60 Minutes Tonight

Who Needs Humans? 60 Minutes Reports on the AI Revolution Reshaping the Future of Work

Ryan Gill

3 weeks ago

Artificial intelligence has already changed how millions of people search for information, write, code and create. The much bigger question is what happens when AI becomes capable of performing the work that people have spent years learning how to do.

That is the question at the heart of “Who Needs Humans?”, a 60 Minutes report from correspondent Jon Wertheim examining the rapidly changing relationship between artificial intelligence and human employment.

Rather than focusing only on engineers building AI systems, Wertheim looks at an increasingly important part of the industry: the doctors, lawyers, musicians and other experienced professionals who are being paid to teach AI the very skills that define their careers.

The report aired as part of the Sunday, October 4 edition of 60 Minutes on CBS.

Humans Are Teaching AI How to Do Their Jobs

Modern artificial intelligence may appear to learn almost magically, but increasingly sophisticated models still depend heavily on human knowledge.

One of the companies featured by 60 Minutes is Mercor, which connects AI developers with people who possess expertise in particular fields. When an AI company wants a model to become better at subjects ranging from medicine and finance to architecture and creative writing, specialists can be brought in to evaluate its work and teach it how an expert would approach a problem.

The work can involve comparing different AI answers, identifying errors, creating challenging examples or explaining what separates a merely plausible answer from a genuinely expert one.

That distinction has become increasingly important. General-purpose AI systems can already generate convincing-looking responses on an extraordinary range of topics. But sounding knowledgeable and actually applying professional judgment are not necessarily the same thing.

Mercor describes its role as connecting human expertise with advanced AI models, arguing that specialist judgment remains one of the key ingredients required to make those systems more capable.

There is an obvious irony, however. The better humans become at teaching AI their jobs, the more capable AI potentially becomes of performing those jobs without them.

A Doctor, Winemaker and Music Producer Train AI

60 Minutes demonstrates the process through several professionals working as AI trainers for Mercor.

Among them is physician Dr. Melania Poonacha, who has worked on projects designed to improve an AI system’s ability to understand and interpret laboratory results.

Music producer Robbie Hiser has performed a different kind of training. His assignments can involve judging competing versions of lyrics and providing feedback intended to teach an AI system what makes one creative choice stronger than another.

Winemaker Shawna Miller is another specialist putting years of accumulated knowledge to work for AI.

The three examples illustrate how dramatically AI training has evolved. Data labeling was once commonly associated with relatively simple jobs such as identifying objects in photographs. Increasingly, however, the industry wants something much harder to reproduce: professional judgment accumulated through years of education and experience.

That has created an unusual new occupation. People can now earn money precisely because they know something an AI system does not yet know — and their job is to reduce that gap.

Could Workers Be Training Their Own Replacements?

That leads directly to one of the most uncomfortable questions in the 60 Minutes report.

If a professional spends hours showing an artificial intelligence system how to perform tasks from his or her occupation, is that person effectively helping create a future competitor?

The AI trainers featured in the report do not necessarily see it that way.

Hiser argues that music still depends upon something machines fundamentally lack: lived experience. A computer can study enormous quantities of music and language, but it has never experienced heartbreak, relationships, ambition or the countless other experiences that human artists draw upon.

It is a distinction likely to become central to the future-of-work debate.

AI may become extremely good at reproducing the observable output of a profession. Whether it can replicate the judgment, experience, empathy, accountability and originality behind that work is another question entirely.

And the answer may differ dramatically from occupation to occupation.

What Is Mercor?

Mercor was founded in 2023 and has rapidly become one of the companies operating at the intersection of AI development and skilled human labour.

Its business reflects a broader change taking place across the AI industry. As models become more advanced, simply feeding them huge quantities of internet text is no longer enough. Developers increasingly need specialist examples, difficult evaluations and expert feedback that can expose weaknesses in a model’s reasoning.

That can mean recruiting professionals from hundreds of disciplines.

Mercor says its network now encompasses millions of domain experts, while other companies have created similar networks. Scale AI, for example, markets access to specialists across fields including finance, law, healthcare, engineering and research.

The result is an entirely new market for knowledge.

A person’s expertise no longer has value only when that person directly performs a job. It can also have value as training material for a machine that is learning to perform parts of that job.

A New Kind of Gig Economy

For workers, there is an immediate upside.

Specialists who participate in AI training can earn additional income using knowledge they already possess. The work is often remote and can fit around an existing career.

According to the 60 Minutes report, Mercor trainers can receive widely varying rates depending upon their expertise and assignments, with some specialist work attracting substantial hourly compensation.

But there is another side to the arrangement.

AI-training work can be project-based and unpredictable. A highly qualified professional may have lucrative assignments available one month and considerably less work the next.

That makes the emerging AI-training economy look in some ways like a highly educated version of the gig economy.

Instead of driving passengers or delivering meals, a physician might evaluate medical reasoning, an attorney might assess legal analysis, or an engineer might examine technical solutions.

The expertise may be highly specialized, but the underlying uncertainty of freelance work remains.

AI Could Remove Tasks Without Eliminating Entire Jobs

One of the strongest arguments against predictions of mass unemployment is that technology often replaces tasks rather than entire occupations.

Calculators did not eliminate accountants. Spreadsheets did not eliminate financial analysts. Computer-aided design did not eliminate architects.

Instead, the tools changed what those professionals spent their time doing.

Supporters of workplace AI believe the same pattern could happen again.

An attorney might use AI to search thousands of documents while concentrating on strategy and advocacy. A physician might allow an AI system to summarize test results while remaining responsible for diagnosis and communication with a patient.

A programmer might delegate routine coding to an AI assistant while concentrating on architecture and more difficult engineering problems.

