World Economic Forum Wants to Help Companies Avoid the Pitfalls of Artificial Intelligence: Eye on A.I.
The World Economic Forum, best known for its glitzy annual conference in Davos, Switzerland, wants to help companies avoid the potential pitfalls that come with deploying artificial intelligence.
Yes, A.I. promises to radically change how businesses operate by opening the door to innovations like driverless vehicles and robots that care for the elderly. But it could also exacerbate inequalities in society and lead to widespread job loss.
The WEF’s solution: A set of guidelines for corporate boards that spells out how companies can use A.I. responsibly.
“We found a lot of boards didn’t really understand A.I., and they were asked to make decisions about implementing A.I. in companies without any tools to do so,” Kay Firth-Butterfield, the WEF’s machine learning chief, told Fortune.
The WEF wants its so-called A.I. toolkit to answer questions like how companies can best implement A.I. in their businesses. The tip sheet will also highlight the importance for businesses to create A.I. ethics councils to monitor their use of A.I. and the public relations black eye and customer backlash companies face if they screw up.
Butterfield hopes the guidelines will help board members “understand a whole set of questions they need to be able to ask and get answers to.”
She and her team announced plans for the A.I. guidelines in January during the Davos summit. Since then, they have gathered feedback from companies and A.I. experts to finish the job.
The WEF plans to release a public version of its A.I. guidelines at next year’s Davos conference. The next step will be to start work on a similar A.I. tip sheet for company executives.
“The C-suite said, ‘What about us?’” Butterfield joked.
Previously, the WEF had made a big push to explain to companies the nuances of cloud computing, another hot technology that gained traction a few years ago. A.I., however, “is slightly more interesting,” Butterfield said.
One example of the technology’s potential downside, she said, involves hiring software that is supposed to speed up the recruitment process. If trained using a company’s previous hiring data, it may exacerbate gender or racial bias by only highlighting white males as the best candidates.
“If you don’t think about bias issues, those could have [negative] effects on your business,” Butterfield said.
EYE ON A.I. NEWS
Hiring A.I. talent in Canada may be easier. Tech companies are having an easier time hiring highly-skilled workers in Canada than in the U.S. because of Canada's more lax immigration policies, Time Magazine reported. One A.I. startup, Finn.AI, for example, “considered locating their new company in Silicon Valley, but ultimately chose Vancouver because they knew they would qualify for a start-up visa there, and that they would be able to quickly hire AI experts from around the world.”
Autism and A.I. Companies like Credit Suisse, Dell Technologies, and Microsoft have established “neurodiversity” programs that include hiring people with autism for A.I.-related jobs, under the belief that “Autistic workers are often hyper-focused, highly analytical thinkers with an exceptional proficiency for technology,” The Wall Street Journal reported. The newspaper said that many of the autistic workers “are capable of working long hours on repetitive AI tasks, such as labeling photos and videos for computer-vision systems, without losing interest.”
A.I. as a creator. The Financial Times reported on a “landmark challenge to the international patents regime” involving an A.I. system that created two designs, one of a “food container capable of changing shape” and the other of “a flashlight system.” The article explores the confusion within the legal community in establishing the devices’ creator, which the patent application attributes to “Dabus,” the computer system that developed the designs.
Sharing faces. Law enforcement agencies in California have “have the capability to run facial recognition searches on each others’ mug shot databases,” tech publication OneZero said. The article explains that tech company DataWorks Plus and its image-sharing service “puts the company in a powerful position in the nation’s largest state.”
AVOIDING FOOL’S GOLD
Patrick Riley, a principle engineer for Google’s accelerated science team, wrote an article in Nature about three pitfalls data scientists should avoid in machine learning. As Riley explains: “machine-learning tools can also turn up fool’s gold — false positives, blind alleys and mistakes. Many of the algorithms are so complicated that it is impossible to inspect all the parameters or to reason about exactly how the inputs have been manipulated.”
EYE ON A.I. HIRES
JPMorgan Chase hired Subhashini Tripuraneni as its executive director of machine learning. Tripuraneni was previously the head of artificial intelligence for 7-Eleven.
Government consulting firm Simple Technology Solutions hired Subhasis Datta to be its chief data scientist and practice lead for data science, machine learning, and artificial intelligence. Datta was previously the chief data scientist for federal consulting firm Analytica.
EYE ON A.I. RESEARCH
Adversarial healthcare A.I. Researchers at Ghent University in Belgium and Ghent University Global Campus in South Korea published a paper about using deep learning to create so-called adversarial examples that trick technology for recognizing ailments like breast cancer and eye disorders in medical imaging data. In one example, the researchers showed how their A.I. techniques covertly manipulated a photo showing breast cancer so that a medical-imaging system classified the photo as “healthy.”
A review of global A.I. ethics. Researchers from consulting firm Dovetail Labs and Princeton University published a paper examining the ethics of A.I. as it pertains to different countries and regions worldwide. One of the paper’s findings touched on how “people from low-and-middle- income countries are likely to be radically underrepresented in the datasets central to developing AI systems.”
FORTUNE ON A.I.
DeepMind’s Latest A.I. Predicts Kidney Injuries 48 Hours in Advance – By Jeremy Kahn
How Intel Hopes to Catch Rivals With Its Latest Chips – By Aaron Pressman
Examining Healthcare. Smithsonian magazine explored the current state of artificial intelligence in healthcare. Despite high hopes that A.I. can improve healthcare, there are still potential problems worth considering. For instance, the Smithsonian notes that “if A.I. services make cost-saving recommendations, human physicians and health care organizations may hesitate to take A.I. advice if they make less money as a result.” One of the most promising but less exciting ways A.I. can aid doctors, the article also explains, is by automatically entering patient data into electronic health records, a burdensome task for physicians and “a main factor behind physical and emotional burnout.”