3 ms·

I think they only announced an early version of their prototype, and it's not clear if it's even implemented yet. Google was supposed to announce their own 49-

by mtgx 9y ago

I think they only announced an early version of their prototype, and it's not clear if it's even implemented yet. Google was supposed to announce their own 49-qubit quantum computer this year - one that actually works - but it seems they haven't had time to finish it, or maybe they weren't satisfied with its performance and didn't want to overhype it. Also, I think IBM proving that 56-qubits can be simulated wasn't a simulation of a universal quantum computer, but a more specific simulation. The record in universal quantum computer simulation seems to be 46-qubits: http://www.fz-juelich.de/SharedDocs/Pressemitteilungen/UK/EN/2017/2017-12-15-world-record-juelich-researchers-simulate-quantum-computer.html http://www.fz-juelich.de/SharedDocs/Pressemitteilungen/UK/EN... Also, interestingly enough, the number of qubits that can be simulated seems to increase by 1 every 2 years. I suppose this is because generally you need 2x the memory for your supercomputer to add an extra qubit to the simulation. Being able to double the memory every two years seems to make sense from Moore's Law perspective. In that case, we should be able to simulate a 50-qubit quantum computer by 2025. So any real quantum computer that has 50 logical qubits and is built by 2025 should be able to prove quantum supremacy.

jngreenlee 9y ago
Sounds like it's actually implemented, as least by this quote: "In both the 50- and the 20-qubit systems, the quantum state is preserved for 90 microseconds—a record for the industry, but still an extremely short period of time."
ineedasername 9y ago
I'm familiar with the 56 qubit simulation from IBM: That does not appear to be significantly restricted in any way, at least not based on the limited understanding I can glean from their publication's results & conclusions (I'm not a quantum computer scientist or physicist). They appear to have found a practical tensor-based work around to the previous exponentially increasing memory requirements. Moreover it is one that can, in limited circumstances, be applied in a paralellized, distributed environments (cloud). That appears to be the limitation: They successfully simulated a 56-qubit computer, but many some workloads using their method cannot be distributed, and must be computed within the same environment. I.e., they made two advancements: 1: A successful simulation, and 2: it can be done distributively, only for certain workloads. Of course it's all roughly a billion times slower than a true quantum computer.

by mtgx 9y ago

jngreenlee 9y ago

Sounds like it's actually implemented, as least by this quote: "In both the 50- and the 20-qubit systems, the quantum state is preserved for 90 microseconds—a record for the industry, but still an extremely short period of time."

ineedasername 9y ago

I'm familiar with the 56 qubit simulation from IBM: That does not appear to be significantly restricted in any way, at least not based on the limited understanding I can glean from their publication's results & conclusions (I'm not a quantum computer scientist or physicist). They appear to have found a practical tensor-based work around to the previous exponentially increasing memory requirements. Moreover it is one that can, in limited circumstances, be applied in a paralellized, distributed environments (cloud). That appears to be the limitation: They successfully simulated a 56-qubit computer, but many some workloads using their method cannot be distributed, and must be computed within the same environment. I.e., they made two advancements: 1: A successful simulation, and 2: it can be done distributively, only for certain workloads. Of course it's all roughly a billion times slower than a true quantum computer.