Satya Nadella, the CEO of Microsoft, delivered the first day keynote at the Microsoft Build conference in Seattle Washington today.
The keynote was divided into two sections. The first part was about the opportunities and responsibilities facing Microsoft and the technology community today. The second part presented the current focus of the conference: the intelligent cloud and the intelligent edge, specifically Azure and Microsoft 365.
Opportunities and Responsibilities
Similar to electricity after the industrial revolution, computing power is becoming invisible within the modern economy. Distributed computing, whether in banking, agriculture, the factory, or vehicles, creates an enormous amount of data that is being analyzed. This is the opportunity.
What is the responsibility?
Technology has to empower everyone to create equitable growth and build trust in technology. The idea is that the effects of technology should be compatible with genuine human life, and should consider its impact on future generations. There are three core principles that should be behind the choices to be made.
Privacy is the first principle. Microsoft has hundreds of engineers working on compliance to the European Union's General Data Protection Regulation (GDPR) which goes into effect this month. In addition, Microsoft is committed to continually help its customers to comply now and into the future. Microsoft has worked with the United States government to create a framework to balance the need for privacy with the needs of law enforcement in the CLOUD act legislation. An international framework is needed to help customers stay in control and preserve privacy.
Cybersecurity is the second principle. Microsoft has formulated a program to help protect democracy by working with political campaigns and civic groups to secure the democratic process. It leads a consortium of 34 companies to ensure that citizens across the world are protected, a first step towards a digital Geneva Convention.
Ethics is the third principle. Artificial Intelligence should be about what computers should do. Microsoft has an ethics board inside of the company to oversee the products that are being built. Tools have to be put into the hands of designers and developers, not only to get good AI, but to help them make good choices. This should be a first class engineering requirement. Data sheets that indicate where the data came from can be used to help "debias" word embeddings. Privacy with cross organizational data flows, using technologies such as homomorphic encryption for learning and training data, are being developed.
Platform Technologies
Microsoft wants to build technologies that empower others to build more technology, to align their products so that Microsoft’s success comes from its customer’s success.
Microsoft Azure and Microsoft 365 power the shift to the intelligent cloud and the intelligent edge, ubiquitous computing from fabric to cloud and to the physical world (edge). The application model for this world is distributed, event driven, and serverless. Azure is being built as the world's computer. It has more regions and certifications than any other public cloud, meeting the demands of digital sovereignty and regulatory requirements.
Azure, Azure Stack, Azure IOT Edge, and Azure Sphere are the basis for this new computing model.
Azure Stack, which is just a year old, supports such scenarios as Chevron which must do computing at disconnected oil rigs. Schlumberger uses it and the public cloud to create one computing fabric so that it can compute wherever the data resides.
Azure IOT Edge, which puts intelligence and analytics on edge devices, runs Linux and Windows equally. As of today the runtime is open source.
A new partnership with Qualcomm was announced for a vision AI developer kit for home security and industrial safety. You can train it in the cloud, and then deploy it on a camera. The developer kit will be available at the end of the year.
DJI, the world leader in drones, will develop a drone SDK for Windows 10. Full flight control and data transfer will allow personal computers to directly control DJI drones.
IOT devices are becoming more intelligent. Rockwell Automation uses them to detect defects and anomalies automatically. The video does not have to be sent to the cloud; it can generate the alert locally.
ML models can be delivered to the camera or a drone. Azure Edge can be part of the drone.
Azure Sphere, announced at the RSA conference, is a secure operating system for secure microcontrollers. It shares the same programming model as Azure. Azure Services - compute and AI - can be put on the device even as small as a Raspberry Pi.
Azure Cognitive Services, such as speech translation, text to speech, OCR Recognition, can be used in customized models that can be put in applications. For example, Twitter uses the language translation to translate every tweet.
The Chinese robot maker, Roobo, has created a smart speaker hardware SDK. This will allow micro speakers for any device. Speech recognition in low-volume scenarios with a lot of ambient noise is a prime use case.
Project Kinect brings back the Kinect technology for software and AI developers. This will include spatial understanding, skeletal tracking, object recognition, powerful sensors, and an ultra-wide field of view.
Conversational AI is another target technology. Companies will be able to build branded agents that will be able to converse across multiple digital assistants to aid customers. Bots are seen as the new apps. The new BOT framework features include conversational interfaces, and more customization such as custom speech and custom personality.
Applications will be put into all channels and digital assistants such as Skype, Facebook, or Cortana. The idea is to have them show up everywhere. To enable this, the conversational system is separate from the enabling technology for different platforms.
Tools and Frameworks
Openness is being built into every layer. You could choose any framework (TensorFlow, CNTK, etc.) to build a model, use ONNX, the open neural network exchange framework developed with Facebook, hardware acceleration, CORE ML as well as Windows ML. You are not locked into one process.
Project Brainwave is a distributed intelligence framework to bring real time AI to the edge. You can deploy machine learning models at the edge.
Microsoft 365 is designed for the multi-sense and multi-device experience. It brings Windows and Office together in a continued experience. You could use an app on your phone, Skype in your car for a meeting, work on a PC, or have a meeting with a large screen device such as Surface Hub. In a single day you use multiple devices, in multiple locations, with multiple senses, and multiple people.
The operating system platform abstracts the hardware to allow the app model to work at the higher level. Microsoft Graph has extended its APIs so that, for example, the phone could be a second screen for a PC, or the PC could be a second screen for the phone. Office is an intrinsically multi-device application. For example, Excel could call a cloud function.
Cortana is part of the windows shell, built into Teams and Outlook. It will be able to converse with other personal digital assistants. The Cortana and Alexa integration, now in beta, is an example.
Microsoft Graph underlies Microsoft 365, with explicitly represented events, and explicit consent has to be given to developers to use the data that belongs to people and companies. Since the graph is extensible, apps can be made part of it.
Currently, digital artifacts have to be schematized to be put into the graph. Eventually, physical world data - spatial data, hospital data, factory data, office data - will be part of the graph. For example, HoloLens brings spatial data into Microsoft Graph.
AI for Accessibility was announced, which will give money to researchers and developers to use AI to build better technology for people with disabilities.
The ultimate goal is to empower more people and organizations.