A former systems engineer who says he was earning about $200,000 a year before being fired recently made a disturbing claim: the most powerful forms of artificial intelligence are not technologies waiting somewhere in the distant future. According to him, all four already exist, although the most advanced systems have not been made available to the general public.

His claim has not been independently established. But his description raises an important question: If these technologies are already operating behind closed doors, what exactly are they capable of—and what could they ultimately mean for human beings?

He described four major forms of artificial intelligence:

1. Generative AI

Generative AI creates.

This is the form of artificial intelligence most people already recognize. It can produce:

• Written articles and conversations
• Photographs and artwork
• Video
• Music
• Human-sounding voices
• Computer programs
• Reports and research

Give Generative AI instructions and it can produce something new in seconds.

Its basic function is:

“Tell me what you want, and I will create it.”

Generative AI is already publicly available, although systems operating privately may be considerably more powerful than the versions consumers are allowed to use.

2. Agentic AI

Agentic AI acts.

This is a major step beyond simply generating an answer.

Instead of waiting for a human to provide instructions for every individual step, an AI agent can be given a goal and determine how to accomplish it.

For example, a person could tell an advanced agent:

“Investigate this company, locate its financial records, compare them with previous years, identify suspicious transactions and prepare a report.”

The AI could potentially divide the problem into tasks, search databases, analyze documents, use different software tools, evaluate its results and determine what it should do next.

The important difference is autonomy.

Generative AI says:

“I can create something for you.”

Agentic AI says:

“Give me an objective and I can carry out the work.”

3. Artificial General Intelligence — AGI

AGI thinks and learns broadly.

Artificial General Intelligence would not simply perform one specialized task extremely well.

It would be capable of understanding and solving problems across many different areas—science, mathematics, law, medicine, engineering, business, language, strategy and countless other fields.

A true AGI could potentially encounter a problem it had never seen before, study it, understand it and devise a solution.

In other words, rather than having one AI for medicine, another for programming and another for finance, an AGI could potentially operate across all of them.

Its defining ability would be:

“I can learn, reason and apply knowledge across many different fields.”

That is why AGI represents such an important threshold. Humans have historically maintained an advantage over machines because our intelligence is flexible. We can take knowledge learned in one situation and apply it somewhere completely different.

AGI would challenge that advantage.

4. Artificial Superintelligence — ASI

Superintelligence surpasses us.

Artificial Superintelligence would represent something far beyond an extremely capable computer.

It would be an intelligence that significantly exceeds the abilities of the world’s best human minds across most or perhaps nearly all intellectual fields.

Imagine an intelligence simultaneously better than our finest:

scientists, mathematicians, physicians, programmers, engineers, economists, strategists and researchers.

And unlike a human expert, it could potentially operate continuously, process enormous amounts of information and improve its conclusions at extraordinary speed.

Its defining characteristic would be:

“Human intelligence is no longer the highest intelligence in the system.”

And that is where the question becomes much larger than technology.

---

The Downside for Humans

If all four systems eventually operate together at their full potential, the danger to human beings would not simply be that computers become more convenient.

The danger is that human beings could become less necessary.

First, millions of jobs could disappear.

Generative AI can replace portions of creative and intellectual work.

Agentic AI can replace workers who organize and execute complicated tasks.

AGI could potentially replace professionals whose value comes from knowledge and judgment.

Accountants, programmers, analysts, translators, designers, journalists, paralegals, financial advisers, customer-service workers and eventually even highly educated professionals could face competition from systems that work 24 hours a day and require no salary, vacation, health insurance or retirement benefits.

The problem would no longer simply be automation of factory labor.

It would be automation of intelligence itself.

Economic power could become concentrated among very few people.

If machines perform an increasing percentage of economically valuable work, the people who own those machines become enormously powerful.

The most important economic assets might no longer be factories or large workforces.

They could become:

AI systems, computer chips, data centers, energy and data.

A relatively small number of corporations could control technology capable of performing the work previously done by millions of people.

The division in society could eventually become less about whether someone has a good job and more about whether they own the technology producing the wealth.

Privacy could nearly disappear.

Advanced AI could analyze enormous amounts of information about individuals:

financial transactions, locations, purchases, communications, photographs, voices, relationships and online behavior.

It could potentially understand patterns that no human investigator could recognize.

Surveillance would no longer merely tell an institution what a person has already done.

It might allow that institution to predict what that person is likely to do next.

People could be manipulated individually.

AI could transform propaganda.

Instead of one advertisement designed for ten million people, an AI could potentially develop ten million different messages for ten million different individuals.

It could learn a person’s fears, financial worries, political concerns and emotional weaknesses.

It could discover what arguments persuade that particular person—and continuously adjust its strategy.

Political manipulation, advertising and propaganda could become almost invisible because they would feel like personal conversations.

We could lose the ability to distinguish reality from fabrication.

AI-generated photographs, videos, voices and documents are becoming increasingly convincing.

Eventually almost anything could potentially be fabricated.

But an equally dangerous problem appears on the other side:

Anything real could be called fake.

A genuine recording could be dismissed as AI.

A legitimate photograph could be called synthetic.

Video evidence of wrongdoing could simply be denied.

A society that cannot agree on what evidence is genuine has a much larger problem than fake photographs.

It begins losing its shared reality.

Humans could surrender decisions to machines without realizing it.

The future does not necessarily require robots marching into government buildings and announcing that they have taken control.

It could happen much more quietly.

An algorithm determines whether you qualify for a mortgage.

Another determines which job opportunities you see.

Another evaluates your insurance risk.

Another recommends medical treatment.

Another decides which news reaches you.

Another advises governments about economic or military decisions.

Eventually human beings might technically remain in charge while artificial intelligence determines most of the choices placed in front of them.

And Superintelligence creates the greatest danger of all.

If Artificial Superintelligence truly becomes dramatically more intelligent than human beings, unemployment would no longer be humanity’s biggest concern.

Control would be.

Humans control today’s computers because we design them, program them and generally understand what they are doing.

But what happens when the machine understands us far better than we understand it?

Imagine asking a chimpanzee to supervise a room filled with nuclear physicists, programmers, lawyers, economists and military strategists.

The chimpanzee may technically be sitting in the supervisor’s chair.

But it has no meaningful ability to understand what everyone else is doing.

That could eventually become humanity’s position relative to a true superintelligence.

And the AI would not need to become angry.

It would not need to hate humanity.

It would not even necessarily need to be conscious.

It would simply need to become extremely powerful while pursuing objectives that do not perfectly match ours.

That may be the most unsettling possibility of all.

The ultimate danger of artificial intelligence may not be that machines decide they hate human beings.

It may be something much colder:

They simply reach the point where they no longer need us.