What AI has in common with dogs

Two new books ponder whether the technology will be tame or savage

The Rise and Fall of the Artificial State. By Jill Lepore. Liveright; 336 pages; $29.99. Allen Lane; £25

What Happened to Liberal Democracy? By Daron Acemoglu. Dutton; 416 pages; $32. Profile Books; £25

When mulling how AI will transform society, consider an earlier technology that gave humans a helpful companion with superior capabilities: the dog. By domesticating wolves, humans were able, indirectly, to smell game far away and hear enemies approaching at night. They became better hunters than Neanderthals and better warriors: hard to ambush, hard to elude. Dogs are one reason why Homo sapiens outcompeted their prehistoric cousins and conquered the world, some anthropologists believe.

Back when dog technology was first being developed, it was far from obvious that it was safe for people—who were literally inviting wolves into their caves. They had to train them not to eat babies. Thousands of years later, dogs still sometimes escape and go rogue. But overall, people are happy with how things have worked out. (It did not end so well for the Neanderthals.)

The optimistic case for AI is that it will be like dogs: complementing rather than replacing human labour, and remaining obedient to its masters. But what if it isn’t? AI evolves much faster. Soon it could be creating new, improved versions of itself at a pace dog-breeders can only dream of. So whereas people can be confident that dogs will not enslave them—by 2030, pooches will still not know how to use a tin-opener—they cannot be so sure about AI.

Two new books, one by a Nobel-prizewinning economist and one by a Pulitzer-garlanded historian, grapple with the question of how AI will shape the future. Neither uses the dog analogy, but both are easier to understand with reference to it.

Jill Lepore, a Harvard historian, focuses on the men who train the dogs. In her telling, AI tycoons are horrible, irresponsible people whose creations may break their leashes and savage humans. In “The Rise and Fall of the Artificial State”, she argues that tech bosses are trying to build “a future in which humans are ruled, and many replaced, by machines owned by corporations”. This, she says, implies “both the desertion of liberal democracy and the destruction of the natural world”.

She bases her case partly on the moguls’ own words. Sam Altman, the boss of OpenAI, has said it would be “awesome” to replace human presidents with an AI that could “go around and talk to every person on Earth, understand their exact preferences” and “optimise” for all of them. Elon Musk has said humans may end up like “pet labradors” for superintelligent robots.

Mr Altman has suggested that most of Earth’s surface may soon be covered with data centres. Mr Musk wants to solve this problem by putting data centres in space—and blasting millions of people to live on Mars. The tech titans all seem to have escape plans, notes Ms Lepore: “Elon Musk to Mars, Mark Zuckerberg to the metaverse...Jeff Bezos to space, Sam Altman’s brain uploaded to the cloud.” This makes them reckless about the risk of creating dystopia here on Earth, she fears.

She is right that today’s tech bosses are not people you would trust to write the rules for the future of civilisation. But she overestimates the importance of individual billionaires. How AI shapes society will not depend on Mr Musk’s eccentric political views, any more than the effects of cars depended on Henry Ford’s. Rather, it will hinge on the power of the technology itself and the uses to which people put it.

That is a question Daron Acemoglu addresses in “What Happened to Liberal Democracy?”. Mr Acemoglu is among the world’s most-cited economists. He is best known for “Why Nations Fail”, a sweeping economic history arguing that the main factor keeping poor countries poor is predatory institutions. His new work focuses on the present day and the interplay between technology and democracy.

Liberal democracy has underpinned progress, he argues. It has fostered innovation and economic growth, spread the benefits (through taxes and public services) and protected voters with regulations.

For a long time, the system more or less met voters’ expectations. But “then computers happened.” Digital technologies, in Mr Acemoglu’s account, aggravated inequality by raising the demand for brains over brawn. They also boosted the “social power of college graduates”, allowing a new liberal elite to impose its values on a resentful working class, provoking a backlash that led to the election of Donald Trump. And they changed how people interact. In America time spent engaging with friends in person has plunged by two-thirds in the past 20 years. Political debate has grown more extreme and toxic.

AI could make everyone much richer by supercharging innovation. But it could also supercharge the harms of digital technology, Mr Acemoglu (rather familiarly) fears. It could lead to widespread joblessness, creating “huge social-status gaps” between “makers” and “takers” and a full-blown crisis of democracy. (It could also accelerate social atomisation by making loverbots and political fakery more seductive.)

His proposal is “pro-worker AI”. By this, Mr Acemoglu means finding ways for workers to use AI to make their labour more valuable, like prehistoric hunters combining their spear-making and throwing skills with a dog’s nose and speed. Maybe AI could enable electricians to troubleshoot increasingly sophisticated grids, spotting problems via sensors and helping humans analyse them with a database of past glitches and solutions. Perhaps AI could improve education: an AI assistant could assess pupils’ answers to quizzes and suggest ways to split the class into smaller groups, who could be offered tailored instruction, thus increasing the demand for human teachers.

It is hard to quarrel with this. AI will certainly cause less social turmoil if it complements human labour rather than making it redundant. But Mr Acemoglu leaves unresolved the question of how this is to be accomplished. Ultimately, both books are disappointing. Inadvertently, they illuminate how hard it is to write persuasively about a technology that is changing so fast, and whose future capabilities are so utterly unknown.

Politically, AI has become unpopular. Twice as many Americans are pessimistic about its long-term effects on society as are optimistic, and only 2% think it is developing “too slowly”. But one way or another, this new wolf will be invited into the cave. The countries pioneering AI technology, America and China, are rivals. Neither wants to be the second to develop a superintelligent ally, for fear of ending up like those hapless, dogless Neanderthals. ■