上海花千坊爱上海

A string of state-directed, targeted mass killings left a bloody stain on the 20th century. A genocide more recent than the Holocaust is providing new insights into why some people join in such atrocities.

Adolf Hitler’s many accomplices in his campaign to exterminate Jews throughout Europe have justifiably attracted the attention of historians and social scientists. But a 100-day spasm of unprecedented violence in 1994 that wiped out about three-quarters of the ethnic Tutsi population in the African nation of Rwanda has the potential to reveal much about how mass killings unfold at ground level.
There is no guarantee that a better, although inevitably incomplete, understanding of why certain members of Rwanda’s majority Hutu population nearly eliminated a Tutsi minority will prevent future large-scale slaughters. The research is worth the effort, though, especially in a 21st century already marked by massacres of hundreds of thousands of people in western Sudan’s Darfur region and in Syria.

Researchers have an advantage in Rwanda. When hostilities ended, Rwanda’s government gathered extensive data on genocide victims and suspected perpetrators through a national survey. And local courts tried more than 1 million cases of alleged involvement in the violence, making the case documents available to researchers.

Genocide studies have often split offenders into organizers — mainly political and community leaders — and “ordinary men” who kill out of blind obedience to central or local authorities and hatred of those deemed enemies. But the extensive data from Rwanda tell a different story: An individual’s willingness to take part in genocidal violence depends on many personal and social factors that influence whether and how deeply a person participates, says sociologist and Rwanda genocide researcher Hollie Nyseth Brehm of Ohio State University in Columbus.

Nyseth Brehm’s findings may not apply to some of Rwanda’s most avid killers, who eluded capture and fled the country as soon as hostilities stopped. But when it comes to the ordinary citizens swept up in the deadly campaign, involvement was not primarily about following political leaders’ orders to eliminate Tutsis.

New reports by Nyseth Brehm and others fuel skepticism about the popular idea that regular folks tend to do as they’re told by authorities. And a fresh look at a famous 1960s psychology study adds further doubt that people will blindly follow orders to harm or kill others.
In reality, only about 20 percent of Hutu men, an estimated 200,000, seriously injured or killed at least one person during the genocidal outbreak, estimates Rwanda genocide researcher Omar McDoom of the London School of Economics and Political Science.

“Why did four in five Hutu men not engage in the killing?” McDoom asks. That puzzle goes against the ordinary man thesis that “implies there are no individual differences in genocide participation,” he says. He suspects participation hinged on personal motivations, such as wanting to defend Rwanda from enemies or make off with a Tutsi neighbor’s possessions. Social circumstances, such as living in high-violence areas or having friends or family members who had already murdered Tutsis, probably played a role too. Nyseth Brehm agrees.

Local triggers
Genocides often fester before exploding. In Rwanda, Tutsi rebels attacked the Hutu-led government and set off a civil war several years before mass killings started. A turning point came when unidentified forces killed Rwanda’s president, shooting down his plane on April 6, 1994. Over the next three months, the government orchestrated a massacre of Tutsis and any Hutus deemed friendly or helpful to Tutsis. Most scholars place the death toll at around 800,000, although estimates range from 500,000 to 1.2 million. Bands of Hutus scoured the countryside for their sworn enemies. Killings took place at roadblocks and in raids on churches, schools and other community facilities. Hutu women killed on a much smaller scale than men did, although they often aided those involved in the carnage.

In many parts of Rwanda, local authorities appointed by the national government recruited Hutu men into groups that burned and looted homes of their Tutsi neighbors, killing everyone they encountered, says political scientist Scott Straus of the University of Wisconsin–Madison. In his 2016 book Fundamentals of Genocide and Mass Atrocity Prevention, Straus describes how Rwandan recruitment efforts coalesced into a killing machine. Politicians, business people, soldiers and others encouraged Hutu farmers to kill an enemy described as “cockroaches” in need of extermination. Similarly, Nazis portrayed Jews as cockroaches and vermin.

Despite the Rwandan state’s best efforts to encourage nationwide Tutsi annihilation, local conditions shaped how the 1994 genocide unfolded, Nyseth Brehm reported in February in Criminology. She looked at 142 of the nation’s 145 municipalities, known as communes. Some experienced as few as 71 killings, while in others, as many as 54,700 people were murdered, she found.

