World's First Programmable Quantum Photonic Chip

MrSeb writes "A team of engineering geniuses from the University of Bristol, England has developed the world's first re-programmable, multi-purpose quantum photonic computer chip that relies on quantum entanglement to perform calculations. With multiple waveguide channels (made from standard silicon dioxide), and eight electrodes, the silicon chip is capable of repeatedly entangling photons. Depending on how the electrodes are programmed, different quantum states can be produced. The end result is two qubits that can be used to perform quantum computing. Most importantly, though, unlike existing quantum photonic setups which require apparatus the size of a 'large dining table,' this new chip is tiny: just 70mm (2.7 inches) by 3mm."

Re:computing power scales exponentially

[Janek Kozicki]

oh, and I forgot to mention - that's also the reason why quantum physics is so difficult to model using our today's computers. Monte carlo and other rough estimations are widely used. Only simplest problems (think harmonic oscilator) have analytical (and crazy complex) solutions.

Re:computing power scales exponentially

[hweimer]

Oh, we already have a quantum version of Moore's law [quantenblog.net]. However, the time constant for doubling is on the order of six years and not 18 months.

Re:computing power scales exponentially

[mikael]

The general rule for qubits seems to be anything that requires a unique solution but has to consider every possible combination of boolean states. Since they are Boolean zero or one values, that leads to cryptography because a relatively few number of bits would be required; 256,512,1024.

GPU's do floating-point calculations in parallel, which is really good for those problems which have to apply the same algorithm to different data points, like CFD, physics, AI, image and signal processing.

To represent floating-point data would require at least 16 qubits for half-floats, 32-bits for IEEE 754 standard floats, and 64-bits for doubles. But to do anything useful like CFD, would require storage of the entire state of the system which would require gigabits of data.

Unless someone could shrink the problem of CFD modelling down to atomic scales using phantom atoms, and overlapping qubits onto the same logic, GPU's won't have any competition.

Re:computing power scales exponentially

[Anonymous Coward]

No, that's false. Quantum computing offers polynomial time algorithms for a very small set of problems for which classically only exponential time algorithms are known, particularly, instances of the hidden subgroup problem [wikipedia.org] (including integer factorization in the form of Shor's algorithm [wikipedia.org]). More generally, Grover's algorithm [wikipedia.org] gives some speedup to general NP problems, but not exponential. As I understand it, you would need a rather large quantum computer before it would actually faster than existing classical computers.

Re:computing power scales exponentially

[Surt]

http://en.wikipedia.org/wiki/Quantum_computer#Potential [wikipedia.org]

Re:computing power scales exponentially

[Anonymous Coward]

It's Mersenne primes. And use powers of 2, not 10.

Re:Bad news for crypto

[Prune]

This is plain wrong. Under a quantum-computer attack, AES256 is as strong as AES128. Thus, you simply need to double your key size. Most symmetric ciphers are safe. Most public-key, on the other hand, is indeed broken by quantum-computation. People often forget that for most things, quantum algorithms can only provide a quadratic speedup--not an exponential one!