AI is advancing faster than robotics because its feedback loop is much shorter. An AI model can train and be evaluated millions of times per second using automated systems—like another AI or hard-coded evaluators—in a technique called Generative Adversarial Networks (GANs). This happens in highly parallelized computing environments, allowing rapid improvement.

Robots, by contrast, face a slow, physical feedback loop. Training a chef robot to fry an egg, for example, means actually frying an egg—a process that takes minutes. Some evaluation can be automated using AI vision to spot undercooked or burnt results, but final assessment often requires a human, adding delays.

However that gap is shrinking. Robots are increasingly trained in fast, physics-based virtual simulations. These “rehearsals” allow embedded AIs to iterate quickly, like the “paint the fence” drills in The Karate Kid. Once an AI performs well in simulation, it can transfer its skills to the real world—already several steps ahead.

A major trend that often escapes public attention is the rise of remotely operated—or teleoperated—robots.

This shift has been unfolding quietly within industry, largely outside the view of everyday life. A prime example is Rio Tinto’s mining operations, where robots are controlled remotely to handle heavy machinery, transport materials, and navigate hazardous environments:

But robots are not just limited to industrial settings. They are beginning to enter everyday life as well. One of the most striking examples is the Optimus robot, which has been remotely operated to work as a bartender—serving drinks, interacting with customers, and performing tasks typically handled by human staff. Note: At first, these robots were marketed as AI-driven. However, it was later clarified by Optimus that they were largely controlled through remote operation:

This development will initially benefit workers. For example, miners will be able to operate robots from a safe, comfortable location—far from the dangerous conditions of mining sites, which include cave-ins, inhalation of hazardous particles, and extreme heat.

Even in less hazardous jobs, remote operation will be a net positive. A high-end bartender in New York City, for instance, would no longer need to live in an expensive apartment or endure long commutes. Instead, they could live in a spacious home near family in a much more affordable region and work from there—using a VR headset, haptic gloves, and other remote control tools to operate a humanoid robot behind the bar.

In the early stages, these robots will be controlled by the same professionals who once performed those jobs in person. Their experience will make them well-suited to guide the robots as the technology matures.

In the second phase, companies will begin to ask a simple question: if robots can be teleoperated from anywhere, why continue paying high wages to local workers?

This realization will trigger a new wave of offshoring. Roles such as cooks, cleaners, construction workers, elderly caregivers, and factory operators will increasingly be filled by robots operated remotely from lower-wage regions—most likely in the developing world.

Alongside this, companies will notice another opportunity. Since every teleoperated interaction can be fully recorded—using cameras, motion sensors, and other data streams—they can capture both the environment and the human response in precise detail. These massive datasets will then be used to train AI systems in virtual simulations, allowing them to gradually learn and mimic human performance.

In the third phase, AI will begin taking over more and more tasks from human teleoperators. These systems will have learned from observing many human interactions and can begin automating the most repetitive and predictable parts of the job. This marks the real convergence of robotics and artificial intelligence—what we can call the era of R&AI.

Given the widespread presence of the Paretto Principle in the world, it is reasonable to expect that R&AI systems will eventually be able to perform around 80% of the tasks in most jobs. However, the remaining 20%—the edge cases—will still require human intervention. These are unusual or complex scenarios that machines cannot yet handle. While full automation of even these outliers is theoretically possible, it will take significantly more time and effort.

The result is the emergence of a new type of job: the "remoter". This is a human operator who supervises and manages multiple R&AIs, potentially in a variety of robot body types. Much like a senior employee oversees junior staff, a remoter watches over fleets of robots, intervening only when necessary.

Take the restaurant industry as an example. A single human chef may monitor a network of robot cooks across several kitchens. The remoter can pause any robot, take over control, or override decisions. If an uncontrolled flame is detected, the human can take over and operate the fire extinguisher. If water stops running, the remoter can investigate the building’s water system or call the utility company.

R&AI systems will also be programmed to ask for help when needed. If a robot encounters a situation it was not trained for—like opening multiple rotten eggs—it will pause and request human support. The remoter will receive a queue of such intervention requests. When it is that robot’s turn, the human might take control, inspect the remaining eggs, dispose of the spoiled ones, check the refrigeration system, and then return the robot to normal automated operation.

Some believe the future will be marked by massive unemployment, where most people are forced into underemployment—accepting gig work, temporary or part-time jobs, or steep salary cuts. This may indeed happen at first, as companies use R&AIs to automate existing roles.

However, it is important to remember what actually creates jobs: competition and innovation.

