Microsoft Paints Picture of Helpful A.I., Deployed to Benefit Humans

Microsoft laid out a higher purpose for its ongoing development of what could someday amount to a so-called general artificial intelligence: helping people and healing the planet.

A flurry of announcements Wednesday touched on the business value of the suite of A.I. technologies the company is racing to develop alongside competitors including Google-parent Alphabet (NASDAQ: GOOG), Amazon (NASDAQ: AMZN), and IBM, as well as a host of well-funded startups and academic institutions.

But top corporate leaders, including Microsoft (NASDAQ: MSFT) executive vice president Harry Shum, also staked out a moral and ethical grounding for Microsoft as it devotes ever more resources to A.I. They add detail to an initial set of principles set forth by CEO Satya Nadella last year, and generally hew to principles established by the broader A.I. research community and the Partnership on A.I. to Support People and Society, which Microsoft co-founded.

“We are responsible for building A.I. advances that amplify human ingenuity, and also that reflect our shared societal values and expectations,” Shum writes. “The A.I. tools and services we create must assist humanity and augment our capabilities.”

Last fall, Shum, pictured above, took the helm of a new group within Microsoft dedicated to A.I. research and products. It now numbers some 7,500 computer scientists, researchers, and engineers, he writes. That group now includes a dedicated team within Microsoft Research, called Microsoft Research A.I. (MSR A.I.), headed by Eric Horvitz and based at Microsoft’s corporate headquarters in Redmond, WA. Some 87 people are listed on the MSR A.I. staff page.

Eric Horvitz

Horvitz

Horvitz, founding co-chair of the Partnership on A.I., will also lead a new Microsoft ethics advisory panel, reporting directly to Nadella, and intended to be a check on A.I. technology development within the company, as Bloomberg reports.

The MSR A.I. lab will, among other things, attempt to create an integrated discipline, from the silos that have emerged in the field, focused on discrete human-like capabilities, including perception, natural language processing, and learning.

It is the unification of those capabilities in a single system that could begin to constitute a general artificial intelligence—which would posses something akin to common sense and be capable of addressing any challenge it might encounter: “complex, multifaceted tasks,” Shum writes. “For example, using this approach we can create methods and systems that understand language and take action based on that understanding.”

That would represent a step-change from today’s state-of-the-art A.I. systems, which can surpass human capabilities in certain narrow contexts: playing chess or Go, for example.

“We believe A.I. will be even more helpful when we can create tools that combine those functions and add some of the abilities that come naturally to people,” Shum writes.

While Shum describes rapid progress in A.I., he also acknowledges significant shortcomings that must be addressed, such as the so-called black box problem. For many machine learning systems today, their creators, let alone their users, often have no insight into how these systems reach their conclusions.

“The people who use those tools should be able to understand how they work and what data they rely on,” Shum writes. (One of the projects under the MSR A.I. group is introduced as “Friends Don’t Let Friends Deploy Black-Box Models: The Importance of Intelligibility and Transparency in Machine Learning.”)

He also calls for development of A.I. technologies that are “as inclusive and unbiased as possible.”

During yet another week of dire climate news (see David Wallace-Wells’ piece in New York magazine, The Uninhabitable Earth, and the trillion-ton iceberg that just cracked free of the Larsen C ice shelf), Microsoft also announced that it would devote its A.I. resources to solving environmental problems.

Specifically, the company intends to provide cloud computing, A.I. technologies, and training to researchers working on “water, agriculture, biodiversity and climate change” issues. The company is committing $2 million in its next fiscal year.

Benjamin Romano is editor of Xconomy Seattle. Email him at bromano [at] xconomy.com. Follow @bromano

Trending on Xconomy