Source: MGM/Everett Collection

Here’s What Inspired Top Minds in Artificial Intelligence to Get Into the Field

Hint: They don’t want to destroy humanity

Artificial intelligence can seem pretty terrifying. On July 28, representatives from the Future of Life Institute presented a letter signed by Tesla Motors Chief Executive Officer Elon Musk, Apple co-founder Steve Wozniak, theoretical physicist Stephen Hawking, and other luminaries of the science and technology world who are urging a ban on autonomous weapons. They expressed concern over a “military artificial intelligence arms race.” That sounds a lot like a plot line to a science-fiction movie, and the fact that we’re having a real conversation about it shows how far AI has come in a brief period of time.

Just a few years ago, artificial intelligence was a field starved for funding, rife with skepticism, and distinguished not by its achievements but by its perennial disappointments. Now machines have the capability to learn, build things, answer questions, and yes, even harm people. But creating the malevolent machine from 2001: A Space Odyssey was not why top scientists pursued a career in AI. (The Stanley Kubrick movie did, however, have an impact on at least a couple of prominent members of the AI community.) We spent the past few months asking some of the field’s most renowned researchers and entrepreneurs what inspired them to pour their intellectual life into something that once seemed so unlikely and ominous.

HAL 9000

HAL 9000 in a scene from 2001: A Space Odyssey.

HAL 9000 in a scene from 2001: A Space Odyssey.

Source: MGM/Everett Collection

Yann LeCun, the director of artificial intelligence research at Facebook, remembers watching 2001: A Space Odyssey when he was about 10 years old. He was enthralled by the hyper-intelligent computer within the spacecraft. LeCun is “not interested, particularly, in how humans function” but says his obsession with developing AI stems from a belief that it could lead scientists to develop a theory for how cognition works, whether biological or digital. “The analogy I always use is: Birds fly, and so do airplanes,” LeCun says. “They use very different details of implementation, but the underlying principles of flight are the same, and that’s based on aerodynamics. What’s the equivalent of aerodynamics for intelligence like this? That’s the big question.”

Fusion power

Interior of the Alcator C-Mod tokamak at the MIT Plasma Science and Fusion Center.

Interior of the Alcator C-Mod tokamak at the MIT Plasma Science & Fusion Center.

Source: Mike Garrett/Wikipedia Commons

Like Lecun, Microsoft Chief Research Scientist Christopher Bishop saw 2001 as a teenager, and HAL 9000 made a similarly big impression. But Bishop decided to not go into AI at that point, because in the 1980s, the field wasn’t very creative. The ’80s version of AI simply programmed machines with instructions to carry out. So Bishop earned a Ph.D. in quantum field theory and then helped devise a novel kind of fusion reactor called a tokamak. Eventually, computers became fast enough that he could apply what’s called neural networks to research fusion power. Neural networks became the basis for modern AI, and now Bishop works on machine learning full time. “Recreating the cognitive capabilities of the brain in an artificial system is a tantalizing challenge, and a successful solution will represent one of the most profound inventions of all time,” he says.

Sci-fi novels

A copy of nineteen eighty four by George Orwell, right, is seen amongst a display of books published by the Penguin publishing house, part of Pearson Plc, at a bookstore in London, U.K., on Friday, April 5, 2013. Bertelsmann SE’s Random House won European Union approval to buy Pearson Plc’s Penguin unit to create the largest book publisher in the U.K. and the U.S.
Photographer: Chris Ratcliffe/Bloomberg

Like seemingly everyone else in AI, Yoshua Bengio spent his youth programming computers and reading science fiction. That combination led him to daydream about whether it’d possible to create intelligent machines. Once Bengio began studying AI in earnest, the University of Montreal computer science professor, along with LeCun and Google’s Geoff Hinton, kept deep learning research alive during a time when AI was out of style. “The reasons I do it have nothing to do with the fashion,” Bengio says. “I do it because I really believe that this is the right direction to reach AI, and I have my intuitions about how brains work.”

A game that tries to guess what number you’re thinking of

BBC Micro computer
Photographer: Sam Radmall/Flickr via Bloomberg

Andrew Ng, the chief scientist at the research arm of Chinese search giant Baidu, started learning to program at age 4, and within a couple of years, he was developing computer games. “I was six years old, writing a program that could play a number guessing game: You would come up with a number, and it would try to guess what it was, and you’d tell it if its guesses were too high or too low each time,” Ng says. “I was amazed that with just a few lines of software, a computer could be made to play this simple game so well. As an adult, I now recognize it to be a simple ‘binary search’ algorithm, but to the 6-year-old me, it was magical.”

Flames on a computer screen

Don't Panic computer
Photographer: Sarah Klockars-Clauser/Flickr via Bloomberg

Matt Zeiler had planned to become a nanotechnologist, but during his second year at the University of Toronto, his adviser showed him a video of a flame flickering, which had been generated by a computer-vision system. “I was like, ‘OK, this is too cool. I’ve got to get into this,’” says Zeiler, who’s now the CEO of Clarifai, a startup that applies large-scale machine intelligence to analyze videos and images. “It was nothing like any program I’d seen in the first two years, and I knew it was going to have a huge impact, because it was going to do much more complex things than you can describe via programming it.” At Zeiler’s AI startup, he splits his time between hiring and programming his self-teaching computers. “I’m wearing multiple hats while the machines are learning in the background,” he says.

A digital fortune teller

Fortune teller
Photographer: Ed Schipul/Flickr via Bloomberg

The first AI-like program Babak Hodjat designed went viral. He was in college and made a game that asked people to input their age, gender, height, and other demographic information to have their fortune read. The system learned over time how to link specific inputs to certain outcomes. Soon, someone asked Hodjat for a copy on a 5¼-inch floppy disk, and his fellow students were going crazy over the software, which seemed fairly simple and unimpressive to him. “Before I knew it—and to my annoyance—the fortune-telling AI was a hit in our college,” he says. Hodjat is now the chief scientist at Sentient Technologies, which makes massively distributed AI tools for various industries. He certainly wouldn’t describe his work today as simple.