By: Dann Anthony Maurno
“We’re not pursuing [artificial intelligence] AI to beat humans at games; we are pursuing AI so we can empower every person and every institution” with the tools of AI. So began Microsoft CEO Satya Nadella’s afternoon keynote at Ignite 2016.
Nadella went on to describe why this lofty goal is necessary, and how Microsoft is achieving it. He also described how Microsoft has stealthily souped-up its global cloud infrastructure, with a ten-fold boost in compute capacity. (More on that below.)
AI an historic “inflection point”
Nadella described AI as the third of three inflection points that democratized intelligence, the first being the invention of the Gutenberg press in the 1400s when perhaps 30,000 books in the world burgeoned suddenly to millions; and the second being the World Wide Web in the ‘90s.
But you can read a book in a few sittings. With information coming fast and furiously from dozens of sources (consider your email inbox alone), how, Nadella asks, are we to make sense of it all?
Time is the key constraint on human productivity; brains don’t work as fast as information comes, and have limited capacity. AI in essence gives an individual a larger, faster brain. That “intelligence” is not data, it is rich, contextual, active, rapidly-accessed knowledge, garnered from the user’s own habits (like contacts used and searches executed). Rather like Netflix knows your viewing habits and offers suggestions, but on a far grander scale.
Nadella outlined a four-pronged approach to “democratizing” AI, including:
- Agents: Harness AI to change how we interact with ambient computing (“agents”).
- Applications: Infuse AI into every application that we interact with, on any device, at any point in time, with intelligence.
- Services: Make Microsoft AI available to every application developer in the world.
- Infrastructure: Build the world’s most powerful AI supercomputer, then make it available to anyone via the cloud.
AI, Cortana for enterprise
Cortana, Microsoft’s personal productivity application, is also its AI interface. Nadella claimed 12+ billion queries to Cortana to date and 133 million active users.
“She” (as Nadella referred to Cortana) is of course a personal productivity tool; and check out Microsoft’s Cortana page and you are assured that “Cortana knows a lot, including how to have fun,” and a “good chuckle when you have time to kill.”
So it knows how to play, but can it work?
It does. To understand Cortana, it doesn’t simply find stuff; it learns about its user. Think of those eerily accurate Google searches, tailored to you, and amp those capabilities up by a factor of ten. Now put it across devices – “she” lives in the cloud, so the agent (smartphone, laptop, etc.) is irrelevant. She’s as ever-present as your shadow.
Nadella went on to describe a few of Microsoft’s own AI-based applications, including:
- The MileIQ location-aware app for travel logging, which says Nadella has logged 11 billion miles and saved its users $1.2 billion.
- The SwiftKey smart keyboard that uses a neural network with highly-intelligent type tracking. Here Nadella claimed 300 million users who’ve saved 100,000 years of time.
- My Analytics (in Office 365), a kind of productivity tracker (which Nadella likened to a fitness tracker) informed by the Microsoft Graph and powered by AI.
The idea, says Nadella, is “taking all of the signal and driving information into intelligent action.”
That approach is one that Microsoft will take into every one of its mainstream products, says Nadella. Beyond personal productivity, Microsoft expects its AI to drive revenue-generating enterprise utility that the market will see in the push ahead on Dynamics 365, which will include much of the Dynamics ERP and CRM landscape.
Nadella promises that Microsoft is turning the idea of CRM “upside down” from a siloed management tool to truly boosting sales productivity. No more stark customer activity: salespeople will be armed with a rich data model of broad, contextual, real-time customer intelligence, from among other sources:
- Social media, where customers readily share their sentiments, also their changes in title; and
- Customer service applications – obviously.
A more intelligent service application has considerable potential for partners. Microsoft is “experimenting with” an “intelligent agent” that troubleshoots customer issues, says Nadella; but it seems beyond experimentation. Microsoft’s own Contact Us page is such an agent with a cheerful-looking avatar. But consider that the best customer service call (or live chat) is no call or live chat in the first place. Those cost labor. An actively-learning virtual assistant deflects them, and frankly, customers prefer to help themselves. That’s quicker. The virtual assistant actively learns from the user’s questions, so presumably, the assistant is that much faster on a second or third call. If Microsoft makes its AI API as readily available as it claims, then, configuring virtual assistants is another much-sought service for partners to build upon.
Those APIs are the same “building blocks” that are behind Office, Cortana and Dynamics 365, and those blocks include the Cortana Intelligence Suite. If the Suite still seems abstract to Dynamics partners or users, Nadella showed a slide listing companies at which it is at work, including Rolls-Royce, Virgin Atlantic, and Accenture, to demonstrate the industry-spanning utility.
Still another platform is Microsoft’s Bot Framework, which as the company describes is “a toolkit that allows you to build intelligence into these new kinds of applications.” The company claims 40,000 developers have used the Bot Framework since the Build conference in April. Among the users is Uber, for intelligent facial recognition (matching driver to car for customer safety).
But can it handle fantasy football? Here Nadella invited surprise guest Deion Sanders of the San Francisco 49ers and the Dallas Cowboys on stage, to demonstrate a fantasy football app that Microsoft is building for the NFL. Sanders at first seemed underwhelmed, but conceded of a quarterback matchup “[It has] more knowledge than I did because I did not consider [that] in a dome, weather would not be a factor.”
So once again, Cortana knows how to play. But that analytical and learning capacity translate well to enterprise. Besides which, the NFL is an enterprise. So is Lowe’s, whose kitchen remodeling application built on the bot framework mines customer data like their Pinterest boards for design suggestions. The audience was treated to a rather long demonstration in which a young couple remodeled their kitchen. If the demonstration was rather tedious, the application’s understanding of the couple’s tastes was spot-on.
Hyperintelligence at hyper speed: FPGA
Legacy frameworks and nets are not sufficient for an AI revolution. But Microsoft claims it foresaw the need and has quietly built the world’s fastest AI supercomputer to manage it.
Here Nadella called Microsoft Distinguished Engineer Doug Burger to the stage, who brought a book. Not just any book, it was the 1440-page War and Peace, and in Russian. Burger used a translator on an ordinary laptop to translate it into English, all in under three seconds.
The laptop was ordinary consumer technology, but it ran on the Azure-based network conceived by Microsoft in 2012 as “Project Catapult.” The result is a return to silicon versus CPUs, with Field-Programmable Gate Arrays (FPGAs) in the Microsoft Cloud; these are ultra-high capacity programmable hardware with eight times the processing speed of traditional CPUs. With the addition of that capacity, Azure is now the world’s fastest AI supercomputer, claimed Burger, capable of processing 25 gigs per second with a 10X reduction in latency. “This new architecture effectively builds FPGA into a global hyperscale cloud, which will change what’s possible for AI,” said Burger. The company has quietly deployed FPGAs in 15 global data centers over the last two years.
So yes – AI is real, has enterprise utility, there is the compute power to handle it and the APIs available to take advantage of it. For Dynamics, the proving ground for native AI will be CRM; but it is already informing Dynamics AX, NAV and GP for companies like Rolls-Royce and Accenture.