Microsoft's CEO talks up how its Azure cloud platform will democratize machine learning.
The advent of easily accessible machine learning services in the cloud will see developers "fuse intelligence into everything they're doing" says Microsoft CEO Satya Nadella.
Each of the major cloud providers, Amazon Web Services, Microsoft and Google are keen to position their platform as having the scale, hardware and diversity of services to make it easier for firms to begin experimenting with machine learning - democratizing AI as Nadella calls it.
"Azure needs to be grounded in what is the next generation of applications," Nadella told the Transform Conference in London.
"What is going to be the core currency of the applications of the future? The answer is very clear, it's going to be data, but more importantly it's going to be the ability to reason over data to create intelligence."
At the event in London today, The Alan Turing Institute, a UK-based data science research group, demonstrated what was possible by exploiting on-demand machine learning.
The institute will use Microsoft's Azure cloud to tune its computer vision system that can take a 2D video of a person speaking and transform them into a lip-synced 3D model in real-time, something that perhaps could one day find a use in a virtual reality telepresence application.
Andrew Blake, director of The Alan Turing Institute, described the task as "very difficult to do and requiring substantial computing power".
"It will be a revolution in how we can do this sort of work when we get access to cloud computing," he said.
The institute has been awarded $5m-worth of Microsoft Azure cloud computing credits to support its research into data science.
Other work at the institute includes creating an automatic statistician using Bayesian statistics -- which recently won the ChaLearn Automatic Machine Learning challenge --- and developing algorithms to recognize human actions from motion.
Nadella namechecked a number of small organizations using Azure to access machine learning, such as the Eyes on the Seas project, which uses computer vision algorithms to monitor satellite images of coastlines for signs of illegal fishing and the Epilepsy Care Alliance, which is collecting data from patients wearing the Microsoft Band and will use Azure Machine Learning to develop algorithms that could help detect seizures.
As with AWS, Azure also offers virtual machines running on clusters of GPUs, via its N-Series platform, which are suited to handling large-scale computing tasks that have been broken down into workloads that can be processed in parallel. This approach is particularly well-suited to tasks related to machine learning and statistical analysis, useful for everything from molecular modelling to seismic analysis.
Sitting on top of this AI-oriented architecture, Microsoft offers access to higher level AI systems via 20 APIs, allowing developers with very little knowledge of machine learning to connect to speech, image, object recognition and natural language processing services.
Every compute node of Azure across all regions has FPGAs (Field Programmable Gate Arrays), which he said means developers can write neural network code, distribute it across the FPGA fabric and "run it at the speed of silicon" -- repeating his description of Azure as "the world's first AI supercomputer".
Nadella talked up the scale of the Azure cloud, saying it had a local presence in more countries around the world than any other provider, with sites in 13 regions. Microsoft are spending $3bn building datacenters in Europe, with datacenters currently in or planned in Ireland, Amsterdam, Germany, France and the UK.
- Microsoft Ignite: Nadella outlines 4 pillars for democratizing AI (TechRepublic)
- How to customize the Windows 10 Start menu with numbered shortcuts(TechRepublic)
- Microsoft Windows Server 2016 to be generally available in mid-October(ZDNet)
- Ditching Microsoft Office? Tips on how to switch to LibreOffice(TechRepublic)
- Microsoft: Windows 10 now on 400 million devices (ZDNet)
- Windows 10: New hardware is coming, but will they be the devices Microsoft really needs? (TechRepublic)