A Warning to the World

I am deeply concerned by the extremely rapid deployment of artificial intelligence in the United States and, potentially, throughout the world.

I am not opposed to artificial intelligence.

On the contrary, AI may become a solution to many of humanity's most difficult problems. It may accelerate science, medicine, education, engineering, and economic development. It may help us accomplish things that are currently beyond the capacity of individual human beings and institutions.

But the same technology may also create new problems — sometimes problems far greater than those it was intended to solve.

My concern is therefore not that humanity is developing AI.

My concern is how quickly we are doing it, how widely we are deploying it, and whether human beings and our institutions are developing quickly enough to remain capable of governing what we create.

I believe we need to slow down.

Not because progress is bad.

Because uncontrolled acceleration increases the probability of catastrophic failure.

The Fundamental Problem Is Larger Than Artificial Intelligence

The problem we are facing is not unique to AI.

Throughout modern history, technological power has repeatedly increased faster than humanity's ability to understand, regulate, and morally govern that power.

Technology gives us new capabilities.

But capability is not wisdom.

Power is not maturity.

And intelligence embodied in a machine does not automatically make the people operating that machine wiser.

There is a fundamental principle that should concern us:

The more powerful the tool becomes, the more sophisticated the human being and the institutions using it must become.

A more complex technology requires a more complex subject capable of operating it responsibly.

The evolution of technology therefore has to be accompanied by an evolution of the human being, culture, ethics, education, law, political institutions, and systems of responsibility.

The disturbing fact is that this process is not merely lagging behind.

It is falling further behind.

We are increasing our technological capabilities faster than we are increasing our capacity to use them safely.

AI Does Not Replace Thinking — It Amplifies It

AI should not be understood simply as a replacement for human intelligence.

It is closer to an exoskeleton for intelligence and thought.

An exoskeleton does not eliminate the importance of the person wearing it. It amplifies the person's physical capabilities.

AI can do something similar for cognition.

It can amplify reasoning, creativity, analysis, memory, communication, planning, programming, scientific research, and decision-making.

But this creates a paradox.

If the person operating the system is careless, the system can amplify carelessness.

If an institution is incompetent, AI can amplify institutional incompetence.

If a government is reckless, AI can amplify recklessness.

If a society is manipulated, AI can make manipulation faster, cheaper, more precise, and more scalable.

The amplification of intelligence does not automatically mean the amplification of wisdom.

It may simply mean that whatever is already present in the human system becomes more powerful.

That is why the question is not only:

How intelligent can we make machines?

It is also:

How intelligent, responsible, disciplined, and morally mature must human beings become in order to use them safely?

This Warning Is Not New

I am not the first person to be worried about this.

Long before artificial intelligence became a mass technology, philosophers, scientists, historians, and technologists were already warning that technological power could outrun human maturity.

Their warnings came from different traditions and different historical circumstances.

But they repeatedly converged on the same problem:

humanity can create more than it can understand, control, or morally justify.

Ten Voices Before the Age of Modern AI

1. Günther Anders — philosopher

Anders described what he called Promethean shame: the gap between what human beings can manufacture and what they are psychologically and morally capable of imagining.

“The incapacity of our imagination to grasp the enormity of what we can produce.”

The warning is remarkably relevant today: our technical imagination can become larger than our moral imagination.

2. Hans Jonas — philosopher

In The Imperative of Responsibility, Jonas proposed a new principle for an age in which human action can affect the distant future:

“Act so that the effects of your action are compatible with the permanence of genuine human life.”

This provides one of the clearest foundations for a precautionary approach to advanced technology.

3. Jacques Ellul — sociologist and philosopher

Ellul warned that technique could cease to function merely as a collection of human tools and become an autonomous system whose internal logic increasingly shapes society.

“Technique is autonomous.”

The question is therefore not only what technology allows us to do. It is whether the logic of technological efficiency eventually begins determining what we do.

4. Lewis Mumford — historian of technology

Mumford warned against treating technological progress as an end in itself. He described the Megamachine — the convergence of technological, economic, political, and organizational power into systems capable of subordinating human purposes to their own logic.

