Why Microsoft’s AI is about helping anyone’s intelligent agent

Автор: | 18.05.2018

In the new age of intelligent agents such as Google Home and Alexa from Amazon, Microsoft sees its primary role as a multi-device AI platform provider. This was among the main messages to come out of the company’s recent Build developer conference in Seattle.

Unlike its two main rivals, its CEO, Satya Nadella, positions the Microsoft AI platform as agnostic, in that it is being designed to support multiple agent devices.

In his opening keynote at Build 2018 in Seattle, Nadella said: “You want to have max reach as a developer – you should be able to publish in Cortana, Skype or Facebook and have your conversational user interface show up.”

Microsoft’s ambition is to separate the logic that goes into the intelligence of an intelligent agent from where it runs and where it is deployed, to enable developers to create agents that can work on any device.

The goal for Microsoft is that these agents will be developed using Microsoft tools; they will integrate with Microsoft products such as Outlook or Teams and call on its growing library of AI service in the Azure cloud.

The idea of separating the intelligent agent from the device that hosts it is conceptually similar to how the web browser worked as a common platform to enable businesses to create websites.

Cross-platform, consistent design

“When the web was born, businesses created websites, not their own browsers,” said David Camona, general manager for AI marketing at Microsoft. “You created a website that worked across multiple channels. The same will happen with AI agents.”

While there is a lot of focus on AI-powered digital assistants, Camona said: “This is not about bringing another device into every home.”

Instead, Microsoft sees an opportunity for developers to build an agent that represents the company they work for across all intelligent agent-based devices. “The device becomes the channel through which you connect to the customer,” he said.

Unlike Google and Amazon, Microsoft may miss out bring front and foremost in people’s homes. But its message is for the enterprise: what is important, at least as far as Microsoft is concerned, is that the business maintain a consistent brand identity across all the channels where it deploys its agent software.

The Microsoft AI strategy is focused on making it easier for developers to start creating their own AI experiences that can work across key products in Microsoft’s software portfolio.

During the Build developer conference, Microsoft showcased how code templates in its Visual Studio integrated development tool environment could be edited to create new AI agent skills. It also demonstrated how a simple JavaScript in an Excel spreadsheet could be used to call a cloud-based machine learning algorithm.

Maturity of enterprise AI

For Camona, there are a number of key challenges businesses need to address when assessing AI. First is how to start to use AI to get a return for the business.

He said the conversation starts with where to use AI in the business. Camona positions Microsoft’s AI platform across three areas, which differ in scope and maturity. First is conversational AI, which embraces a conversational user interface. “A conversational user interface is not simple but it is a definite first step to engage with customers and employees,” he said.

“We believe the core essence of a conversational AI, it is not about the actual agent. Our core strategy in business is to enable businesses to have their own conversational AI assistant. The difficult part is natural language understanding and how to connect the outcome for a user.” He said Microsoft was making a platform available to developers to help them achieve this.

As part of this platform strategy, two new services were unveiled at Microsoft Build 2018. The first, Conversational Learner, is targeted at supporting richer conversations. The service trains the agent with real conversations. The second is Personality Chat, which makes bots have a personality –  they say good morning or ask about the weather, said Camona.

These form part of the next generation of Conversational AI tools from Microsoft, to help developers build, connect, deploy and manage intelligent bots to interact naturally with users.

Non-intuitive queries

During Build, Microsoft demonstrated how a user would be able to get Alexa to do something by asking Cortana to ask Alexa. Such queries in a speech user interface are non-intuitive. But a speech UI can be made more seamless.

“In Cortana you have the concept of semantic skills. Depending on the context of the conversation, Cortana will connect to the appropriate service or skill,” said Camona.

“So you could ask to book a hairdresser appointment and the agent in Cortana would connect to the most appropriate service, based on your location, without you having to explicitly invoke the skill associated with hairdresser bookings.”

The success of the speech user interface to communicate with intelligent agents will largely depend on how well the skill  being developed understands the context of a query, and the level of integration with the intelligent device, like Alexa. Microsoft’s demonstration at Build only showed a lightweight integration.

The second level of AI maturity is intelligent applications, where AI is integrated into enterprise or customer-facing applications to improve them. “This is how Microsoft exposes very advanced, complex AI through Cognitive Services. Developers can then use AI within their applications,” said Camona.

The highest level of maturity for Microsoft, is where AI is used as part of a wider business transformation. “This has a lot to do with how you use AI to optimise and redefine business processes,” he said.

Industry experts often talk about the quick wins and low-hanging fruits that make up the use cases in business for AI. For Microsoft, this is the golden opportunity to enable third parties to develop intelligent assistants that seamlessly integrate with its office productivity and collaboration software.

In doing so, Word, Excel, PowerPoint et al, become more than a tool for creating, editing and reviewing Office documents. Rather, the software actually helps people to write better letters, manipulate spreadsheet calculations quickly and efficiently and enables them to draw in relevant information to create slide decks much easier than before.

It is the art of the possible, at least for delegates at Build. During the event, Microsoft showed how AI integration between Cortana and Teams could be used to manage interaction in a video conference. A conference room device running Cortana was able to listen to requests from people and present the required information. She also identified each speaker using facial recognition.

The work Microsoft has done on its intelligent agent technology forms part of Microsoft’s broader AI strategy for business. At Build 2018 conference, Microsoft also launched a suite of pre-trained models from Microsoft Cognitive Services, that allow developers to add AI across vision, speech, language, knowledge and search to their applications.

Many of these pre-trained models are now customisable to meet the specific needs of companies and their customers, said Microsoft.

Source

Добавить комментарий