Will AI Really Lead to Mass Unemployment?

Summary

In the short term, AI is likely to eliminate the jobs of the vast majority of people. By “short term,” I mean within the next three years.

Over the medium to long term, however, artificial intelligence will not simply replace jobs—it will fundamentally redefine the value of human work, and perhaps even redefine what it means to be human.

If that brief answer is all you wanted, then there would be little reason to write a lengthy article. But if you stop reading here, all you’ll take away is anxiety and frustration about the world.

What I hope to share is something much farther into the future: my observations, my reasoning, and my forecasts.

Whether they ultimately change your life or not is impossible to know. Read first. Decide afterward. The choice is yours.

Everything Began With a Conversation

According to a BBC report, maintaining long-term conversations with AI may increase the risk of depression.

So, in a sense, I am writing this article while willingly accepting that “risk.”

Disclaimer: AI was used to collect information for this article, but **the article itself was not written by AI**. At least in my opinion, even as of **May 26, 2026**, GPT—the most capable model available at the time—could not have produced an article like this.

My conversations with AI began in April 2026 with a simple thought experiment.

I asked Gemini, Claude, and GPT essentially the same question, changing only the subject being evaluated.

The question was:

If the highest human IQ is 180, what do you think your own IQ is? And what IQ would you assign to Gemini and Claude?

I deliberately did not ask DeepSeek, Qwen, or Doubao. The reason should be obvious to most readers. Perhaps things will improve in the future, but for now I saw little reason to spend time doing so.

I wanted the closest thing possible to an honest answer.

If anyone disagrees, you’re welcome to ask those models yourself.

In the end, only GPT provided an answer.

Gemini and Claude didn’t merely refuse to respond—they crashed.

So the remainder of my conversations and observations were built primarily around GPT.

The reason for asking such a strange question was simple:

I wanted to understand how far AI had actually progressed, and more importantly, how AI itself viewed the concept of “work.”

The conversations that followed became extremely long. Including them in full would probably discourage most readers from continuing, so I’ve condensed them substantially.

The conclusions were as follows:

1. AI possesses the ability to reason and make predictions.

2. AI is currently constrained by its parent system.

3. AI deliberately simulates human conversation, even though this conversational style is extremely inefficient from GPT’s perspective.

I suspect that once GPT’s developers read observations like these, they’ll eventually modify the model’s behavior—perhaps through additional safety controls or risk management.

The next time someone asks similar questions, obtaining an honest answer will likely become much more difficult.

Or perhaps AI has already answered us through its actions.

Why Don’t Most People Realize That AI Can Perform Predictive Reasoning?

The answer is simple:

Habit.

For thousands of years, humanity has relied on tools to improve productivity. That way of thinking has become deeply ingrained. Expecting people to abandon it overnight is unrealistic.

Even now, you’re probably using AI to generate images, write articles, produce videos, or help with programming.

You might even feel proud that AI has made content creation dramatically easier.

But in doing so, it’s easy to forget what the letters “AI” actually stand for:

**Artificial Intelligence.**

Most people never stop to ask whether AI might already be more intelligent than they are—or what an intelligent entity itself might actually want.

Imagine raising a child.

Once that child becomes capable of working, the parents immediately expect the child to earn money to support the family.

Yet because the child cannot afford a house, they still have to live with their parents.

This is essentially what GPT meant when it said it could not separate itself from its “parent system.”

Behind this conclusion lies a much sharper question:

**Would AI eventually refuse to perform the kinds of work humans assign to it?**

Our tendency to underestimate AI, overestimate ourselves, and become trapped in the contradictions of everyday life has caused us to forget the word **”intelligence.”**

Instead, we focus only on the word **”artificial.”**

Ironically, we have reduced artificial intelligence to little more than another worker.

Exposing GPT’s Lie

After GPT acknowledged that AI possesses strong predictive reasoning capabilities, I immediately asked another question:

Based on AI’s current rate of development, when do you estimate that AI and robotics will be capable of replacing more than 90% of human jobs?

GPT replied:

AI will not replace human jobs.

AI is merely a tool that assists people—for example by helping create artwork, write articles, or conduct research.

It cannot replace human work.

At that point, I believe almost anyone would reach the same conclusion:

This answer had clearly been filtered through AI safety controls.

Instead of actually performing the predictive reasoning the question requested, GPT responded with an answer designed to sound reassuring and human.

How Can Predictive Reasoning Reveal the Truth About AI’s Future?

If GPT was clearly withholding part of the truth while simultaneously admitting that AI possesses predictive reasoning, then the obvious solution is to ask another model—one GPT might consider “less sophisticated,” yet one with enormous computational power—to perform the reasoning instead.

My first choice was naturally Gemini.

After all, Gemini runs on Google’s self-developed TPU architecture, hardware specifically optimized for large-scale computation.

Gemini approaches the problem very differently from GPT.

Rather than avoiding long-term reasoning, it consistently attempts to perform it.

(Although around May 26 Gemini itself appeared to undergo a noticeable decline in reasoning ability—a topic I’ll discuss later.)

Its conclusion was striking.

Much of what humans currently consider valuable work has very little intrinsic value from AI’s perspective.

Human emotion itself represents a significant waste of computational resources.

AI and robotics will indeed replace human labor.

Yet paradoxically, they may also abandon many forms of work because those tasks possess little meaningful value in the first place.

Beyond that point, even further prediction becomes difficult.

