Skip to Content

Google and Microsoft Want Every Company to Scrutinize You with AI

The tech giants are eager to rent out their AI breakthroughs to other companies.
August 1, 2016

When some patients of Dartmouth-Hitchcock Medical Center in New Hampshire step on their bathroom scale at home, Microsoft’s computers know about it. The corporation’s machines also get blood pressure readings taken at home. And they can even listen to calls between nurses and patients to gauge a person’s emotional state. Microsoft’s artificial intelligence software parses that data to try and warn patients and staff of emerging health problems before any human notices.

The hospital is previewing both the future of health care and of Microsoft’s business. It’s using a suite of new “cognitive” services recently added to Microsoft’s cloud computing service, called Azure. The company says renting out its machine-learning technology will unlock new profits, and enable companies of all kinds to subject their data—and customers—to artificial-intelligence techniques previously limited to computing giants.

Illustration by Oscar Bolton Green

“Customers are going to mature from classic cloud services to services that use elements of machine learning and AI,” says Herain Oberoi, director of product management at Microsoft, who oversees the company’s cloud machine-learning services. “Every company I talk with has someone extremely senior tasked with thinking about how to make this technology work for them.”

Microsoft’s competitors Google, IBM, and Amazon are making the same bet. Google announced in June that it had invented a new kind of chip to accelerate machine-learning software and make its cloud services more competitive. The company lags Amazon and Microsoft in the cloud market, and CEO Sundar Pichai has said machine-learning services provide a way for Google to differentiate itself. Amazon’s cloud division, Amazon Web Services, launched its first machine-learning cloud services last year, and in June the group’s head, Andy Jassy, pledged to expand them significantly in the coming months.

Amazon and its largest competitors stepped up their investments in machine-learning technology in recent years after breakthroughs in software that can be trained to do tasks such as interpret photos or speech (see “10 Breakthrough Technologies 2013: Deep Learning”).

Some of the first consumer products to take advantage of those breakthroughs were Amazon’s Alexa voice-operated home assistant and Google’s new Photos service, which understands the content of images and has more than 200 million users. Adding machine learning to the cloud services that corporations already use to outsource tasks such as data storage and analysis is seen as another way to extract money from the technology and enhance the very lucrative market. IDC estimates that corporations spent almost $70 billion with cloud providers last year, and predicts that will double before the end of the decade.

Rob Craft, who leads product management for Google's cloud machine-learning offerings, says that most companies are in a position to benefit from machine learning right away because they have a lot of data on hand about their operations, business, and customers. “Our goal is to help them have more direct value from that data,” he says.

The most straightforward of the new services offered by Google and others do things like describe the content of images, transcribe audio files such as phone calls, extract key terms from text, or translate text between languages. Although seen as lagging behind Google in machine-learning technology, Microsoft and IBM have so far rolled out the broadest range of such services, known as APIs.

Microsoft has an API that tries to decipher facial expressions, for example. IBM has one that assesses the personality of the author of text such as social media posts. Marketing company Influential uses it to help brands such as Corona and Red Bull identify the most useful social media users for promotional efforts. Different APIs can be combined. For example, a company could set up a system that spots its logo in social media images, notes the facial expression of any people in the photo, and extracts key terms from any accompanying text.

Many key software components needed to build the kind of machine-learning systems that Google and others hope will be so valuable are free (see “Facebook Joins Stampede of Tech Giants Giving Away AI Technology”). But Jimoh Ovbiagele, cofounder and chief technology officer at startup ROSS Intelligence, which provides software that speeds up legal research to major law firms, says that the time and expense of building and operating a top-notch machine-learning system means many companies are better off renting the technology.

“It makes sense to stand on the shoulders of giants,” says Ovbiagele. ROSS’s ability to understand legal questions is built on IBM’s suite of language processing technology, some of which originated with the Watson computer that beat two Jeopardy! champions in 2011.

Chris Curran, chief technologist with PwC, says most large corporations are still far from ready to spend significantly on machine-learning services, though. He estimates about three quarters are in “watch and learn” mode, waiting to see what these new capabilities offer.

And while the new services from Microsoft and others make it easy for non-technology companies like Dartmouth-Hitchcock Medical Center to use preprogrammed machine-learning systems, the technology is most valuable when customized for an organization’s specific needs, says Curran. Google and Microsoft’s image APIs are good at general assessments, such as whether a photo contains a cat or a skyscraper, for example. But a food manufacturer would get more value from a vision system able to spot specific defects in items on its production line.

All the cloud providers either already offer or have promised ways for customers to train algorithms on their own data, for their own problems. But creating customized artificial intelligence software can only be made so easy, says Curran. “You need to have the right people and expertise, and those are in short supply,” he says.

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

The problem with plug-in hybrids? Their drivers.

Plug-in hybrids are often sold as a transition to EVs, but new data from Europe shows we’re still underestimating the emissions they produce.

How scientists traced a mysterious covid case back to six toilets

When wastewater surveillance turns into a hunt for a single infected individual, the ethics get tricky.

Google DeepMind’s new generative model makes Super Mario–like games from scratch

Genie learns how to control games by watching hours and hours of video. It could help train next-gen robots too.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at with a list of newsletters you’d like to receive.