NVIDIA CEO Jensen Huang on Why AI Won't Replace Your Job: The Radiology Paradox

About 10 years ago, a world-famous computer scientist asserted:

"The first job to disappear due to AI will be the radiologist."

Ten years later, this prediction has only come half true.

This anecdote was shared by NVIDIA CEO Jensen Huang during a talk at the Milken Institute in May 2024.

First, the half that came true.

Computer vision has become completely superhuman in the narrow task of interpreting scans. It can maintain focus longer than a human and pick up tiny abnormalities that humans might miss. According to Huang, ten years later, AI has 100% penetration in radiology.

But the half that missed was profound.

The job of a radiologist did not disappear. In fact, the opposite happened.

By having AI handle the reading of scans, doctors were able to read more scans. They could take on more patients and diagnose diseases more accurately.

As a result, hospital revenue increased, radiology became one of the largest revenue-generating departments in hospitals, and now hospitals want to hire even more radiologists.

Huang also pointed out that if everyone had believed that prediction and stopped aspiring to be radiologists, the world would have lost this critically important talent.

I want people who feel uneasy every time they see news about AI potentially taking their jobs to understand the reason behind this reversal.

By the time you finish reading, that vague anxiety should turn into a concrete action: "Tonight, I'll divide my work into two parts on paper and try handing one task to AI tomorrow."

The "Task" Disappeared, but the "Purpose" of the Work Did Not

Why did the prediction fail?

Huang's answer was essentially one sentence.

"What everyone overlooks is that the purpose of a job and the tasks of a job are related, but they are not the same thing."

The purpose of a radiologist is not to stare at scans on a workstation in a dark room.

It is to collaborate with other doctors to diagnose diseases and cure patients. Reading scans is merely a "task" to achieve that.

Therefore, even if the task was automated, the job did not disappear. Instead, the time that could be devoted to the purpose increased, the number of patients seen increased, and the work shifted toward growth.

Huang then used himself as an example, which was quite interesting.

"100% of the tasks I do for work are typing and talking. AI has already completely automated both typing and talking and is completely superhuman at them. If so, I should be unemployed. Yet, we are working harder than ever."

He says the same applies to software engineers.

The purpose of an engineer is not to type code, but to solve problems and create new things. He even joked that no child who moved to America at age nine did so because they dreamed of a life spent typing keys at a small screen from morning to night.

This applies directly to your work as well.

Creating documents. Summarizing meeting minutes. Transcribing numbers. Replying to emails. Those are tasks.

Delighting customers. Moving the team forward. Generating sales. Those are purposes.

AI is coming to eliminate the tasks.

Huang Discusses the Current State: "AI Has Finally Become Useful in These Last Few Months"

So, where is AI now?

I think the true nature of anxiety is the feeling that "evolution is so fast that the big picture is invisible." Huang's organization of this was very easy to understand as a roadmap.

Two years ago, ChatGPT appeared and generative AI was born. According to Huang, there were two essential aspects to being able to "generate."

First: Thinking is the generation of thoughts (tokens) in the mind. So the moment generation became possible, the path opened for AI to think and reason.

Second: To use tools, one must generate commands. Even to operate a browser, you have to generate words to control something.

This reasoning AI emerged last year, and we are now in the stage of "agentic AI" that can understand, reason, plan, and use tools to accomplish useful things.

The symbol cited was Anthropic's Claude Code.

He noted it was the first agentic system to handle truly productive work like software coding. The important thing here is that Huang emphasized that "coding is not just for engineers."

Coding is "codifying things you want to automate repeatedly into a program." There is no company in the world that has nothing it wants to automate. Therefore, coding is actually critically important for every company.

And this change has created an explosion in computation.

The amount of computation required for agentic AI is about 1,000 times that of generative AI. Multiply that by the fact that "the number of people who want to use it has increased 100-fold."

That's why the demand for GPUs is exploding, and there was even an anecdote that GPUs sold 4-5 years ago are appreciating in value faster than fine wine.

Furthermore, Huang pointed out that the gross margins of AI-native companies like OpenAI and Anthropic have turned significantly positive in the last 3-6 months, and he stated flatly:

"AI has finally become useful in these last few months."