Competition pushes standards higher. When companies automate part of their operations, they free up resources to innovate and differentiate themselves. This leads to new roles being created. This principle applies both to individuals competing for status and businesses competing for customers. For example:

• Washing Machines: The introduction of washing machines reduced the time required to do laundry. Previously, laundry might take four hours per load; today, it takes just over 40 minutes. But total laundry time has not dropped significantly—people wash clothes more frequently, and own more larger wardrobes. As people compete socially, cleanliness and varied clothing become status markers, creating more work, not less.
• ATMs: When ATMs were introduced, many feared bank teller jobs would vanish. Instead, competition between banks forced them to introduce more complex and personalized services to stand out, requiring human representatives to explain and manage these new offerings. Even the ATM itself also became a job generator—new roles emerged for updating security and adding features.

Innovation also drives job creation through what I described in my book as the Four Horsemen of Creative Destruction. Let us consider them through the example of websites:

• Breakthroughs: These are new possibilities that suddenly open up, pushing companies to act adopt them to outpace their competittors. When the internet became widespread, every business needed a website to compete.
• Complexity: These are competing technologies and standards that emerge, forcing companies must stay compatible with many of them. For example, as social media platforms rose, businesses had to maintain a presence across multiple channels.
• Obsolescence: These are newer technologies that render old ones outdated, forcing companies to upgrade. The rise of smartphones made it necessary for websites to be mobile-friendly.
• Trends: These include changing design styles, language, and customer expectations. To avoid looking outdated, companies must frequently refresh their websites and brand presence.

These forces are especially visible in fast-moving industries like software, where patents are rare and innovation thrives. But they also apply, in lesser intensity, across other sectors—like automotive, healthcare, real estate, and consumer goods.

In summary: widespread unemployment or underemployment is not inevitable—so long as we encourage competition and innovation. Abolishing the patent system would remove the biggest structural barrier preventing new ideas from arising and SMEs from creating jobs. Until such structural reforms are made, we will experience worsening job conditions.

Some people, argue that Universal Basic Income (UBI) may be necessary to cope with widespread job loss due to R&AI.

However, UBI is unlikely to work if it is rolled out to large portions of the population. It would likely fuel inflation.

The United Kingdom’s "Help to Buy" program aimed to support first-time homebuyers, but critics argued it inflated housing prices. The IMF and the Bank of England warned that the scheme risked triggering a housing bubble by boosting demand without addressing supply shortages.

Another example was the massive amount of money governments printed to manage the COVID-19 crisis, both in the U.S. and other countries. This resulted in significant inflation in the following years—even after supply adjusted—proving that supply limitations worsen inflation, but are not a requirement for it to occur.

An important consequence is that R&AI will probably cause an economic crash, if measured by GDP or tax revenue. This may seem unintuitive, but it is the natural result of a sharp drop in the price of most goods and services, as they become cheaper to produce through R&AI.

In reality, the economy will be far more productive, but because we still rely on primitive measures like GDP to track economic health, it will appear as if the economy is getting worse, even when it is not.

Without the creation of new products and services that are hard to automate—and therefore remain expensive—governments will struggle to collect tax revenue. They will be forced to print money to fund UBI or other support programs, which leads to more inflation.

Targeted support programs like UBI can be useful—but only when given to a small percentage of the population and when supply can scale with demand. Otherwise, inflation will erase the benefits.

Conclusion: There is no shortcut around the future of work. When R&AIs destroy jobs, we must unleash competition and innovation to create new ones. It is the only sustainable path to prosperity in the age of R&AI.

Some argue that R&AI will create a post-work society, where all our needs are met by machines and we no longer need to work.

This scenario might become reality if R&AIs reach sentience and can perform any human task. However, as mentioned earlier, we are not assuming this will happen.

Another version of this argument claims that even without sentience, R&AIs will be smart enough to produce most goods and services we require, such as food and basic health care. With our basic needs covered, work would no longer be necessary.

However, this belief overlooks two psychological patterns:

• Arrival Fallacy: The illusion that reaching a goal will bring lasting happiness or fulfillment.
• Hedonic Adaptation: The human tendency to return to a baseline level of happiness shortly after positive or negative changes.

If a post-work society were truly satisfying, we would already see widespread adoption. Many people today live with more comfort than their great-grandparents could have imagined. Basic food and health care are affordable in many places, requiring little work to maintain. And yet, most people do not choose a minimalist lifestyle.

There are exceptions—some choose to live in ashrams, kibbutzim, or similar communities, trading work for food, lodging, and participation in communal life. But most people are not satisfied with just the basics.

We want more—new smartphones, travel to exotic places, tennis lessons, gourmet meals, trendy clothes, and other markers of personal growth or status. R&AIs may commoditize many goods and services, making them cheaper and more accessible. But over time, these will become the new baseline. The desire for more will remain.

R&AIs might reduce the number of hours we need to work, especially for those content with a simpler life. But it is unlikely they will eliminate work altogether or even in large parts. Human nature drives us to grow, strive, and improve—and that will always require work.