Communes with the fewest killings were those that had the highest marriage and employment rates, Nyseth Brehm says. In those settings, mainly farming communities where people knew and trusted each other, most citizens valued a peaceful status quo and discouraged a descent into mass killing, she suspects.
Curiously, violence was worse in areas with the largest numbers of educated people. That points to the effectiveness of anti-Tutsi teachings in Rwandan schools, Nyseth Brehm suggests.

Her study relied on data from a postgenocide survey, published in 2004 by Rwanda’s government, intended to document every person killed during the atrocity. Citizens throughout Rwanda told interviewers about individuals in their communities who had been killed during the outburst of slaughter. Reported and confirmed deaths were checked against records of human remains linked to the 1994 genocide. Comparisons were also made to Rwanda’s 1991 census.

However, any data on killings during mass violence, including from the Rwandan survey, will be incomplete, Nyseth Brehm cautions. So she also analyzed data from 1,068,192 genocide-related cases tried in local Rwandan courts from 2002 to 2012. Of particular note, although most nongenocidal murders in Rwanda are carried out by men in their 20s, the average age of accused genocide perpetrators was 34.7 years old, Nyseth Brehm reported in the November 2016 Criminology.

Hutu men in their 30s joined the genocidal fray as a way to fulfill adult duties by defending their communities against an outside threat, she suggests. Preliminary analyses show that perpetrators tended to cluster in families; if one of several brothers killed Tutsis, the others were far more likely to follow suit.

Additional scouring of court data indicated that Rwandans who had siblings convicted of genocide killings were especially likely to have murdered Tutsis themselves. In earlier interviews of 130 Rwandans, some who had killed Tutsis and others who hadn’t, McDoom similarly found that perpetrators tended to cluster in families.

Missing murderers
Unfortunately, the Rwandan genocide’s most prolific players have eluded both the law and science, says political scientist Cyanne Loyle of Indiana University Bloomington. Investigators have so far interviewed only a handful of the powerful “big fish” who orchestrated the genocide, plus several hundred people tried and imprisoned for genocide participation. Survey and court data are limited to killers who either stayed in Rwanda after atrocities ended or were caught trying to flee the country.

But perpetrators with the most blood on their hands traveled in bands, wiping out tens of thousands of people at a time before hiding abroad, Loyle says. For instance, local officials lured large numbers of Tutsis to a school near the town of Murambi, where Hutu militias used machine guns, explosives and other weapons to kill more than 40,000 people in just three days.

“Scholars have studied Rwandans who killed on the sidelines while a larger and deadlier campaign was under way,” Loyle says. “They have mistaken a sideshow for the main event.”

Perpetrators of colossal atrocities at Murambi and elsewhere were less powerful than the government’s genocide masterminds, Loyle says. These “murderers in the middle,” however, were better equipped and far more effective at killing than common folk who got caught up in events, she contends.

There are no good estimates of how many members of large-scale killing squads escaped Rwanda and now live elsewhere. From 15,000 to 22,000 members of the Rwandan army and local militia groups were at large in the Democratic Republic of the Congo, near Rwanda’s border, in January 2003, according to a report by the International Crisis Group, a nonprofit organization.

Nyseth Brehm acknowledges the difficulty of accounting for genocide perpetrators who eluded justice. She and others, including Straus, have interviewed genocide offenders who stayed in Rwanda, often imprisoned for their crimes. Many of those who fled must have traveled in groups that murdered on a grand scale, she says. Those mass killers represent crucial missing data on who participates in genocide, and for what reasons.
Vicious virtue
In interviews by Nyseth Brehm, McDoom and others, perpetrators listed many reasons for joining the 1994 killing spree — hatred of Tutsis, a perceived need to protect nation and family, a desire to claim a neighbor’s property or a decision to join a suddenly popular cause, to name a few. Blind obedience to brutal leaders was far from the only reason cited.

That finding conflicts with the late psychologist Stanley Milgram’s interpretation of his famous “obedience to authority” experiments. Milgram described those trials, in which volunteers were told to administer increasingly intense shocks to another person, as a demonstration of people’s frequent willingness to follow heinous commands. He saw the experiments as approximating the more extreme situations in which Germans had participated in the Holocaust.
On closer inspection, though, Milgram’s study aligns closely with what’s known about Rwandan genocide perpetrators, says S. Alexander Haslam, a psychologist at the University of Queensland in Australia.
In Milgram’s experiments, as in Rwanda and Nazi Germany, “those willing to harm others were not so much passive ciphers as motivated instruments of a collective cause,” Haslam says. “They perceived themselves as acting virtuously and doing good things.”

