This is the web version of Eye on A.I., Fortune’s weekly newsletter covering artificial intelligence and business. To get it delivered weekly to your in-box, sign up here.
Convenience store giant 7-Eleven revealed last week during the Re-Work A.I. conference in San Francisco that it had opened an experimental cashierless store in Dallas in November. Shahmeer Mirza, a 7-Eleven machine learning engineer, said people can pick up what they want and walk out, much like what shoppers can do at Amazon’s Go growing chain of cashierless stores.
Computer vision technology tracks what customers take and then, in theory, ensures that they’re charged.
Meanwhile, McDonald’s said that it’s testing kiosks at a small number of U.S. restaurants that let customers order Happy Meals and shakes by talking to a digital assistant. It’s part of a broader push by McDonald’s to incorporate advances in computers that understand language into its franchises and into its mobile apps, said German Parisi, director of applied A.I. for McDonald’s research and technology unit.
As Parsi explained, McDonald’s views A.I. “as a driving force” that can “improve the customer and employee experience.” Parsi joined McDonald’s when it bought the natural language processing startup Apprente in the fall, and he is now helping the company build a new A.I. lab in Mountain View, Calif.
For anyone waiting for 7-Eleven and McDonald’s to detail the financial rationale of their A.I. projects, this was not the time. They said little on stage about the cost or potential payoff. Instead, the companies took the opportunity to highlight their “wow, that’s cool” technology—just like Amazon, Google regularly do. But it will likely be years before 7-Eleven and McDonald’s A.I. bets pay off.
As Fortune has previously reported, consulting and analyst surveys indicate that most businesses are struggling with their A.I. projects and expect any significant financial returns to be years down the road. Moreover, in many cases, those payoffs are turning out to take far longer than initially expected.
One top data scientist from a major retailer told Fortune on the conference sidelines that it’s a big challenge for corporate data-crunching staff to convince the business-side that A.I. projects are worth it. Technologists have difficulty conveying the significance of A.I. to financial staff, who prefer business projects that increase revenue or profits in the short term.
Less tangible aspects of A.I.’s impact on a company—like helping it be perceived as technologically savvy, which can help with recruiting—are difficult for accountants to quantify.
Regardless, companies like 7-Eleven and McDonald’s, are making big A.I. bets. Although A.I.’s financial benefits are still debatable (at least in the short term), it’s clear that they don’t want to be left behind.
A.I. IN THE NEWS
Apple says ‘no’ to Project Maven. When Apple recently bought the A.I. startup Xnor.AI earlier this year, the iPhone-maker inherited the company’s work with the Pentagon’s Project Maven initiative, which encompasses the use of A.I. to analyze imagers collected by military drones. Now, tech publication The Information reported that Apple terminated the startup’s work with Project Maven, implying that Apple has no interest in pursuing lucrative military contracts like Amazon or Microsoft.
Facebook pays again for privacy lapses. Facebook paid $550 million to settle claims that the social networking giant violated Illinois law governing the collection of biometric data, Bloomberg News reported. The lawsuit alleged that Facebook stored images of people’s faces without their consent; the facial data was used to power the company’s A.I.-powered Tag Suggestions feature that could suggest the names of people in photos.
Intel nixes Nervana chips. Intel plans to phase out its Nervana Spring Crest computer chips used for data training in favor of chip technology offered by Habana Labs, which Intel bought in December for $2 billion, reported tech publication The Register. The article states that it’s “no doubt, a blow for the folks at Nervana,” which Intel bought in 2016 and pitched as key to the company’s A.I. chip efforts. Analyst Karl Freund of Moor Insights & Strategy wrote, "I suspect these customers pointed to Habana as the preferred platform that can compete with Nvidia."
Who needs a map? Facebook A.I. researchers developed a reinforcement learning system—in which computers learn through repetition—that could robots navigate spaces without requiring a digital map, MIT Technology Review reported. The research could be useful to creating more capable robots that can move autonomously without running into obstacles.“As ever, the team doesn’t know exactly how the AI learned to navigate, but a best guess is that it picked up on patterns in the interior structure of the human-designed environments,” the article said.
TELL ME ABOUT YOUR DATA
Multiple startups have recently emerged that are selling software and services that help companies label and annotate their data used to train A.I. systems. Fortune takes a look at the data-labeling startup Labelbox, which just received $25 million in funding. “Where the rubber hits the road is what does the data these A.I. systems are trained on look like?” Scale AI CEO Alexandr Wang said. “Is that data biased? Is that data high quality? Does that data have noise? Is that data comprehensive?”
EYE ON A.I. RESEARCH
Deep learning in space. Researchers from the University of Oxford and Samsung’sA.I. unit published a paper describing how deep learning could benefit various scientific activities in space, including analyzing satellite imagery and improving spacecraft operations like communications and navigation. “As space devices become more affordable to launch and their hardware becomes more powerful to run non-trivial workloads, deep learning in space will continue to grow as a topic within mobile and embedded machine learning,” the authors wrote.
Helping drones understand what they see. Researchers with the German Aerospace Center, Technical University of Munich, and The Institute of Electrical and Electronics Engineers published a paper and related dataset intended to help researchers create more capable A.I. software that can help drones analyze images in real-time. The researchers compiled and annotated 290 hours of video of scenes captured in the air that showed various events like floods, fires, mudslides, and traffic congestion. The authors said that the “proposed dataset serves as a new challenge to develop models that can understand what happens on the planet from an aerial view.”
FORTUNE ON A.I.
Founder of Google’s Moonshot Factory and Udacity makes a big new bet—By Jonathan Vanian
How marketers are increasingly using A.I. to persuade you to buy—By David Z. Morris
Privacy, bias and safety: On facial recognition, Berlin and London choose different paths—By David Meyer
After eight years, what did IBM’s outgoing CEO accomplish?—By Aaron Pressman
Warehouse party. The New York Times visited a German warehouse belonging to electrical parts manufacturer Obeta and saw advanced robots that were able to pick and sort objects blazingly fast. The robots, built by the startup Covariant, were outfitted with reinforcement learning technology, which allows them to “pick and sort more than 10,000 different items.” A day after the story published, the Times reported on the possible negative and beneficial effects of A.I. and advanced robotics to the workforce. Advances in A.I. have given rise to companies like Uber and Grubhub, that can use the technology to “monitor and control workers, while still maintaining a formal arms-length relationship that skirts federal employment status guidelines. Some of these issues are explored in this piece by The Atlantic, which probes a harrowing lawsuit and examines whether “Uber’s system—which uses algorithms to manage drivers’ ride loads—is a form of control that should designate it as an employer.”