Individuals should learn: How to Work with AIs, and Areas of Exclusive Mastery

The first step is learning how to make the best use of AIs—understanding their strengths, adapting them to your needs, and recognizing their limits. Human-AI interaction happens in three directions, and each one requires different techniques and approaches:

• Above: Receiving instructions from AI to learn new knowledge/skills or improve existing ones.
• Alongside: Collaborating with AI as a partner in your creative or analytical projects.
• Below: Delegating tasks to AI, so it handles repetitive or derivative work while you focus on higher-level thinking.

R&AIs will dominate well-documented, repetitive, and derivative tasks. To remain valuable, individuals must pursue skills that are rare, complex, or evolving. These areas that few have mastered, will tend to fall into two categories:

• Hyper-Specialization: Becoming an expert in a very narrow field. Examples:
  • Instead of learning general cooking, specialize in rare areas like barbecuing or exotic mushrooms.
  • Instead of being a general frontend developer, master accessible web design for users with vision impairments.
• Hyper-Generalization: Gaining working knowledge in many fields to understand how parts fit together and evaluate AI-generated work. Examples:
  • Instead of focusing only on savory dishes, understand full meal design—nutrition, flavor balance, and complementary beverages.
  • Instead of focusing on frontend programming, become a “full stack developer” who oversees and integrates design, accessibility, SEO, database, frontend and backend.

Those caught in the middle—neither highly specialized nor broadly knowledgeable—are most at risk, as their work is often easy to document, repeat, derive and automate.

It is important to emphasize the word mastery. In the age of R&AI, there will be very few opportunities for junior-level workers with only surface-level skills. The market will reward individuals who hold deep, comprehensive knowledge in a specific field—true mastery.

As Naval Ravikant once said: “If you need a degree to do it, it’s not going to make you wealthy.” In the age of R&AI, this insight becomes even more urgent. If you could not become wealthy doing it before, you may soon find it impossible to even earn an average income. Knowledge and skills that are well-known, standardized, and easily taught in a degree program will soon be absorbed and replicated by machines. Everyone will need a differential—an area of exclusive mastery that sets them apart.

Achieving mastery might seem overwhelming. The bar is being raised, but thankfully, R&AI is also offering a ladder with gradual steps to help you reach it. While human teachers often struggle with time constraints, limited ability to adapt explanations, or the challenge of identifying a student's unique learning bottlenecks, R&AI does not suffer from these limitations.

For many students, missing a foundational concept can snowball into academic failure, sometimes leading to emotional trauma and a belief that they are incapable of learning.

R&AI will not suffer from the limitations of human teachers. These systems can explain subjects in many different ways, spend unlimited time focused on a single student, identify learning bottlenecks, propose tailored solutions, and generate infinite custom exercises to reinforce understanding. This allows for a level of educational support previously unimaginable.

In a poetic twist, the same fast feedback loops that allow us to train advanced machine learning systems will now be used by those systems to accelerate human learning. R&AIs will speed up our own slow feedback loops in education, enabling individuals to develop areas of mastery much faster than before.

In more technical terms, R&AI can better provide Formative Assessment (ongoing feedback), as opposite to summative assessment (feedback in the end)

Imagine an AI tutoring children through difficult math lessons using real-time diagnostics. Picture a humanoid robot spending hours helping you refine a single tennis move. Or a surgeon practicing complex procedures in a VR simulation, complete with unexpected complications to build real-world resilience.

True mastery, however, is never instant. It will still require time, discipline, and long-term planning. Perhaps the best advice comes from physicist Richard Feynman: "Study hard what interests you the most in the most undisciplined, irreverent, and original manner possible."

Finally, individuals must pressure governments to take effective action to adapt to the R&AI age.

To replace the jobs lost during the Great Automation, governments must do everything they can to unlock maximum competition and innovation. The clearest first step is to eliminate the biggest barrier to both: the patent system.

Patents have become a tool for entrenching power, not encouraging innovation, for a variety of reasons, including:

• They allow powerful companies to use AI to preemptively patent every conceivable invention, locking small and medium-sized businesses out of entire markets.
• They let companies shelve disruptive inventions to protect their oligopolies, rather than implementing them to create value and new jobs.
• They give corporations with large legal teams the ability to intimidate or bankrupt competitors through expensive litigation over vague patent boundaries.
• They slow down innovation, allowing automation to destroy jobs faster than new ones are created.
• They limit exposure to new technologies, making it harder for workers to learn, adapt, and become future innovators or raise their income.

Other pro-competition and innovation measures can be:

• Reduce red tape and simplify business regulations to encourage entrepreneurship.
• Streamline government processes so new businesses can launch and scale quickly, and existing businesses can operate with less overhead.
• Removing artificial scarcity in sectors like housing is also crucial. Many laws limit high-density construction and prevent cities from meeting demand. Ending these restrictions would create millions of jobs while bringing down rent and mortgage costs—giving struggling workers and companies more financial breathing room.

Governments may also need to launch public information campaigns. These would help workers and businesses understand which industries are most vulnerable to automation, and which emerging sectors are growing or show long-term promise.