This week, independent tests have been passed by DWaveSys. A company, that has been working on quantum computing for a few years now. It was the source of some controversy in the past, but it seems that the doubts were not founded.
The principle they use is quantum annealing. It's the property of matter to rest in the state of lowest energy when put near 0 degree Kelvin. Using some digital analog converters, they can define some energy landscapes corresponding to the problem they wish to solve, then cool the computer, then read the states of the qubits.
The results are in. When running full-hardware (i.e. best case) it's 3600 times faster than a software version at solving the particular minimisation problem (a version of the Ising problem (which is NP-Complete so it holds the promise of solving all NP-hard problems in polynomial time ) ). But what's important is not that it's more than 3 orders of magnitude faster, it's that up to now the scaling is exponential (more exactly it scales in exp( sqrt( number of qubits) ). This means that when the number of qubits will be 4 times greater, the chip will be 10 000 000 times faster than software. If the trend continues as it seems it would, moore's law will no longer hold, but contrary to what pessimistic minds could have thought, it would no longer hold because the scaling would now be doubly exponential.
Singularity, may just be a few years away...
No comments:
Post a Comment