His warning was ultimately about what happens when technical and organizational power becomes greater than the maturity of the institutions directing it.

5. Norbert Wiener — founder of cybernetics

Wiener understood very early that technological progress changes not only machines but also the human environment.

“We have modified our environment so radically that we must now modify ourselves.”

If we change the environment of human thought radically enough, the human being must adapt as well.

6. Erich Fromm — philosopher and social psychologist

Fromm warned about the transformation of human beings into objects within technological and economic systems.

His critique of modernity repeatedly returned to alienation and the danger of treating the human being as a commodity rather than as a person.

The lesson for AI is obvious: we must not allow efficiency to become a substitute for human value.

7. Alvin Toffler — futurist

In Future Shock, Toffler described the psychological consequences of “too much change in too short a period of time.”

The problem was not technological change itself.

It was the velocity of change exceeding the capacity of individuals and institutions to adapt.

That problem has not disappeared. It has become faster.

8. Ivan Illich — philosopher

Illich distinguished between tools that expand human autonomy and systems that gradually make human beings dependent upon them.

His concept of convivial tools was based on the idea that tools should remain under meaningful human control.

His warning is particularly relevant to AI:

A tool should expand human agency, not replace it.

9. Karl Popper — philosopher of science

Popper rejected the idea that society should be redesigned according to grand, irreversible plans.

His preferred method was piecemeal social engineering:

“He will make his way, step by step, carefully comparing the results expected with the results achieved.”

This is precisely the principle that should govern technologies capable of affecting entire societies:

Experiment → Measure → Learn → Correct → Proceed

10. Zygmunt Bauman — sociologist

Bauman examined the weakness of human responsibility when modern systems combine enormous organizational power with technical efficiency.

His analysis emphasized the danger created when technology, rationalized systems, and economic effectiveness become detached from moral responsibility.

Technology can make immoral action extraordinarily efficient.

That is not a hypothetical problem.

Human history has already demonstrated it.

And Now Artificial Intelligence Has Brought the Old Problem to a New Level

The warnings did not disappear when AI arrived.

They became more explicit.

And what is particularly significant is who is making them.

These are not warnings coming only from people who oppose technology. They include people who helped create the technological world we inhabit.

Ten Voices Specifically Warning About AI

1. Bill Joy — co-founder and Chief Scientist of Sun Microsystems

In 2000, Joy published Why the Future Doesn't Need Us, warning that robotics, genetic engineering, and nanotechnology could create technologies whose destructive potential exceeded our ability to control them.

“We must limit development of the technologies that are too dangerous.”

Joy's argument was not anti-technology. It was a warning against treating technological acceleration as an unquestionable good.

2. Elon Musk — technology entrepreneur

Musk has repeatedly warned about AI and called for regulatory oversight.

“With artificial intelligence we are summoning the demon.”

The important part of the warning is not the metaphor. It is the underlying question:

Can we safely create something whose capabilities may eventually exceed our ability to control it?

3. Stephen Hawking — theoretical physicist

“The development of full artificial intelligence could spell the end of the human race.”

His concern was particularly relevant to the problem of speed: technological development can occur much faster than biological evolution.

Human beings cannot simply evolve cognitively at the same rate as our machines.

4. Yuval Noah Harari — historian

Harari has focused on a different danger: not only superintelligence, but control over information, narratives, institutions, and human relationships.

“A social weapon of mass destruction.”

He has also argued that AI is the first technology in history that can take power away from us.

This is a crucial warning because catastrophe does not require a machine to become conscious.

Human societies can be destabilized long before that.

5. Nick Bostrom — philosopher

Bostrom's central warning is famous:

“Machine intelligence is the last invention that humanity will ever need to make.”

If we create a system capable of outperforming humans across intellectual domains, we may also be creating the system that designs what comes next.

Therefore the transition deserves extraordinary caution.

