To seek the origins of Microsoft’s interest in artificial intelligence, you need to go way back–well before Amazon, Facebook, and Google were in business, let alone titans of AI. Bill Gates founded Microsoft’s research arm in 1991, and AI was an area of investigation from the start. Three years later, in a speech at the National Conference on Artificial Intelligence in Seattle, then-sales chief Steve Ballmer stressed Microsoft’s belief in AI’s potential and said he hoped that software would someday be smart enough to steer a vehicle. (He’d banged up his own car in the parking lot upon arriving at the event.)
From the start, Microsoft Research (MSR for short) hired more than its fair share of computing’s most visionary, accomplished scientists. For a long time, however, it had a reputation for struggling to turn their innovations into features and products that customers wanted. In the ’90s, for instance, I recall being puzzled about why its ambitious work in areas such as speech recognition hadn’t had a profound effect on Windows and Office.
Five years into Satya Nadella’s tenure as Microsoft CEO, that stigma is gone. Personal determination on Nadella’s part has surely helped. “Satya is–let’s put it very positively–impatient to get more technology into our products,” says Harry Shum, Microsoft’s executive VP of artificial intelligence and research. “It’s really very encouraging to all of us in Microsoft Research.” That’s a lot of happy people: more than 1,000 computer scientists are in MSR’s employ, at its Redmond headquarters as well as Boston, Montreal, Beijing, Bangalore, and beyond.
CEO determination, by itself, only goes so far. Microsoft has gotten good at the tricky logistics of identifying what research to leverage in which products, encouraging far-flung employees to collaborate on that effort, and getting the results in front of everyone from worker bees to game enthusiasts.
Get your 70-740 Exam Dumps Question And Answers | RealExamDumps.com
For the record, Shum argues that Microsoft’s old rep for failing to make use of its researchers’ breakthroughs was unfair–but he doesn’t deny that the company is much better at what he calls “deployment-driven research” than in the past. “The key now is how quickly we can make those things happen,” he says.
Judging from my recent visit to its campus, Microsoft is making things happen at a clip that’s only accelerating. I talked to Shum and some of his colleagues from across the company about the process of embracing AI as swiftly and widely as possible. And it turned out that it isn’t one process but a bunch of them.
MEETING OF THE MINDS
On the most fundamental level, ensuring that Microsoft AI innovations benefit Microsoft customers is about making sure that research and product teams aren’t siloed off from each other. That means encouraging teams to talk to each other, which Microsoft now does in a big, organized fashion. Every six months or so, for instance, an event called Roc is devoted to cross-fertilization between research efforts and Office product development.
“We have those two-, three-day workshops where we have 50 people from Microsoft Research, 100 people from Office, all coming together,” says Shum. Everybody shares what they’re working on, and the whole affair ends with a hackathon.
Harry Shum [Photo: courtesy of Microsoft]
Another ongoing meeting of the minds, the Distinguished Engineering Leadership Lecture Series, brings executives responsible for products to the Microsoft campus’s Building 99, where MSR is headquartered. “I say, ‘You have to come in and do three things for me,'” Shum says. “Number one is tell us your product roadmap. Number two is [list] 10 things you need Microsoft Research to solve for you. And then number three is before you leave the building, commit to one or two projects that we will work on together.”
More than a decade after Office eradicated Clippy, it still wants to detect if you’re performing a task where AI might be useful. It just wants the experience to be subtle rather than intrusive. As Ronette Lawrence, principal product planning manager for Office, says, “One of our core principles is making sure that the human stays the hero.”
PowerPoint’s Design Ideas feature can use AI to dress up your slides–but it works hard to stay out of your way unless you want to use it.
Lawrence says that nearly everything Microsoft is adding to Office these days has an element of AI and machine learning to it. In PowerPoint, for instance, the company wants AI to be “the designer that works in the cloud for you.” If you’re using a pen-equipped PC such as Microsoft’s own Surface, PowerPoint can convert your scrawled freehand words and shapes into polished text blocks and objects. And if the software notices that you’re entering a sequence of dates, it will realize that it might make sense to lay them out as a timeline.
Instead of shoving unsolicited advice in your face, however, “we’re careful to make sure that these kind of suggestions come out as a whisper,” says Lawrence. The “Design Ideas” feature analyzes your presentation in process and shows thumbnails of possible tweaks–such as that timeline layout for a sequence of dates–to the right of your slides. They’re equally easy to implement or ignore.
Though many Office features are dependent on Microsoft Research’s latest work, certain brainstorms make their way out the lab more easily than others. “Some of it feels like science fiction,” says Lawrence of AI in its raw form. “And other [examples] feel closer to product readiness.”
During one workshopping session between the Office product team and MSR, the fact that people typically rough out Word documents and fill in holes later–or ask coworkers to do some of the filling–came up. What if Word helped wrangle the process?
The first Office users to get access to this to-do feature in its initial form are Windows and Mac users who have signed up for Office’s early-adopter program. (It’s due for general release by the end of the year.) Oftentimes, however, new AI functionality shows up first in Office’s web-based apps, where it’s easier for Microsoft to get it in front of a lot of people quickly–and learn, and refine–than if the company needed to wait for the next release of Office in its more conventional form.
“It’s pretty important for us to listen to feedback and see how people are using it to train a model,” says Lawrence. “That is part of the new era of Microsoft, where it isn’t just about the usability of the functionality when you release it. The web gives us that feedback mechanism.”
A current series of online ads is devoted to showing that the Office 365 service has an array of handy features that are absent in Office 2019, the current pay-once version of the suite. All these features leverage AI, but the ads don’t mention that. After all, the human is supposed to be the hero
No comments:
Post a Comment