Although Milgram’s tests upset some volunteers, most participants identified with his scientific mission to understand human behavior and wanted to prove themselves as worthy of the project, Haslam and psychologist Stephen Reicher of the University of St. Andrews in Fife, Scotland, conclude in a research review scheduled to appear in the 2017 Annual Review of Law and Social Science.

Milgram conducted 23 obedience experiments with New Haven, Conn., residents in 1961 and 1962 (SN: 9/21/13, p. 30). Most attention has focused on only one of those experiments. Volunteers designated as “teachers” were asked by an experimenter to continue upping the intensity of what they thought were electric shocks to a “learner” — who was actually in league with Milgram — who erred time and again on a word-recall test. Through screams, shouts and eventually dead silence from the learner, 26 of 40 volunteers, or 65 percent, administered shocks all the way to a maximum of 450 volts.

But experiments that undermined participants’ identification with the scientific mission lowered their willingness to deliver the harshest shocks, Haslam and Reicher say. Fewer volunteers shocked to the bitter end if, for instance, the study was conducted in an office building rather than a university laboratory or if the experimenter was not physically present. An analysis of data available from 21 of the 23 experiments finds that 43.6 percent of 740 volunteers shocked learners to the limit.
Participants were most compliant when an experimenter encouraged them to continue shocking for the sake of the experiment (by saying, “The experiment requires that you continue”), the psychologists add. Participants never followed the order: “You have no choice, you must continue.”

Milgram’s archives at Yale University contain letters and survey responses from former participants reporting high levels of support for Milgram’s project and for science in general. Many former volunteers told Milgram that they administered shocks out of a duty to collaborate on what they viewed as important research, even if it caused them distress at the time. Still, Milgram’s recruits often admitted having had suspicions during the experiments that learners were not really being zapped.

Milgram was right that his experiments applied to real-world genocides, Haslam concludes, but erred in assuming that obedience to authority explained his results. From Milgram’s laboratory to Rwanda’s killing squads and Nazi concentration camps, orders to harm others are carried out by motivated followers, not passive conformists, he asserts.

If anything, that makes genocide all the more horrifying.

When the Aug. 21 solar eclipse unveils the sun’s normally dim atmosphere, the corona will look like an intricate, orderly network of loops, fans and streamers. These features trace the corona’s magnetic field, which guides coronal plasma to take on the shape of tubes and sheets.

These wispy coronal structures arise from the magnetic field on the sun’s visible surface, or its photosphere. Unlike the corona, the photosphere’s magnetism is a complete mess.
“It’s not a static surface like the ground, it’s more like an ocean,” says solar physicist Amir Caspi of the Southwest Research Institute in Boulder, Colo. “And not just an ocean. It’s like a boiling ocean.”

Because the corona’s loops and streamers all originate in the turbulent photosphere, their roots should get twisted and turned around.

“And yet these structures in the corona are not tangled and snarled and matted like kelp or seaweed in the ocean,” Caspi says. “They seem to still be these organized, smooth loops. Nobody understands why.”

To unknot the photosphere’s tangled mats, the corona must release some of the energy stored there, Caspi says. So during the eclipse, he and his colleagues will be looking for the release valves that set the corona free.
One possibility is that wave motion in the corona’s magnetic field lines helps untie the snarls. Magnetic waves in plasma, called Alfvén waves, are thought to ripple through the sun’s magnetic field lines like vibrations in a guitar string. Researchers have directly observed Alfvén waves in the lower corona, within about half a solar radius of the surface (SN: 4/11/09, p. 12), but not farther out where similar waves with higher amplitudes would travel. Those close-in waves were too weak to explain the corona’s features, but perhaps more distant waves could shake things up enough.
Another option is that little hypothetical spurts of magnetic energy could help release the tangles. These nanoflares and nanojets would be like solar flares but with a billionth of the energy. By going off all the time, nanoflares and nanojets could collectively release enough energy to give the corona some structure, simulations have shown.

“Both are symptoms of tiny rearrangements of the magnetic field — magnetic reconnection,” says solar physicist Craig DeForest, also at the Southwest Research Institute. Solar flares and bigger outbursts called coronal mass ejections are also signs of magnetic reconnection, but they’re not frequent enough to account for the corona’s smoothness. “Nanojets and/or nanoflares in the middle corona would be a smoking gun that would explain why the corona is so organized,” DeForest says.

