Eye on A.I.— Artificial Intelligence: From Mars to Earth
If you need convincing that artificial intelligence will transform the world, I’d like to take you on a trip to Mars.
Well, not the planet, but Amazon CEO Jeff Bezos’s annual invite-only MARS conference last week in Palm Springs that takes its name from its focus on machine learning, automation, robotics, and space. Over 200 of the world’s leading scientists and technologists gathered to discuss their latest far-out research, a nerve-racking experience for those who presented in front of Bezos himself.
A.I., and its ability to make sense of data, was a common theme. But while it’s easy to dream about the future of A.I., and all the benefits it will supposedly bring, our present day version has room for improvement.
Speakers covered topics like the future of personalized medicine, the use of drone sailboats to forecast catastrophic storms, and the possibility of sending humans to Mars. All of these scenarios depend on computers crunching enormous amounts of data.
As Daphne Koller, co-founder of online learning service Coursera, explained, the cutting-edge analysis used to parse data collected by computers and sensors will likely bleed further into other areas—in her case the biological sciences. This means breakthroughs in healthcare, finance, and any industry that’s informed by data, which is to say, every industry.
At least that’s the hope.
But many conference attendees, including A.I. researchers, executives, and even artists, told me they’re concerned about some of A.I.’s recent stumbles. Consider the problem of A.I. trained on datasets in which certain ethnic groups are underrepresented. This problem can lead to facial-recognition software that works better on people with lighter than darker skin.
While the conference clearly inspired the technologists and scientists to think big, their casual conversations on the sidelines revealed that they are still scratching their heads over many facets of artificial intelligence and the potential problems the technology creates. It’s a confusing, complicated world, and these top researchers are still as perplexed as many of us civilians about how to solve these riddles.
But ultimately, these top scientists, like many business leaders, are hopeful that A.I. will lead to a better future.
As the curator for Fortune’s new weekly Eye on A.I. newsletter, my goal is to give you an overarching view of artificial intelligence, showing you where it excels and where it fails. That includes all of its derivatives like machine learning, deep learning, and neural networks. Don’t worry, I’ll explain what they are and how they’re different from one another.
I’ve covered business and technology for nearly a decade, and am still learning new things every day about artificial intelligence. I’m looking forward to exploring and sharing this critical topic with you.
Please let me know what you want to hear about, what fascinates you, and what it is about A.I. that keeps you up at night. You can sign up for Eye on A.I. here.
A.I. IN THE NEWS
Unhealthy A.I. Insurance companies, medical device makers, and hospitals could use A.I. to overcharge patients and other organizations, reports The New York Times. Harvard and M.I.T. researchers worry that healthcare companies may use so-called adversarial attacks to fool A.I. systems that are trained to recognize the contents of medical images. “If an insurance company uses A.I. to evaluate medical scans, for instance, a hospital could manipulate scans in an effort to boost payouts,” the Times wrote.
India’s data-labeling boom. India is home to several companies focused on labeling and annotating data, a necessary task for machine learning to work well, according to India tech news outlet Factor Daily. Some of the companies profiled in the story include iMerit, which labels data for customers like Microsoft and eBay, and Playment, whose clients include Samsung and Alibaba.
About our China A.I. plans… A top Pentagon official plans to meet with Google to discuss his concerns about the search giant’s A.I. efforts in China, Politico reported. Gen. Joseph Dunford said during an event that Google’s work “indirectly benefits the Chinese military and creates a challenge for us in maintaining a competitive advantage.”
Facebook’s A.I. can’t recognize violent videos. Facebook said that its A.I. systems failed to detect a video of the recent mosque shootings in New Zealand. One reason was because the social networking giant doesn’t have enough similar shooting videos to train its A.I. systems to automatically recognize violent attacks, an executive explained.
Colorado companies are getting “smarter.” Colorado is experiencing a rise in startups specializing in A.I. that helps workers (although their technology may replace some as well), reports Colorado Public Radio. One company, AMP Robotics, uses deep learning to train robots to recognize and sort through garbage at waste facilities.
CUTTING THROUGH THE HYPE
Invest in the basics. Although many companies want to quickly put machine learning to use, it’s important that they first have some basic data crunching technology in place. Technical publisher O’Reilly hammered that point home in a report that found companies interested in machine learning are spending most of their time building or evaluating tools for collecting, cleaning, and storing data — all of which is a critical for getting quality results.
Chatbots are mostly all talk. Machine-learning powered chatbots were all the rage a few years ago, with some companies saying that Siri-like digital assistants could replace customer service workers. Trade publication DevOps.com, however, believes that chatbots are still an “aspirational” technology that are only able to respond to the most rudimentary and simplest of questions.
EYE ON A.I. TALENT
Pinterest has hired Jeremy King to be the social media service’s head of engineering. King, formerly the chief technologist at Walmart, will help oversee Pinterest’s data-intensive technology, including computer vision projects and personalized shopping recommendations.
Former Pinterest CTO Vanja Josifovski has joined Airbnb to become the CTO of home-rental and travel service’s core business, Airbnb Homes. Airbnb told tech news site VentureBeat that the hire is intended to bring the company’s tech projects, like artificial intelligence, under one leader.
Sacramento Bee reporter Brad Branan is leaving journalism to become a data scientist for the state of California. He said on Twitter that he would lead an unspecified study for California’s Department of Veterans Affairs.
EYE ON A.I. RESEARCH
A.I. in the air. Iowa State University researchers published a paper on how A.I. could benefit the airline industry. Their paper describes how the A.I. technique of reinforcement learning—in which computers learn by trial and error—could help airlines better predict their revenue by taking into account variables like people cancelling their trips or airlines overbooking flights.
Fine-tuning cellular networks with A.I. University of Texas researchers published a paper detailing how reinforcement learning could be used to calibrate certain antenna settings to improve cellular networks. This new approach may be better than current methods in which telecom workers manually tweak the settings using their own “trial and error,” the paper said.
Deep learning as a parking attendant tool. Researchers from M.I.T.’s Senseable City Lab published a paper detailing the use of deep learning for monitoring cars in parking lots. The researchers, who used deep learning techniques to analyze three days of video footage of a street parking near M.I.T.’s campus, said their system was able to better guess the number of cars in nearby parking lots than conventional methods.
FORTUNE ON A.I.
Google Celebrates Bach's Birthday With Its First A.I.-Powered Doodle - By Natasha Bach
To Design Great A.I., Get 'Covered in Blood' - By Timothy McDonald
Nvidia Has a Cheap A.I.-Focused Computer for the Do-it-Yourself Crowd - By Aaron Pressman
Buying Stock in Tech Building Blocks - By Robert Hackett
Make Your March Madness Bracket Pick With Big Data's Help - By Chris Morris
Most Companies Aren't Ready for California's Tough New Privacy Law - By Danielle Abril
Who owns A.I. creations? As A.I. matures, it’s likely that computers will get better at making original art, songs, articles, and other creative work that people have typically created. But law policy expert Krista Cox explains in the legal publication Above The Law that these A.I. creations may test the limits of existing intellectual property and copyright laws in the U.S.
Consider an A.I. system that’s been trained on a corpus of copyrighted novels to create its own literature. Cox writes that it might be impossible to obtain permission to use every single book to train the neural networks, the software at the heart of deep learning. She asks, “Would the use of copyrighted works in these training sets be considered a fair use?”
We’ll have to let the courts decide when the time comes.