Learning with Microsoft's Azure Machine
Our machines aren't quite as self-aware as Skynet yet, despite Elon Musk's fears, but that doesn't mean they aren't learning. Behind many modern systems is some amount of machine-learning; Google, for example, uses it to build their vast "knowledge graph" to link data sets to one another and create an engine which can find almost any answer available on the internet (and can you think of much that isn't?)
Microsoft also uses it for the system behind Bing, called 'Satori', which helps to power the company's impressive new virtual assistant 'Cortana'. This is all based in Azure, Microsoft's Cloud. The solutions for internal projects, such as Xbox, were built specifically to tackle the barriers of machine learning today – which generally consists of specialist data scientists having to be on-site at all times. 'Azure Machine Learning' is a fully-managed Cloud service which enters preview next month and simplifies building predictive analytics solutions to build data-driven applications "in mere hours" instead of weeks and months, as was previously the case.
It enables the possibility of tapping into the power of machine-learning to smaller customers instead of being reserved to large companies who can afford all the startup and maintenance costs of implementing such powerful and generally complex technology. Visual workflows and startup templates make common machine-learning tasks simple, whilst the ability to publish APIs and web service in minutes alongside collaboration quickly turns analytics into powerful, enterprise-ready solutions.
As we delve deeper into the "Internet of Things"-era where all our devices are connected to the web in some form or another, Machine Learning will become ever more important to understand how each device is working, what data is being collected, and how it can be used to automate previously manual tasks and create more efficiency.
Gartner's hype cycle report shows Big Data is in the "trough of disillusionment" due to it not delivering the benefits to companies it has claimed. A lot of this is due to the complexity in working with it effectively which Azure Machine Learning is hoping to simplify. One of Microsoft's teams applied it to fraud detection around license keys and seen an extra 20 - 30% savings.
Outside of Microsoft, machine-learning is becoming increasingly important. Amdocs offers a solution which can predict when a customer may be unhappy before they call so action can be taken to rectify the situation before anger and frustration is caused which helps retain their custom, as an example.
The Azure ML public preview will be unleashed in July.
Will you be taking a look at utilising Azure Machine Learning? Let us know in the comments.
- » GitHub announces native Android and iOS apps alongside new features
- » StrongSalt’s new Open Privacy API offers ‘encryption as a service’
- » Sophos launches a security analysis platform for developers
- » Google simplifies trawling through Android Open Source Project code
- » Apple launches dedicated mobile app for developers – to eventually replace WWDC app