers will be discussed in a separate mail.
Quantum logic gate style quantum computers use qubits, which have two orthogonal quantum state |0> and |1> (which may be a horizontally and vertically polarized photon, correspondingly, which is familiar for us, communication engineers) used as bit values of 0 and 1. As they are orthogonal, we can consider |0>+|1>, (which may be a diagonally polarized photon, easy to understand classically). But such addition is improperly called quantum superposition. Two qubit states are represented as |00>, |11> and |00>+|11>. Then, if we can construct a linear circuit to compute f, a 3 input and 3 output bits function, f(a, b, c)=(d, e, f) as f(|abc>)=|def>. Then, we can compute f(|000>)+f(|001>>+...+f(|111>) as f(|000>+|001>+...+|111>), that is evaluating f not 8 times but only once, which is quantum parallelism. It can be extended for N qubit cases for 2**N parallelism by 2**N terms. However, though an N qubit state, like an N bit string, can naturally distinguish 2**N cases, 2**N term quantum superpositioned states require distinguishing 2**(2**N) cases, which is obviously impossible because of noise/error (theory of Shannon). So, QEC (Quantum Error Correction) was invented, which was expected to reduce noise/error. QEC certainly work for single term states. Small non linear error on single term states is no different from linear error. The problem, however, is that, QEC is non-linear, which means not applicable to 2**N term superpositioned states. But, it is implicitly and wrongly assumed that all the 2**N terms will suffer from identical error, in which case, QEC become linear. That is single, so called, bit flip error on the first qubit of: |000>+|111> occurs as: |100>+|111> or: |000>+|011> not (unless strong correlation exists, which is an improper assumption): |100>+|011> which is the reason why quantum logic gate quantum computer with practical number of qubits is impossible. A complication is that it is possible to construct complex QEC scheme applicable for 2 term states. So called Shor code is such QEC. But Shor code itself use states with more than 2 terms. Anyway, once QEC scheme is fixed, it is not applicable to 2**N term states for large N. It should also be noted that though non-linear error on two or more term states imply errors caused by interactions of multiple terms, they are ignored. Masataka Ohta
participants (2)
-
Masataka Ohta -
Nanog News