No one has actually seen any nanoflares or nanojets. Theories suggest that they’re too small and quick to see individually — but they should be visible as a cacophony of little pops when the solar eclipse reveals the lower corona.

The shaking from Alfvén waves and the flickers of nanoflares could not only loosen up the tangled skein of magnetism, but also transfer heat high up into the corona. Caspi, DeForest and their colleagues hope to see both effects on August 21, when they fly a pair of telescopes on twin NASA WB-57 high-altitude research jets along the path of the eclipse (SN Online: 8/14/17).

“We’re taking high-speed movies of the sun and analyzing them for things that look like waves,” Caspi says. “We’re just overall looking at the structure of the corona.”

A boy who lived in what’s now South Africa nearly 2,000 years ago has lent a helping genome to science. Using the long-gone youngster’s genetic instruction book, scientists have estimated that humans emerged as a distinct population earlier than typically thought, between 350,000 and 260,000 years ago.

The trick was retrieving a complete version of the ancient boy’s DNA from his skeleton to compare with DNA from people today and from Stone Age Neandertals and Denisovans. Previously documented migrations of West African farmers to East Africa around 2,000 years ago, and then to southern Africa around 1,500 years ago, reshaped Africans’ genetics — and obscured ancient ancestry patterns — more than has been known, the researchers report online September 28 in Science.
The ancient boy’s DNA was not affected by those migrations. As a result, it provides the best benchmark so far for gauging when Homo sapiens originated in Africa, evolutionary geneticist Carina Schlebusch of Uppsala University in Sweden and her colleagues conclude.

In line with the new genetically derived age estimate for human origins, another team has proposed that approximately 300,000-year-old fossils found in northwestern Africa belonged to H. sapiens (SN: 7/8/17, p. 6). Some researchers suspect a skull from South Africa’s Florisbad site, dated to around 260,000 years ago, qualifies as H. sapiens. But investigators often place our species’ origins close to 200,000 years ago (SN: 2/26/05, p. 141). There is broad consensus that several fossils from that time represent H. sapiens.

Debate over the timing of human origins will continue despite the new evidence from the child, whose remains came from previous shoreline excavations near the town of Ballito Bay, says Uppsala University evolutionary geneticist and study coauthor Mattias Jakobsson. “We don’t know if early Homo sapiens fossils or the Florisbad individual were genetically related to the Ballito Bay boy,” he says.

Thus, the precise timing of humankind’s emergence, and exact patterns of divergence among later human populations, remain unclear. Researchers have yet to retrieve DNA from fossils dating between 200,000 and 300,000 years old that either securely or possibly belong to H. sapiens.
However early human evolution played out, later mixing and mingling of populations had a big genetic impact. DNA evidence from more recent fossils, including those studied by Schlebusch’s group, increasingly suggests that Stone Age human groups migrated from one part of Africa to another and mated with each other along the way (SN: 10/20/12, p. 9), says Harvard Medical School evolutionary geneticist Pontus Skoglund. In the Sept. 21 Cell, he and his colleagues report that DNA from 16 Africans, whose remains date to between 8,100 and 400 years ago, reveals a shared ancestry among hunter-gatherers from East Africa to South Africa that existed before West African farmers first arrived 2,000 years ago.

That ancient set of common genes still comprises a big, varying chunk of the DNA of present-day Khoisan people in southern Africa, Skoglund’s group found. Earlier studies found that the Khoisan — consisting of related San hunter-gatherer and Khoikhoi herding groups — display more genetic diversity than any other human population.

Schlebusch’s team estimates that a genetic split between the Khoisan and other Africans occurred roughly 260,000 years ago, shortly after humankind’s origins and around the time of the Florisbad individual. Khoisan people then diverged into two genetically distinct populations around 200,000 years ago, the researchers calculate.

Ancient DNA in Schlebusch’s study came from seven individuals unearthed at six South African sites. Three hunter-gatherers, including the Ballito Bay boy, lived about 2,000 years ago. Four farmers lived between 500 and 300 years ago.

Comparisons to DNA from modern populations in Africa and elsewhere indicated that between 9 percent and 30 percent of Khoisan DNA today comes from an East African population that had already interbred with Eurasian people. Those East Africans were likely the much-traveled farmers who started out in West Africa and reached southern Africa around 1,500 years ago, the researchers propose.

For Chong Liu, asking a scientific question is something like placing a bet: You throw all your energy into tackling a big and challenging problem with no guarantee of a reward. As a student, he bet that he could create a contraption that photosynthesizes like a leaf on a tree — but better. For the now 30-year-old chemist, the gamble is paying off.