In that version of the future, artificial intelligence does not eliminate professionals. It makes each professional dramatically more productive.

Mercor co-founder Brendan Foody presents a broadly optimistic version of this argument in the 60 Minutes report: repetitive portions of jobs may disappear, while human workers move toward other activities that remain valuable.

But Productivity Creates Another Problem

There is, however, an uncomfortable mathematical consequence of enormous improvements in productivity.

Imagine a company once needed 20 employees to complete a certain amount of work.

If AI allows five employees to produce the same output, the remaining question becomes whether the company suddenly needs four times as much work performed.

If demand grows enough, employment can remain stable or even increase.

If it does not, fewer workers may be required.

That distinction helps explain why the impact of AI could vary enormously between industries.

Professions experiencing rapidly growing demand may absorb productivity gains easily. Industries with relatively fixed demand may have far less reason to retain the same number of workers once each employee can accomplish considerably more.

Clara Shih Warns About Entry-Level Jobs

One of the more cautious voices featured by 60 Minutes is technology executive and entrepreneur Clara Shih.

Having previously overseen major AI operations at companies including Salesforce and Meta, Shih has seen firsthand how dramatically the technology can shrink the number of people required for certain projects.

She told Wertheim that work that previously required substantial teams can increasingly be accomplished by only a few people using AI agents.

Her concern extends particularly to jobs built around preparing information, conducting research and analysing data for other people within an organisation.

Those responsibilities frequently belong to younger and entry-level employees.

That creates another potential problem beyond immediate unemployment.

What Happens If AI Eliminates the Bottom Rung of the Career Ladder?

Entry-level jobs do more than provide inexpensive labour.

They are where future experts learn.

A junior lawyer reviews documents before becoming a senior attorney. A young analyst prepares reports before becoming an executive. A trainee programmer fixes relatively simple problems before becoming capable of designing complex systems.

If AI performs much of that introductory work, companies may eventually face a paradox.

They may need experienced professionals but have fewer opportunities through which inexperienced workers can become experienced professionals.

The question then becomes not merely which jobs AI can perform, but how people will acquire the expertise required for the jobs that remain.

It could eventually force companies and universities to rethink professional training itself.

An Economist Offers a Much More Cautious View

60 Minutes also turns to Daron Acemoglu, the MIT economist and Nobel laureate whose research has extensively examined technology, automation and labour.

Acemoglu disputes the assumption that technological progress automatically guarantees that enough new jobs will emerge to compensate for those that disappear.

One reason is speed.

Previous technological revolutions transformed the economy over generations. The current AI transition is potentially spreading through numerous white-collar professions simultaneously and at a much faster pace.

Acemoglu tells Wertheim that what AI will be capable of doing 15 years from now is difficult to predict — an important qualification in a debate where confident forecasts are common on both sides.

This Time, White-Collar Workers Are in the Spotlight

Automation has traditionally been associated with factories.

Machines replaced certain forms of physical labour, while computers eliminated huge amounts of clerical work.

Generative AI is different because many of the tasks it performs belong to occupations once considered relatively protected from automation.

AI systems can already draft contracts, write computer code, analyze financial information, create advertising material, summarize scientific papers and generate images, music and video.

That means lawyers, programmers, designers, consultants, analysts and other professionals are confronting questions factory workers have faced during previous waves of technological change.

The knowledge economy itself is now being automated.

The Human Advantage May Keep Moving

One tempting response is to identify things AI supposedly cannot do and assume they will remain permanently human.

Creativity. Judgment. Empathy. Complex reasoning. Strategy.

The problem is that the boundary has repeatedly moved.

Tasks that looked extraordinary for AI only a few years ago are increasingly routine.

That does not mean machines will inevitably become better than humans at everything. It does mean that defining a profession around what AI currently cannot do may be risky.

The more durable advantage may instead come from learning how to work alongside increasingly capable systems.

Human Expertise Is Becoming More Valuable — and More Vulnerable

There is a fascinating contradiction running through the entire AI-training industry.

Advanced models need human expertise precisely because that expertise is valuable.

A medical AI needs physicians because physicians understand medicine. A legal model needs lawyers because lawyers understand law. A financial system needs experienced professionals because years of real-world knowledge cannot simply be invented.

Yet every successful transfer of that knowledge can make the machine more capable.

Human expertise is therefore simultaneously becoming more valuable as training material and potentially less scarce as a service.

That tension is one of the most important economic questions surrounding artificial intelligence.

A Revolution That Has Already Started

Debates about artificial general intelligence can sometimes make AI sound like a distant issue.

The workplace transition documented by 60 Minutes is much more immediate.

Companies are already deploying AI. Professionals are already training it. Businesses are already reorganizing jobs around it.

Workers are already asking which parts of their occupations can be automated and which skills will remain valuable.

Even AI training itself has emerged as a growing category of employment, creating jobs that would scarcely have existed before the recent generative-AI boom.

The transformation therefore does not require some hypothetical future machine capable of thinking exactly like a person.

It is happening with the technology that exists now.

Who Needs Humans?

The provocative title of the 60 Minutes segment ultimately has no simple answer.

AI still depends heavily on humans. People develop the systems, supply training material, evaluate their output and provide the specialist knowledge needed to improve them.

But the relationship is changing.

Humans are no longer merely telling computers what instructions to execute. They are increasingly showing computers how humans themselves perform complex work.

Whether that process creates a world in which workers become vastly more productive or one in which fewer workers are needed may become one of the defining economic questions of the AI era.

For now, one thing is clear: the contest between human expertise and artificial intelligence is not really a contest at all.

Humans are helping build the competitor.

More 60 Minutes October 4 2026

More Feature Articles

Exit mobile version