Microsoft, Google, and Others are Accelerating Machine Learning the Quantum Way

How long is it until the machines take over? Unclear, but it’s certain to be sooner with the development of quantum computing. In late April 2021, Microsoft was granted a patent for “adversarial quantum machine learning.” Microsoft isn’t the only company researching quantum machine learning, with Google, D-Wave Systems, Accenture, and IBM all having patent assets integrating the quantum realm with AI. Quantum machine learning will continue to grow and with good reason.

Quantum machine learning is able to use parameters represented as joint probability distributions. The learner therefore doesn’t need to discretely check all permutations of all parameters, but rather deals with entangled states of those parameters. By trading out vectors of training data for eigenvectors, quantum machine learning algorithms are capable of learning much faster than their classical counterparts. But it’s not just that they’re faster. Microsoft also mentions that quantum machine learning is better able to ensure privacy of the training data, as it allows an experimenter to send out a data set and have participants apply a number of tiny perturbations to the data set that are individually unnoticeable, but help to train the model in aggregate. Participants therefore do not give away enough information for even the person training the model to determine information about them.

Ok, but what does the adversarial part of Microsoft’s patent mean? Well, a common problem with machine learning models, especially when they train using data not entirely controlled by the person running the models, is that they are vulnerable to being intentionally mistrained. For example, one could perhaps train a chatbot to primarily speak in obscenities and slurs by having its primary inflow of information be laden with obscenities and slurs. If that sounds familiar, it’s because it happened to Microsoft in 2016, when its Tay chatbot was trained to say a number of controversial statements. However, an “adversarial” machine learning algorithm includes specific measures to detect and prevent such attacks. Microsoft, like its AI, has learned.

For more developments regarding quantum machine learning, check out the Quantum Computing Patent Forecast®!

Related Posts

Microsoft Debugs Quantum Software in Virtual Reality, Catches Up To Minority Report

In almost every Sci-Fi film, there is a scene where some scientist is manipulating holographic images with just their hands. Microsoft is…


IBM and Google are building it, but who will come?

Unlike many other sectors, the Quantum Computing sector has so far been dominated by a few major hardware developers: IBM, Google, D-Wave…


The Quantum British Invasion

In the race to develop Quantum Computing, in addition to facing challenges across the Pacific in the form of China’s Jiuzhang Computer, the…