6. Max Tegmark — physicist

Tegmark helped organize the 2023 open letter calling for a pause in the development of increasingly powerful AI systems.

The letter warned against an “out-of-control race” to develop systems that humans could not reliably understand, predict, or control.

This is perhaps the most important political problem:

Even responsible actors can behave irresponsibly when they believe everyone else is accelerating.

7. Jaron Lanier — computer scientist and pioneer of virtual reality

Lanier approaches the issue from the opposite direction. Rather than treating AI as an alien intelligence, he argues that we should understand it as a human collaboration system.

His warning is fundamentally about responsibility:

“We can take responsibility.”

The danger may be that technology drives society toward mutual unintelligibility and collective irrationality.

This is an essential correction to the mythology surrounding AI:

Machines must never become an excuse for humans to abandon responsibility.

8. Stuart Russell — AI researcher, University of California, Berkeley

Russell has been particularly direct about the danger of competitive AI development.

“Play Russian roulette with every human being on earth.”

The underlying argument is that technologies with potentially irreversible consequences cannot be governed solely by market competition.

9. Geoffrey Hinton — pioneer of modern neural networks

Hinton is perhaps the most significant warning voice precisely because he helped create the foundations of modern AI.

He has estimated a 10–20% chance of human extinction from AI within the next three decades, while emphasizing that the number is deeply uncertain.

“We've never had to deal with things more intelligent than ourselves before.”

His warning is therefore not merely about the probability of a particular scenario. It is about entering a domain for which humanity has no historical precedent.

10. Sam Harris — neuroscientist and philosopher

Harris has argued that the existential-risk scenario should be taken seriously enough to justify a pause.

“I would definitely pause it.”

His position is not that AI can never be useful. It is that the alignment and control problem must be understood before humanity moves further into territory from which retreat may be impossible.

The Pattern Should Concern Us

These people disagree enormously.

They disagree about AI timelines. They disagree about the probability of catastrophe. They disagree about whether superintelligence is possible. They disagree about whether regulation, alignment research, international cooperation, or technological restraint is the correct solution.

Some are optimists.

Some are pessimists.

Some are building AI themselves.

Some are asking us to stop.

But there is one point on which their warnings repeatedly converge:

The consequences of advanced technology can become larger than the institutions governing it.

And that is precisely where caution becomes rational.

We Should Not Wait for Catastrophe to Prove That Caution Was Justified

There is a dangerous asymmetry here.

If we slow technological deployment unnecessarily, we lose time.

If we accelerate irresponsibly and something catastrophically goes wrong, we may lose far more than time.

Some consequences may be irreversible.

This means the normal logic of technological competition is insufficient.

In ordinary markets, failure is often survivable.

  • A company launches a bad product.
  • A project fails.
  • An investment loses money.
  • A competitor overtakes you.
  • You learn and try again.

But some technological failures do not provide that luxury.

There may be no second attempt.

There may be no rollback.

There may be no functioning institution left to correct the mistake.

When failure can be irreversible, speed must no longer be treated as an unquestionable virtue.

What I Am Asking of Those in Power

I am not calling for the abolition of artificial intelligence.

I am calling for responsible acceleration — progress with brakes, feedback mechanisms, reversibility, and institutional preparation.

1. Do not confuse acceleration with progress

A technology becoming available sooner does not automatically mean that humanity is better prepared to use it.

2. Introduce powerful systems gradually

Large-scale deployment should follow controlled stages rather than a single irreversible leap.

3. Preserve the possibility of retreat

Critical AI systems should be designed so that society can disable, restrict, or replace them.

Reversibility should be treated as a design requirement.

4. Do not delegate ultimate responsibility to machines

An AI system can make a recommendation. It can calculate. It can analyze. It can assist.

But where human lives, fundamental rights, military force, critical infrastructure, or civilization-scale decisions are concerned, responsibility must remain human.

5. Test societal consequences before mass deployment

Technical benchmarks are not enough.

We need social, political, psychological, economic, and security benchmarks.

