The concept of artificial intelligence extends to antiquity, when Hephaestus, the Greek god of craftsmen and artisans, built metal automatons to assist him, or when Aphrodite brought Pygmalion’s ivory statue—whom we know today as Galatea—to life. With the exception of a few rudimentary human-made automatons, AI remained the subject of myth and fiction for millennia, but humans were not content to leave to the gods the ability to instill intelligence in inanimate objects. Humans wished to turn the process around and, as Pamela McCorduck put it in Machines Who Think, “… forge the gods.”
That began to seem a possibility in the 20th century, with the Dartmouth Summer Research Project on Artificial Intelligence in 1956 providing a significant boost. Research in AI proceeded fitfully, weathering two “AI winters” as funding rose and fell. But since the second AI winter faded in 1993, AI has progressed steadily in tandem with Moore’s law, which provides the compute power necessary to support machine learning and deep learning, named by MIT Technology Review as one of 10 technology breakthroughs for 2013.
With success comes concern. After all, “forging the gods” may have unintended consequences. Stephen Hawking, Stuart Russell, Max Tegmark, and Frank Wilczek warned against dismissing AI-capable machines as science fiction. “Success in creating AI would be the biggest event in human history,” they wrote in the May 1, 2014, edition of The Independent. “Unfortunately, it might also be the last, unless we learn how to avoid the risks.” They cited the potential dangers of autonomous weapons systems that choose their own targets, and they likened the arrival of AI to the arrival of a superior alien civilization from outer space.
At the time, many observers contended that Hawking and his coauthors were overreacting. Gary Marcus, a professor of psychology at New York University, wrote in The Wall Street Journal that computers can balance checkbooks and land airplanes but lack a teenager’s ability to learn to play video games. And Andrew Leonard in Salon wrote that machines don’t think—they just do what they are told.
Leonard may have had a point for traditional computing, in which computer scientists provide an algorithm and an input, and the machine generates an output. But with machine learning, humans provide inputs and outputs, and the computer creates its own algorithm. And brain-inspired deep learning, as described by Karim Arabi, vice president of engineering at Qualcomm, at the International Test Conference last October, will only increase the possibility for unintended results. “The computer will eventually outpace our intelligence and logic in all possible ways,” he said.
While Hawking, et al., sounded the alarm, a group of technologists including Elon Musk and Peter Thiel announced in December that they have committed $1 billion to a new initiative called OpenAI. According to CTO Greg Brockman and research director Ilya Sutskever, the goal of the nonprofit research organization is “… to advance digital intelligence in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate financial return.”
As an alternative to addressing narrow tasks through hand-coding (like teaching a computer to play chess), they see deep learning as the key to discovering general human-level intelligence algorithms. They write, “It’s hard to fathom how much human-level AI could benefit society, and it’s equally hard to imagine how much it could damage society if built or used incorrectly.” OpenAI, they say, will grow into an institution that can prioritize a good outcome.
There are few details so far as to how and when the organization will invest its $1 billion. And I share some of the skepticism of Leonard at Salon as to AI’s immediate threat. But it can only be beneficial to have a nonprofit like OpenAI providing inputs that can inform the efforts of for-profit companies.
Rick Nelson, Executive Editor.