人工智能会抢走人类的所有饭碗，将人类变成无用阶层么？Will AI Turn Us All Into the Useless Class?

"People will always be the core factor—not robots."

Piece No. 5

A lot of readers had their whole worldview "upended" by the argument in Homo Deus: A Brief History of Tomorrow. According to Harari, AI systems will, in the not-too-distant future, take away all of humanity's jobs, and in doing so turn all of humankind into a "useless class."

Fully "convinced" by Harari, some people started to doubt the point of working hard and studying hard, and resolved to exile themselves early into the ranks of the "useless class," settling into a life of eating their fill and doing nothing with their minds.

But will AI really take away all of humanity's jobs and turn us into a useless class?

The whole purpose that keeps the intelligent-robot economy in motion lies in humans, not in the robots themselves

Here's the reasoning of the people who believe all of this will come to pass: robot couriers are already in service in parts of Hokkaido, Japan, and back home, JD.com will soon roll them out based on big data and its use cases. In the near future, human couriers will all be out of work. In industrial production, there are more and more robots on the assembly line, so blue-collar workers won't be needed. In warfare, militaries lean more and more on drones and other unmanned combat platforms, so human soldiers won't be needed. On top of that, as self-driving technology spreads, human drivers won't be needed; as super-powered automated medical-diagnosis software arrives, human doctors won't be needed; and so on and so forth.

Putting all this together: the professional training and education programs across every industry won't need to exist anymore, because those institutional arrangements were designed for human characteristics in the first place—and once human labor no longer needs to exist in large numbers, those very institutions ought to be laid to rest. So in the end, everyone will be out of a job.

This scenario looks logically airtight, and pretty appealing too.

But we can flip the question around. In the society imagined above, all the cooks are robots. So does the robot cook make its braised-chicken-over-rice and its Shaxian snacks for itself—or for humans?

The answer is beyond doubt: for humans. Which means this already concedes that in that imagined future society, the products turned out by the great labor army of intelligent robots must, in the end, face a consumer market made up of humans. And there's no such thing as a free lunch. After human consumers eat the food made by the robot cook, they reasonably still have to pay for it—which in turn presupposes that they have the ability to pay, and therefore that they have an income.

But by the logic above, humans as a "useless class" are useless, so they shouldn't have any income. And if that's the case, how would this "useless class" have the income to buy the things they need? If no one can afford these products, then where does the "final cause" of the robots' production even show up? Produce a heap of food and then dump it? And the material costs that the robots' own production activity generates—raw materials, electricity, and the like—who pays for those?

Clearly this leads to an infinite regress in the logic. And the only way to prevent that regress is to concede this: the entire "final cause" that keeps the "intelligent-robot economy" in motion lies in humans, not in the robots themselves. That conclusion obviously topples the conclusion that "humans will become a useless class" on the spot.

A broad participation of human intelligence is one of the crucial logical preconditions for a robot-led economic structure to exist in a healthy way

Some people might not quite agree with this reasoning. They think there's another possibility: the future society is a super-welfare society, where we can all get large amounts of Bitcoin for free to buy food and life's necessities. A highly developed robot economy can guarantee an abundant supply of goods, and that is bound to make satisfying humanity's various material needs a breeze.

It still sounds lovely. But how does a super-welfare society come about? The mass arrival of intelligent robots doesn't necessarily lead automatically to the arrival of an ideal society.

Because they haven't ruled out one possibility: they've underestimated human nature.

Those who control and monopolize intelligent-robot technology will make that technology serve their own class's interests first, and press the welfare of the great majority of people in society down to the lowest possible level. If that happens, the effective consumer population of the whole society shrinks sharply, leading to overproduction and, with it, an anemic, faltering economic cycle.

You could try to counter this by strengthening the oversight system that the commoners—who make up the vast majority of humanity—exercise over the machine-production-and-distribution mechanism. But at this point a hole opens up in their argument: we just conceded that one of the crucial logical preconditions for an intelligent-robot-led economic structure to exist in a healthy way is human oversight of the relevant production-and-distribution mechanism. And that oversight is itself an expression of human intelligence. Put another way, a broad participation of human intelligence is one of the crucial logical preconditions for an intelligent-robot-led economic structure to exist in a healthy way. So we can immediately conclude: even within such a structure, humans are certainly not a "useless class."

Pulling it all together, then, we arrive at this: the relationship between today's AI and human intelligence is like the relationship between hair and hide—if the hide is gone, what will the hair cling to?

The existence of an AI economy depends, at the level of the "final cause," on the existence of humanity's ability and habit of consuming; otherwise this economy cannot complete a healthy cycle and thereby persist in a stable way. And preserving humanity's ability and habit of consuming depends, in turn, on the existence of human intellectual labor; otherwise the "conservation of energy" in how the social economy operates cannot be guaranteed. So human intellectual labor—not merely in the narrow sense of designing robots—is, from beginning to end, a logical precondition for the existence of an AI economy.

The arrival of self-aware robots would, on the contrary, prove humans won't become a "useless class"

Someone might raise another question: why must intelligent robots take satisfying humanity's physical, psychological, and cultural needs as their "purpose for working"? Couldn't a self-aware super-robot "fully be its own thing"? In that case, why must we assume that the workings of the robot economy take the universal existence of human intellectual labor as a logical precondition? For instance, robots themselves don't need to eat; if they became self-aware of this, they could go pursue their own life goals—so wouldn't their activity then be decoupled from human material needs?

But this scenario is actually a false counterexample. Suppose robots with this kind of self-awareness do appear. Then, on the contrary, it proves humans won't become a "useless class"—because humans themselves still have to eat. If the robots go on strike, humanity's own cooks, long famished, are more than ready and can't wait to step in.

So now we've arrived at a prisoner's-dilemma line of reasoning:

Option A: If future robots aren't yet smart enough to realize their needs differ from humans', they'll obediently serve as humanity's slaves, producing all of life's necessities according to shifts in human taste. In this case, the human way of living remains the precondition for the intelligent economy's existence, and humans won't become a "useless class."

Option B: If future robots are already smart enough to "realize they have independent needs of their own," then humans have two options: keep using those less-smart "robot slaves" to serve them, or go do the work themselves. But whichever they choose, humans won't become a "useless class."

That said, technological innovation and social change really will put a lot of people out of work—that's an irreversible historical process. And the new jobs created by new technology won't be enough to bring the number of unemployed back down. This too is the social upheaval that any technological upgrade brings, and it isn't uniquely tied to AI. On top of that, and more importantly, since a fully realized artificial intelligence doesn't appear in this scenario, no wholesale replacement of the full range of human intelligence will occur under these conditions either.

A large rise in the unemployed population is enough to shrink the effective consumer population, leading to the sustainability problem of the economic cycle described earlier. This is a big problem humanity has to face, and one humanity will need its own wisdom to solve. And that, precisely, demonstrates how important human wisdom is.

So, to sum all of this up: the day humans become a useless class will never come.