Microsoft is teaching machines to read and understand context in pursuit of building smarter AI
Tech News

Microsoft is teaching machines to read and understand context in pursuit of building smarter AI

Microsoft is going past the ideas of picture recognition and machine studying to make synthetic intelligence smarter, and building methods that may read textual content, comprehend the context behind it and even ask and reply questions.

This effort is in half the fruits of its acquisition in January of Maluuba, a Montreal-based firm that makes use of deep studying to develop natural-language understanding. According to a Microsoft blog post, two analysis groups, one on the firm’s Redmond, Wash., headquarters and one in its Beijing, China, analysis lab, are concerned in a competitors run by Stanford University that makes use of info from Wikipedia to take a look at how effectively AI methods can reply questions.

“We’re making an attempt to develop what we name a literate machine: A machine that may read textual content, understand textual content and then learn the way to talk, whether or not it’s written or orally,” Kaheer Suleman, co-founder of Maluuba mentioned in the weblog publish.”

This form of data may help individuals shortly get by dense materials reminiscent of libraries of authorized paperwork, automotive manuals and medical research.

Within the weblog publish, Microsoft’s Allison Linn writes that machine studying is tougher than ideas like picture recognition as a result of it requires data outdoors of the particular set of knowledge that the system is making an attempt to interpret.

Microsoft mentioned this type of know-how may revolutionize the fashionable search engine. It additionally may match into the voice assistant work that Microsoft, Amazon, Google, Apple and others are deeply investing in. Such context-heavy studying may give Cortana a lift because it seeks to unseat Amazon in the digital voice market.

In accordance to the weblog publish, the roots of Microsoft’s present machine studying efforts return greater than 20 years. The elevated energy that comes from cloud computing, giant swaths of out there knowledge to take a look at and advances in deep studying algorithms are enabling breakthroughs in this subject.

Whereas Microsoft cautions that its AI is nonetheless behind the comprehension of a knowledgable human, it is making nice strides. Microsoft final fall mentioned its researchers had examined software program that acknowledges human speech in addition to people do, with a record low error rate of 6.3 p.c. Twenty years in the past, the most effective price achieved was better than 43 p.c.

Related posts

Samsung’s Galaxy Z Fold 3 and Z Flip 3 revealed in new leak


Oculus users are getting a new metaverse home


‘Tools and Weapons’ at the GeekWire Summit: Hear Microsoft President Brad Smith on tech’s big issues