“He opened up a new field,” says Peidong Yang, a chemist at the University of California, Berkeley who was Liu’s Ph.D. adviser. Liu was among the first to combine bacteria with metals or other inorganic materials to replicate the energy-generating chemical reactions of photosynthesis, Yang says. Liu’s approach to artificial photosynthesis may one day be especially useful in places without extensive energy infrastructure.

Liu first became interested in chemistry during high school, and majored in the subject at Fudan University in Shanghai. He recalls feeling frustrated in school when he would ask questions and be told that the answer was beyond the scope of what he needed to know. Research was a chance to seek out answers on his own. And the problem of artificial photosynthesis seemed like something substantial to throw himself into — challenging enough “so [I] wouldn’t be jobless in 10 or 15 years,” he jokes.
Photosynthesis is a simple but powerful process: Sunlight helps transform carbon dioxide and water into chemical energy stored in the chemical bonds of sugar molecules. But in nature, the process isn’t particularly efficient, converting just 1 percent of solar energy into chemical energy. Liu thought he could do better with a hybrid system.
The efficiency of natural photosynthesis is limited by light-absorbing pigments in plants or bacteria, he says. People have designed materials that absorb light far more efficiently. But when it comes to transforming that light energy into fuel, bacteria shine.

“By taking a hybrid approach, you leverage what each side is better at,” says Dick Co, managing director of the Solar Fuels Institute at Northwestern University in Evanston, Ill.

Liu’s early inspiration was an Apollo-era attempt at a life-support system for manned space missions. The idea was to use inorganic materials with specialized bacteria to turn astronauts’ exhaled carbon dioxide into food. But early attempts never went anywhere.

“The efficiency was terribly low, way worse than you’d expect from plants,” Liu says. And the bacteria kept dying — probably because other parts of the system were producing molecules that were toxic to the bacteria.

As a graduate student, Liu decided to use his understanding of inorganic chemistry to build a system that would work alongside the bacteria, not against them. He first designed a system that uses nanowires coated with bacteria. The nanowires collect sunlight, much like the light-absorbing layer on a solar panel, and the bacteria use the energy from that sunlight to carry out chemical reactions that turn carbon dioxide into a liquid fuel such as isopropanol.

As a postdoctoral fellow in the lab of Harvard University chemist Daniel Nocera, Liu collaborated on a different approach. Nocera had been working on a “bionic leaf” in which solar panels provide the energy to split water into hydrogen and oxygen gases. Then, Ralstonia eutropha bacteria consume the hydrogen gas and pull in carbon dioxide from the air. The microbes are genetically engineered to transform the ingredients into isopropanol or another liquid fuel. But the project faced many of the same problems as other bacteria-based artificial photosynthesis attempts: low efficiency and lots of dead bacteria.
“Chong figured out how to make the system extremely efficient,” Nocera says. “He invented biocompatible catalysts” that jump-start the chemical reactions inside the system without killing off the fuel-generating bacteria. That advance required sifting through countless scientific papers for clues to how different materials might interact with the bacteria, and then testing many different options in the lab. In the end, Liu replaced the original system’s problem catalysts — which made a microbe-killing, highly reactive type of oxygen molecule — with cobalt-phosphorus, which didn’t bother the bacteria.

Chong is “very skilled and open-minded,” Nocera says. “His ability to integrate different fields was a big asset.”

The team published the results in Science in 2016, reporting that the device was about 10 times as efficient as plants at removing carbon dioxide from the air. With 1 kilowatt-hour of energy powering the system, Liu calculated, it could recycle all the carbon dioxide in more than 85,000 liters of air into other molecules that could be turned into fuel. Using different bacteria but the same overall setup, the researchers later turned nitrogen gas into ammonia for fertilizer, which could offer a more sustainable approach to the energy-guzzling method used for fertilizer production today.

Soil bacteria carry out similar reactions, turning atmospheric nitrogen into forms that are usable by plants. Now at UCLA, Liu is launching his own lab to study the way the inorganic components of soil influence bacteria’s ability to run these and other important chemical reactions. He wants to understand the relationship between soil and microbes — not as crazy a leap as it seems, he says. The stuff you might dig out of your garden is, like his approach to artificial photosynthesis, “inorganic materials plus biological stuff,” he says. “It’s a mixture.”