The way we use computers is also going to change.

Until now, it was about "retrieving" things that someone had previously created and saved.

From now on, when you convey your intent as if talking to a person, the AI will think of the method, make a plan, master tools like browsers, Excel, or Photoshop, and return the finished product.

While you are afraid, the tools are continuing to advance in this direction.

Jobs Won't Disappear, but "Everyone's Job" Will Be Affected

Up to this point, it might sound like mere optimism. However, Huang also spoke clearly about the reality of employment.

First, what AI is doing initially is creating a massive number of jobs.

Chip factories, computer factories, AI factories. A multi-trillion dollar re-industrialization is occurring across these three types of factories.

Last year, $100 billion—what Huang calls the largest investment in human history—flowed into AI-related startups, and all of it turned into jobs.

There is also an interesting paradox.

Even though coding was the first thing AI became good at, job openings for software engineers are actually increasing. The reason is that ambition has increased. With AI, more can be done. Therefore, more people are hired.

However, Huang also explicitly mentioned "dislocation."

"If a student graduating from university today cannot master AI, they will not be able to take jobs from graduates who can."

"Skills that were unnecessary yesterday become essential today."

Operations where the task itself was the job will indeed be replaced. Huang cited telephone reservations for restaurants as an example. The task of just taking a call and accepting a reservation will move to AI. But that person, instead of being a phone operator, will be able to face the customers in front of them.

The conclusion is this:

"Many jobs will be created. Some jobs will disappear. But all jobs will be affected."

In other words, there is no windless safe zone. But it's not a story of despair either. The dividing line isn't the occupation; it's whether you are on the side of using AI. That's the point.

The Biggest Loser is the Person Who is Too Afraid to Touch AI

In the talk, the conflict between AI pessimists (doomers) and optimists (boomers) was also a topic.

When asked if he was the leading optimist, Huang's answer was, "I am a pragmatist."

His response when faced with the theory by the "Godfather of AI" Geoffrey Hinton—that there is a 20-30% chance AI ends human existence—was also impressive.

"Where he is completely wrong is in thinking that a large number of smart people are not working to prevent that."

For every person trying to make a car faster, there are ten people trying to make it safer. For every person trying to make AI smarter, there are ten people working on guardrails and safety.

Furthermore, the "biggest concern" Huang raised was unexpected.

It wasn't about other countries possessing AI. It was that people in his own country, continuously fed sci-fi horror stories, would become too afraid to touch AI, and as a result, the country would lose its lead.

"The reason America benefited from the previous industrial revolution wasn't because we invented it, but because we applied it."

This is a story about a country, but I think it can be translated directly to individuals. The time spent being frightened by sensationalist articles and waiting to see is the biggest cost. Making it safe is the industry's job. Applying it is your job.

"That Ambition is Too Low. Increase Your Expectations 100-Fold"

At the end of the talk, when asked what he was thinking about now, Huang shared this story.

Researchers who used to spend months exploring new ideas can now do it in a day using AI. Months turned into one day.

Breakthroughs are happening in every field: energy science, climate science, biology, drug discovery, and physical science.

"If you could see what I see every day, you would be excited and realize this: no matter what ambition you had in the past, it wasn't enough. There is only one thing to change: raise your expectations about 100-fold."

So, what should we do tomorrow?

If we turn this talk into action, I think there are three things:

1. Tonight, write your work on paper and divide it into "Purpose" and "Task." For a radiologist, curing patients is the purpose, and reading scans is the task. Which is which for your job?

1. Tomorrow, try handing just one task to AI. It could be meeting minutes, a draft for a document, or research. Even if it doesn't go well, the moment you touch it, you've moved from the side of waiting to the side of using.

1. Devote the time you've freed up to the purpose. Customers, planning, results. Your value comes out in the parts you can't hand over to AI.

Will you be on the side that disappears or the side that increases?

The dividing line isn't occupation or age; it's whether you do this decomposition. That's how I took it.

Finally, a question.

What is the "purpose" of your work? And which "task" will you hand over to AI first tomorrow?

I'd love to hear from you in the quotes or replies.