6. Do not allow an arms race to determine the speed of civilization

If every actor accelerates because they fear that someone else will accelerate, humanity has created a mechanism in which everyone can become trapped into doing what nobody would choose independently.

This is precisely the kind of problem that requires international coordination.

7. Invest in the evolution of the human subject

Education must change. Political institutions must change. Professional standards must change. Ethics must change.

People must learn not merely how to use AI, but how to think when AI is available.

The ability to ask a machine for an answer is not the same as the ability to understand whether the answer should be trusted.

8. Protect human cognitive independence

If AI becomes an exoskeleton for intelligence, humanity must ensure that it does not simultaneously become a prosthesis replacing the underlying muscles.

A society that can no longer think without its machines is not necessarily more intelligent.

It may simply be more dependent.

9. Create global mechanisms for discussing technological limits

There are technologies whose consequences are too large to be determined solely by individual corporations or individual states.

Humanity already understands this principle in relation to nuclear weapons, biological weapons, and other civilization-scale risks.

AI increasingly belongs in the same conversation.

10. Adopt the principle of gradualism

Here Popper's lesson is especially relevant:

Proceed step by step, compare expected results with actual results, identify unintended consequences, and avoid changes whose complexity makes their consequences impossible to understand.

That is not fear.

That is engineering.

The Real Race Is Not Between America, China, Europe, or Anyone Else

The most dangerous framing would be:

Who will build the most powerful AI first?

The more important question is:

Who will learn how to use powerful AI without losing control of themselves?

The winner of the technological race should not be the society that reaches the most powerful technology first.

It should be the society that learns how to combine technological power with human maturity.

If one country develops powerful AI rapidly while its institutions remain immature, it may gain a short-term strategic advantage.

But if every major power follows the same logic, the result may be a global acceleration in which nobody feels capable of slowing down.

That is not progress.

That is a collective failure of coordination.

A Final Warning

We should remember that the greatest danger of technology may not be that machines become too intelligent.

It may be that human beings become too powerful without becoming sufficiently wise.

We have already seen this pattern.

We developed technologies capable of destroying cities before developing political institutions capable of eliminating war.

We developed industrial technologies capable of transforming the planet before fully understanding their ecological consequences.

We developed global communications before developing sufficient cultural mechanisms for resisting mass manipulation.

And now we are developing systems capable of amplifying human intelligence itself.

The scale of the next transformation may therefore be unprecedented.

We should not assume that because we created the technology, we automatically understand its consequences.

We should not assume that because we can build something, we are ready to deploy it everywhere.

We should not assume that because a technology can solve one problem, it cannot create another — perhaps larger — problem.

And above all, we should not allow the speed of technological development to become faster than the speed at which humanity can learn to govern it.

The Principle

I would therefore formulate the central principle of this memorandum as simply as possible:

Technology must not outrun humanity.

And for artificial intelligence:

The development of machine intelligence must be accompanied by the development of human intelligence, human responsibility, and the institutions capable of controlling the power we are creating.

We do not need to stop progress.

We need to make progress survivable.

We do not need to fear intelligence.

We need to make sure that intelligence — human and artificial — remains connected to responsibility.

We do not need to reject the future.

We need to make sure that we are mature enough to enter it.

Slow down where the consequences may be irreversible.
Experiment where the consequences can be contained.
Measure what happens.
Learn.
Adapt.
Then proceed.

The future does not belong to those who accelerate the fastest.

It belongs to those who can move forward without losing control of the direction in which they are going.

A Note on the Quotations

The quotations above have deliberately been kept short and traceable. Several formulations commonly circulated online and presented as direct quotations from Anders, Jonas, Wiener, Popper, and others are actually paraphrases or translations of longer arguments.

For a public memorandum, it is more responsible to use short quotations and distinguish them from the author's own formulation.

The intellectual lineage is therefore not being presented as evidence that all these thinkers agreed with one another. They did not.

It is being presented for a different reason:

The question of whether technological power can outrun human maturity is much older than artificial intelligence.

AI has not invented this problem.

It has made the problem considerably more urgent.