Liu is ready to place a new bet — this time on re-creating the reactions in soil the same way he’s mimicked the reactions in a leaf.

MINNEAPOLIS — Birds don’t need to be drenched in crude oil to be harmed by spills and leaks.

Ingesting even small amounts of oil can interfere with the animals’ normal behavior, researchers reported November 15 at the annual meeting of the Society of Environmental Toxicology and Chemistry North America. Birds can take in these smaller doses by preening slightly greasy feathers or eating contaminated food, for example.

Big oil spills, such as the 2010 Deepwater Horizon disaster, leave a trail of dead and visibly oily birds (SN: 4/18/15, p. 22). But incidents like last week’s 5,000-barrel spill from the Keystone pipeline — and smaller spills that don’t make national headlines — can also impact wildlife, even if they don’t spur dramatic photos.
To test how oil snacks might affect birds, researchers fed zebra finches small amounts of crude oil or peanut oil for two weeks, then analyzed the birds’ blood and behavior. Birds fed the crude oil were less active and spent less time preening their feathers than birds fed peanut oil, said study coauthor Christopher Goodchild, an ecotoxicologist at Oklahoma State University in Stillwater.

Oil-soaked birds will often preen excessively to try to remove the oil, sometimes at the expense of other important activities such as feeding. But in this case, the birds didn’t have any crude oil on their feathers, so the decrease in preening is probably a sign they’re not feeling well, the researchers say.

Exactly how the oil affects the birds’ activity levels isn’t clear. Researchers suspected that oil might deprive birds of oxygen by affecting hemoglobin, which carries oxygen in the blood. Blood tests didn’t turn up any evidence of damaged hemoglobin proteins but did find some evidence that oil-sipping birds might be anemic, Goodchild said. At the higher of two crude oil doses, birds’ blood contained less hemoglobin per red blood cell, a sign of anemia.
The findings, while preliminary, add to a growing pile of evidence that estimates of the number of animals impacted by oil spills might be too low. For instance, even a light sheen of oil on sandpipers’ wings makes it harder to fly, costing birds more energy, a different group of researchers reported earlier this year. That could affect everything from birds’ daily movements to long-distance migration.

Movies and other media are full of mixed messages about the risks and rewards of building machines with minds of their own. For every manipulative automaton like Ex Machina’s Ava (SN: 5/16/15, p. 26), there’s a helpful Star Wars droid. And while some tech titans such as Elon Musk warn of the threats artificial intelligence presents, others, including Mark Zuckerberg, dismiss the doomsayers.

AI researcher Toby Walsh’s Machines That Think is for anyone who has heard the hype and is seeking a critical assessment of what the technology can do — and what it might do in the future. Walsh’s conversational style is welcoming to nonexperts while his endnotes point readers to opportunities for deeper dives into specific aspects of AI.
Walsh begins with a history of AI, from Aristotle’s foundation of formal logic to modern facial-recognition systems. Excerpts from computer-composed poetry and tales of computers trouncing humans at strategy games (SN: 11/11/17, p. 13) are a testament to how far AI has come. But Walsh also highlights weaknesses, such as machine-learning algorithms’ reliance on so much data to master a single task.

This 30,000-foot view of AI research packs a lot of history, as well as philosophical and technical explanation. Walsh personalizes the account with stories of his own programming experiences, anecdotes about AI in daily life — like his daughter’s use of Siri — and his absolute, unapologetic love of puns.

Later in the book, Walsh speculates about technical hurdles that may curb further AI development and legal limits that society may want to impose. He also explores the societal impact that increasingly intelligent computers may have.
For instance, Walsh evaluates how likely various jobs are to be outsourced to AI. Some occupations, like journalist, will almost certainly be automated, he argues. Others, like oral surgeon, are probably safe. For future job security, Walsh recommends pursuing careers that require programming acumen, emotional intelligence or creativity.

AI also has the potential to revolutionize warfare. “Like Moore’s law, we are likely to see exponential growth in the capabilities of autonomous weapons,” Walsh writes. “I have named this ‘Schwarzenegger’s law’ to remind us of where it will end.” Walsh isn’t resigned to a Terminator-like future, though. If governments ban killer robots and arms developers use automation to enhance defensive equipment, he believes military AI could actually save many lives.

In fact, Walsh argues, all aspects of AI’s future impacts are in our hands. “Artificial intelligence can lead us down many different paths, some good and some bad,” he writes. “Society must choose which path to take.”