There are two main claims behind this question worth responding to...
Claim 1: Certain capabilities are intrinsically beyond machines
Capabilities such as creativity, empathy, judgment, craft, a genuine human touch are typically considered to be beyond machines, but the list of things we are confident only a human can do keeps getting shorter, and it does so faster than almost anyone expects. We tend to underestimate both how capable machines will become and how often people will actually prefer a machine once one is available. Some of the clearest early examples come from exactly the domains we assume are safest. A large share of patients already report preferring AI for sensitive health and mental-health conversations, and in head-to-head comparisons people often rate AI responses as more empathetic than those from human professionals. As capability grows, this will reach even occupations we think of as quintessentially human, including performers and caregivers. The category of "only a human can do this" is real today, but shrinking.
Claim 2: The surviving human-only roles would be enough to sustain the labor market.
Suppose we grant that a handful of occupations remain reserved for humans because being human is the whole point of the job, such as certain political, diplomatic, ceremonial, coaching, or companionship roles. Fine. We do not need four billion of them. A global economy currently employs billions of people across the full range of tasks, and a residue of roles where humanity itself is the qualification cannot absorb that workforce. So even in the most generous version of this objection, where machines never quite close the last gap, the labor market as we know it still comes apart.
The disruption of labor operates on tasks, not jobs. Almost every uniquely human job is in fact a bundle of many tasks, most of which are not uniquely human at all, and which can be peeled off and automated one by one. And because AI can move in and out of robotic bodies, any single robot can become any kind of expert on demand, from a farmhand in the morning to a clinician in the evening, which means scarce expertise stops being scarce. Our analysis indicates that by the 2040s there will be virtually nothing a person can do that a machine cannot do as well or better for a tiny fraction of the cost, and this holds even if no machine is ever given fully sapient general intelligence.
Witness the transformation
In the span of just fifteen years, the working horse went from providing the vast majority of road travel to a tiny fraction of it. The automobile had arrived, and the fate of the horse was sealed. We are now on the cusp of a disruption every bit as swift and complete, except this time, we humans are the horses.
A convergence of sensors, computing, actuators, and batteries now gives humanoid robots the capability to perform both cognitive and physical work. AI is already taking on cognitive tasks once reserved for people, and humanoid robots are bringing the same capability to physical tasks. For the first time, the supply of available labor can expand as fast as machines can be built and trained. These systems are already approaching cost parity with human labor across much of the global economy, and their cost will keep falling while their capability keeps rising.
This is about far more than cheaper labor. Robots will create an entirely new and vastly larger labor system in which the marginal cost of labor approaches zero. The result will be a sweeping tide of falling costs, rising quality, and explosive productivity that forms the foundation of an era of superabundance. The nations, industries, and individuals who recognize this early, and who choose to protect people rather than jobs, will be best positioned to navigate the transformation and capture its extraordinary benefits.
Learn more about the disruption of labor and its implications for jobs, society, and the economy.