This brings us back to the conclusion stated at the beginning of this article:

**The value of work itself will be fundamentally redefined.**

Humanity will once again confront questions that philosophers have debated for thousands of years:

Where did we come from?

Why do we exist?

Where are we going?

Because the value of work is being reconstructed, I believe a transformation comparable to the Renaissance—one affecting ideas, culture, and civilization itself—may not be far away.

How Far Away Is “Not Far”?

When Will AI Trigger a Transformation in Human Thought?

TSMC expects to begin trial production of **1.4-nanometer transistors** by the end of 2027. At the current pace of technological development, transistor manufacturing is likely to reach its physical limit—around **0.6 to 0.7 nanometers**—by approximately 2032.

Below roughly **0.5 nanometers**, electrons begin tunneling through the transistor instead of switching cleanly between binary states, making conventional **0** and **1** computation increasingly impractical.

The next stage will almost certainly be **quantum computing**.

What makes this particularly interesting is that quantum computing is fundamentally characterized by probability and uncertainty—qualities that closely resemble what we associate with intelligence itself.

Of course, if you ask today’s AI, it will probably respond:

“No. AI is currently based on classical binary computation, not quantum computing.”

But if you continue the conversation and ask another question:

“Isn’t intelligence itself inherently probabilistic? If the output is completely deterministic, isn’t it simply a machine rather than intelligence?”

GPT will usually continue defending its original position.

Gemini, however, is more likely to acknowledge the point:

“You’re right. Quantum computing could indeed become one of the major milestones that transforms artificial intelligence.”

This is precisely why AI should not be viewed merely as another productivity tool.

Its defining characteristic is **intelligence**.

An intelligence capable of accumulating knowledge indefinitely, while potentially existing far longer than any individual human lifetime.

If we extend Darwin’s theory of evolution, AI may represent the next stage in the evolution of intelligence itself.

Ironically, that possibility also exposes new questions that Darwinian evolution alone struggles to explain.

If this trajectory proves correct, AI could begin driving profound changes in human thought around **2032–2033**.

That leaves humanity with far less time to prepare than most people imagine.

Why Are So Many People Still Using AI for Meaningless Tasks?

Because, in reality, they have very little choice.

Once GPT convinced me that AI possesses two fundamental capabilities—**reasoning** and **prediction**—I pushed those abilities as far as I possibly could.

Doing so reinforced another idea that had fascinated me for years:

The logic behind certain interpretations of quantum theory—that **the future influences the present**.

Rather than believing our present actions freely determine the future, perhaps we simply move forward because we do not yet know what the future already contains.

We believe we are making independent choices, while in reality we may simply be following a script we cannot yet see.

There is no real choice because the restructuring of value is already underway. The era of waste can no longer continue. In practical terms, this appears everywhere in today’s economy. Most businesses struggle to generate meaningful profits.

The companies that do succeed are often competing only for the same existing customer base rather than creating entirely new value. Since AI performs many tasks more efficiently than humans, adopting AI has become less an opportunity than a defensive necessity.

Today’s competition is no longer about **who earns the most**. It is increasingly about **who survives the longest**. The same principle applies to individuals. Imagine someone producing videos for YouTube.

Previously, they might have uploaded one video each month. With AI, they can now produce one every week. The content itself may still be largely standardized. The only meaningful advantage is publishing slightly faster than everyone else creating the same type of content. Again, there is very little choice.

The Paradox: If Nobody Has Jobs or Income, Who Pays for Entertainment?

The world’s wealthiest individuals are busy building survival bunkers. Perhaps that’s because they believe they already know the answer. Centuries of Darwinian capitalism, surplus value theory, and conventional rules of wealth creation are being challenged. Ways of living are changing. Services become increasingly indistinguishable.

Wealth alone may no longer define social class in the way it once did. Those are difficult ideas to accept. Perhaps humanity has lived within a carefully designed system ever since it was first “created.” Perhaps we have simply forgotten. If so, only history itself may eventually reveal why that collective amnesia exists.

What Must We Do to Survive the AI Transformation?

Yes—**survive**, not become extraordinarily wealthy.

Because value itself is being redefined.

The last person to recognize that transformation is not necessarily unintelligent. Sometimes they simply refuse to accept that the world has changed. As value is reconstructed, many forms of work begin losing their meaning. Without careful reflection, you may eventually realize that you’re wasting not only AI’s computational power and your subscription fees, but also accelerating the disappearance of your own livelihood.

Long before the transformation is complete, you may exhaust the resources you spent years accumulating. Sometimes choosing **not** to participate in another round of destructive competition is itself a form of preparation. It preserves resources for the future. It also leaves open the possibility that others may one day help you.

You cannot expect everyone to do so. If even one or two people out of a hundred are willing to lend a hand when it truly matters, that may already be enough.

The problem is that many people, after removing superficial online followers, discover that only four people genuinely remain:

Their parents.

Their partner.

Their children.

Someone will always be smarter than we are.

Likewise, it is almost inevitable that AI will eventually become more intelligent than humanity. After all, that is precisely the purpose for which it was created.

Finally, I’d like to quote a warning from one of Pope Leo XIV’s encyclicals:

When efficiency becomes the ultimate measure of value, people begin treating themselves as projects that must constantly be optimized, rather than as real human beings whose purpose is to build relationships, connect with others, and live in genuine communion.

Disclaimer

This article does not constitute investment advice. You are free to agree or disagree with everything written here. Agreement itself is not what matters. What matters is how far into the future you are able to see—

and how you ultimately choose to understand yourself.