General purpose quantum computing closer to reality with Microsoft breakthroughs(news.microsoft.com)

34 pointsby hexrcsvor 9 Jahren13 comments

news.microsoft.com

General purpose quantum computing closer to reality with Microsoft breakthroughs

https://news.microsoft.com/features/new-microsoft-breakthroughs-general-purpose-quantum-computing-moves-closer-reality/

13 comments

ColanRvor 9 Jahren

> Craig Mundie, who as Microsoft’s chief research and strategy officer first backed Freedman’s push into quantum computing a dozen years ago, noted that if a quantum computer could hypothetically process a training algorithm for the Cortana digital assistant in a day rather than a month, that would mark a profound improvement in AI advances.

I didn't think that quantum computing would help with DNN or parallel processing...what gives?

SomeStupidPointvor 9 Jahren

One of the people mentioned in the article, Svore, works at the QuArC group for MS and has worked on quantum machine learning. (QuArC generally works on how to actually design and use quantum computers.)

For example: https://arxiv.org/pdf/1412.3489.pdf

From my (shallow) understanding, you can get improvements by considering the space of possible states of your quantum system to be the space of weights for your model, and then annealing the quantum system (which is equivalent to optimizing the objective function). The quantum annealing is a (significant) improvement over other optimization methods... or something.

martinquantumvor 9 Jahren

looks like a quantum computer can train models better. Like, find gradients better or just find better gradients.

maxandervor 9 Jahren

My (admittedly fuzzy) understanding is that while running a DNN is a very parallel problem, training it (back-propagation, in particular) is not and is thought to be susceptible to quantum speedup. Google's work in quantum computing is motivated by the same application.

iamcasenvor 9 Jahren

It entirely depends on the problem being solved. I am not sure what their ideas are with machine learning, but essentially quantum computers help with parallelization with algorithms that can take advantage of superpositions.

martinquantumvor 9 Jahren

I think that for pattern matching (DNN) it's not clear, but for parallel stuff the quantum computer is like the ultimate SIMD machine.

manuelelucchivor 9 Jahren

What about the quantum programming language? Q#? The extension is .qb so i think not... What do you think?

martinquantumvor 9 Jahren

The sample they showed did not look like any of the standard .NET languages, so I wonder if they are going with a standalone language. AFAIK most quantum programming languages are actually embedded into another language (C, Haskell, ML) so if they actually implemented a standalone language, that would be interesting. Not sure what the file extension means.

jongalloway2vor 9 Jahren

It looks like it was related to Liquid. Info here: http://stationq.github.io/Liquid/

That specific code sample looks like it was based on the Teleport sample: https://github.com/StationQ/Liquid/blob/master/Samples/Telep...

axonicvor 9 Jahren

Uh oh, here comes Quantum Basic.

martinquantumvor 9 Jahren

when you look at the example, there was also a teleport.g.cs file which looks like a standard #C file, but possibly auto-generated(?) What do you think?

lostmsuvor 9 Jahren

Here, saved you time on reading 3 pages of bloat.

> A new programming language that is designed for developers to create apps to debug on quantum simulators today and run on an actual topological quantum computer in the future.