At every speech he’s made so far, Microsoft’s new CEO Satya Nadella has talked about his vision of the future of computing. Having a vision is good for Microsoft, a company that seemed lost, struggling to find relevance in a world dominated by the new giants of Silicon Valley. Now the sleeping giant has awoken, first with Office on the iPad, then with new versions of Windows, then with Cortana, Surface 3, and on and on, five months of announcement after announcement.
Much of what Microsoft has delivered was born under Nadella's predecessor Steve Ballmer, but much of it has come from the groups touched by Nadella’s career, from Bing to Azure, to Office 365 and beyond. Much of it has also come from Microsoft’s secret sauce, the scientists and engineers of Microsoft Research.
Having a big vision is important. Microsoft used to talk about “a computer on every desk and in every home,” a vision it clearly succeeded in delivering. But what do you do when you’ve delivered that vision? What do you do when you’ve walked on the moon? It’s hard to find a new vision that will motivate 100,000 people from all around the world. You need to shoot for another moon, another dream that’s almost attainable, yet currently out of reach.
So what does that new vision entail? What does Nadella mean when he talks about “ubiquitous computing and ambient intelligence”?
The ubiquitous computing part is easy. We might call it the “Internet of Things” now, but back in the 90s we started talking about ubiquitous computing. Even then it was clear that the growing networked world would subsume the old computing paradigms, replacing hardware with software and connecting machine to machine, machine to human, and human to human. What had been discrete computers, lumps of metal and plastic on our desks, were going to go away, disappearing into the fabric of everyday life and work.
The computer was going to be ubiquitous. It was going to be everywhere and -- above all -- it was going to be everything. That vision got lost somewhere along the long and winding road from the research labs I worked in. But suddenly it’s here: embodied in smartphones, in tablets, in smart homes, and in our newly connected cars. It’s a world of WiFi, of Bluetooth-connected devices in our pockets, of processors with everything.
Those ubiquitous computers are disappearing into the background of our lives, even as the devices we access them from become handier. We don’t think about what happens when we turn on a light switch, and we don’t think what happens when we talk to Siri or Cortana. We’re surrounded by invisible computers; from the sensors that track buses and cars, to our phones and tablets, to the hyperscale data centers of the cloud.
The ubiquitous computing world is in our pockets, and just around the corner. It’s visible, understandable, and attainable. But what of the other part of Nadella’s vision, the world of “ambient intelligence”?
That’s the blue sky part of the vision, Microsoft’s new moonshot. It’s a vision that comes from Nadella’s work at Bing and at Azure, one where that hyperscale cloud works with the data generated by the ubiquitous computing fog we live in. To understand what that moonshot is building we need to look at the work that’s been going on in Microsoft’s research labs for more than two decades, and to the inspiration of one computing’s lost greats.
The loss of Microsoft Research’s Jim Gray in a sailing accident was a tragedy in many ways, but his legacy of research and thinking that has shaped much of the new Microsoft. A database pioneer, he focused much of his later research on big data, and on its effects on society and science. Describing a next step in science, Gray coined the phrase “the fourth paradigm” to talk about how data exploration was changing the way researchers approached science, exploring large amounts of data to find new and interesting results.
That fourth paradigm is at the heart of Nadella’s “ambient intelligence,” using insights from the ambient sea of big data to power a new generation of artificial intelligences. It’s those insights that are powering tools like Microsoft’s personal assistant Cortana, or the upcoming Skype language translator.
Microsoft Research head Peter Lee recently talked about some of the AI breakthroughs that were powering the new tools. Discussing the concept of “transfer learning,” he revealed that by training a speech recognition neural net on multiple languages, its performance improved with each new language -- even on previously trained languages.
We won’t get ambient intelligence without the big data we’ll get from ubiquitous computing. Cortana might tell me that it’s going to take me 10 minutes longer to get to a meeting than I’d budgeted, but it can only do that with access to the millions of sensors out there on the roads -- from the government data feeds to the ad hoc networks of phone GPS. Personalized traffic reports like this are an early benefit of the working with ambient big data sources; we’re no longer limited to hourly traffic reports on crackly AM talk radio stations.
Business users are getting access to another early form of ambient intelligence in the shape of the Power BI tools in Office 365. With cloud-powered big data analysis delivered to Excel, and natural language query building powered by the Bing engines, you can start to see the shape of an automated insight engine that can extract meaning from personal, local, regional, and global big data. It’s the future promised by apps like Project Oslo that infer context from our emails and documents, and then deliver the information we need, when we need it.
Managing context needs ambient intelligence, and the ability for AI to use that sea of big data to infer what we’re doing, and how. It’s a complex task, and requires understanding the difference between correlation and causation -- still one of the biggest problems in AI. But Microsoft's researchers seem to have found some routes to handling context, with neural networks that watch and learn before they start operating.
Peter Lee tells the story of the neural network that runs the elevators in Microsoft Research’s Redmond building, which spent months observing and learning the patterns of staff around the lifts before taking over control. Now, if you walk to the elevators, they open for you; if you’re walking past on the way to the canteen, or are just having a conversation, the doors stay closed. It’s learned to look at ambient behavior and work out what people want.
We’re living in a sea of information that’s becoming an ocean. As more and more sensors plug into the ubiquitous computing environment we’re building, the more data we’ll be generating. Building on the classic model of information theory, that data becomes information, the information becomes knowledge, which becomes insight, which becomes wisdom. Ambient intelligence is about tapping into that information hierarchy to give us the insights we need in our day to day personal and business lives.
It’s quite the moonshot. Despite the advances in the cloud, in machine learning, and in big data, we’re still a long way from a set of tools that can deliver on Nadella’s vision. But then, that’s the point of a big vision: it gives Microsoft’s reshaped carrier battlegroup a destination, and that gives 100,000 Microsoft employees a shared vision that can break down the barriers between the old Microsoft silos.
Nadella’s ubiquitous computing Microsoft isn’t the old Microsoft. It’s one where mobile and the cloud dominate, where Azure and an everywhere-Office dominate the old Windows hegemony (which itself starts to smear across the ubiquitous computing device spectrum), and where Bing and Microsoft Research navigate the complex road that leads to ambient intelligence. A brave new world, indeed.