There is a saying in the field of artificial intelligence: “Hard things are easy; easy things are hard.” Called Moravec's paradox, after Hans Moravec, founder of robotics company Seegrid, it is explained in detail in a recent book by computer science professor Melanie Mitchell entitled Artificial Intelligence: A Guide for Thinking Humans. Activities that most people find very hard, such as playing chess or doing higher mathematics, have yielded fairly readily to computation, yet many tasks that humans find easy or even trivial resist being conquered by machines.
Twenty-five years ago Garry Kasparov became the first world chess champion to lose to a computer.* Today computer programs can beat the world's best players at poker and Go, write music and even pass the famous Turing test—fooling people into thinking they are talking to another human. Yet computers still struggle to do things most of us find easy, such as learning to speak our native tongue or predicting from body language whether a pedestrian is about to cross the street—something that human drivers do subconsciously but that can stymie even the most advanced self-driving cars.
AI researchers will tell you that chess turned out to be comparatively easy because it follows a set of rigid rules that create a finite (albeit large) number of possible plays. Predicting the intentions of a pedestrian, however, is a more complex and fluid task that is hard to reduce to rules. No doubt that is true, but I think there is a bigger lesson in the AI experience that applies to more urgent problems. Let's call it the vaccine-vaccination paradox.
On supporting science journalism
If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
Anyone familiar with biology is hugely impressed by the agile scientific work that in under a year yielded astonishingly effective vaccines to fight COVID-19. Both the Moderna and the Pfizer-BioNTech vaccines use messenger RNA (mRNA) to deliver instructions to cells to generate the spike protein found on the novel coronavirus, which prompts the body to make the antibodies needed to fight an actual infection. It is a brilliant piece of biotechnological work that bodes well for similar uses of mRNA in the future.
Yet even several months after those vaccines were cleared for use, it is extremely hard to get the American population fully vaccinated. In the U.S., the difficulties have included the vexed politics of the past year, but the logistical challenges turned out to be great as well. Before the vaccines were authorized, some health experts were concerned that there might not be enough vials and syringes or cold storage. Others noted the problem of vaccine hesitancy. And since the vaccines became available, a host of new problems, including such quotidian tasks as scheduling, have plagued the program. The hard task of creating a vaccine proved (relatively) easy; the easy task of vaccination has proved very hard.
Maybe it is time to rethink our categories. We view chess as hard because very few people can play it at a high level, and almost no one is a grand master. In contrast, there are nearly four million nurses in the U.S. alone, most of whom presumably know how to deliver inoculations. If we had to, nearly all of us could probably learn to drive a truck to deliver vaccines. But this perspective confuses difficulty with scarcity. As the AI example shows, many things that all of us can do are in some respects remarkably difficult. Or perhaps we are conflating what is difficult to conceive with what is a challenge to do. Quantum physics is conceptually hard; administering 600 million shots in a large, diverse country with a decentralized health system is a staggeringly difficult practicality.
We call the physical sciences “hard” because they deal with issues that are mostly independent of the vagaries of human nature; they offer laws that (at least in the right circumstances) yield exact answers. But physics and chemistry will never tell us how to design an effective vaccination program or solve the problem of the crossing pedestrian, in part because they do not help us comprehend human behavior. The social sciences rarely yield exact answers. But that does not make them easy.
When it comes to solving real-life problems, it is the supposedly straightforward ones that seem to be tripping us up. The vaccine-vaccination paradox suggests that the truly hard sciences are those that involve human behavior.”
*Editor’s Note (7/1/21): This sentence was edited after posting to correct the description of the “first” that was achieved when a computer beat Garry Kasparov at chess.