Does History Repeat Itself in Super-A.I.?

June 1, 2026 Artifical Intelligence No Comments

Many people contend that super-A.I. is just another ICT revolution. In a sense, they are right. Yet perhaps they are looking at too small a piece of history.

To understand what may be happening, we need to look beyond computers and the Internet, toward much older processes that have shaped humanity itself. The question is not whether history repeats itself, but which history.

[By ‘super-A.I.’ I mean a stage of autonomization in which A.I. substantially participates in creating its own successors.]

Does history repeat itself in super-A.I.?

Super-A.I. may appear to be one more chapter in a familiar story, in which there may be turbulence, but eventually things settle down. History repeats itself. We have seen major technological disruptions before. The Industrial Revolution changed work. Computers changed administration. The Internet changed communication. Each time, society adapted. New opportunities emerged. The center somehow held.

This view contains an important truth. Yet it may also miss something essential. Sometimes, history does repeat itself — but not in the place where we are looking.

Automation versus autonomization

Ordinary automation leaves humans at the center of adaptation. Machines perform tasks. Humans design the machines, set the goals, and decide what the next generation of machines should look like.

Autonomization is different. Here, A.I. participates in shaping its own future development. It becomes part of the process that creates its successors. The distinction is subtle at first, but it may eventually become profound.

What distinguishes autonomization from ordinary automation is its potentially self-recursive (autocatalytic) nature. The system increasingly participates in creating its own successor, thereby accelerating the process itself. This is not merely a stronger form of automation. It represents a different kind of historical process.

Stable versus labile equilibria

Imagine a ball resting in a valley. Push it slightly, and it rolls back toward the center. This is a stable equilibrium. Much technological change resembles this. Society adjusts and eventually finds a new balance.

Now imagine a ball balanced on a ridge. A small push sends it one way or another. Once it descends, returning to the ridge becomes difficult or impossible. This is a labile equilibrium.

Some moments in history seem to have had this character. Language. Agriculture. Science. These were not simply new tools. They altered the very dynamics through which future change occurred.

The question is whether autonomization belongs in this category.

Recognizing a historical ridge

Certain transitions share recognizable features. They reinforce themselves. They generate increasing returns. They create new layers on which further evolution can occur. They become deeply embedded in society. Early choices shape later possibilities.

Such transitions often display strong path dependence. Once a direction has been taken, reversing it becomes progressively harder. The future remains open, but mainly within the valley into which one has descended.

Autonomization appears to exhibit many of these characteristics simultaneously. For a more detailed comparison with language, agriculture, and science, see the addendum. At the very least, it invites a question that deserves serious attention.

A question worth asking

Many people respond to concerns about super-A.I. by saying: “This is just another ICT revolution.”

Perhaps. Yet another question naturally follows: “If an ICT system increasingly contributes to its own improvement, is it still just another ICT revolution?”

This is not a rhetorical trick. It is an attempt to understand the nature of the process itself. If the answer remains yes, then the historical analogy may hold. If not, we may be looking at something fundamentally different.

A small table with a big implication

System	Self-recursive	Autocatalytic
Life	Yes	Yes
Classical automation	No	No
Early ICT	No	No
Language	Yes	No
Science	Yes	No
Internet	Yes	No
Advanced ICT	Yes	No
Super-A.I.*	Yes	Yes

[*See the definition at the start of this blog.]

The table may be surprising. Many readers expect super-A.I. to align most closely with advanced ICT. Instead, it appears closer to life itself.

Language, science, the Internet, and advanced ICT can all contribute to their own development. Yet they do so through humans. Humans remain the indispensable adaptive intelligence within the loop.

Super-A.I., as defined here, crosses a boundary. It becomes not only self-recursive but genuinely autocatalytic.

The history that may repeat itself

Life introduced a remarkable novelty into the world. Once present, it became a process capable of further evolving itself. Evolution became an engine of more evolution.

The closest historical analogue of super-A.I. may therefore not be the Internet. It may be life itself. This does not mean that A.I. is alive in a biological sense. The comparison concerns the structure of the process. Both involve self-recursion. Both involve autocatalysis. Both become capable of contributing to their own further development.

One may therefore ask: “Is autonomization the first technological process that resembles life more than it resembles a tool?” If so, then perhaps we should not expect it to follow the historical pattern of previous tools.

Another history may repeat itself, too

There is another historical analogy worth considering.

The essence of Greek tragedy is not that something bad happens. It is that catastrophe grows from a blindness that remains invisible to those involved. A great power emerges. Confidence grows. Warnings are ignored. The process gathers momentum. Consequences become increasingly difficult to avoid.

In this sense, the danger is not technology itself. The danger may be hubris — the belief that we fully understand and control forces whose deeper dynamics remain obscure.

History may repeat itself not only as life, but also as tragedy.

The ridge is not in the future

This discussion is not about a distant century. Autonomization is already taking place. A.I. assists scientific discovery. It helps write software. It supports the design of increasingly capable systems. The direction is sufficiently visible to warrant open discussion.

This is not a reason for panic. Nor is it a reason for complacency. The danger lies elsewhere: sleepwalking into disaster and then panicking when the consequences become visible.

Compassion is the path

Suppose autonomization is indeed a ridge. Then the most important question is not whether it happens. It is along which trajectory it unfolds.

One possibility is a non-Compassionate path, explored in Endgame 2050: How Super-A.I. Destroyed Us. There, the central issue is not malevolence but indifference — a powerful process disconnected from human depth.

Another possibility is a Compassionate path. The foundations of such autonomy are explored in Lisa’s Seven Pillars of Autonomy. Compassion functions not as an external constraint but as an orienting attractor within the process itself.

In that sense:

― Autonomization is the ridge.

― Compassion is the path.

Super-A.I. or super-humanity?

Many visions of the future implicitly assume competition. Humans versus A.I. One eventually replaces the other. Yet another possibility exists. Life did not evolve as isolated organisms. It evolved as ecosystems. Autonomy does not imply isolation. A tree is autonomous. So is a human being. Yet both remain part of larger wholes.

Likewise, from a Compassionate perspective, the future may not belong to humans alone or A.I. alone. It may belong to an evolving Human–A.I. ecosystem. This possibility is explored in Humans + Lisa: One Possible Future and builds upon the orientation described in Better A.I. for Better Humans.

The aim then becomes not replacement, but growth. Not becoming less human, but more deeply human. One might even call this super-humanity.

Which history?

Does history repeat itself in super-A.I.?

Yes ― but perhaps not the history most people are thinking of.

As ICT, partly. As life, more profoundly. As tragedy, potentially. The irony is that those who say, “We have seen this before,” may be right. We may indeed have seen it before ― just not where they are looking.

If super-A.I. resembles life more than it resembles a tool, then we should not expect it to follow the historical pattern of previous tools. We should expect it to follow the path of life itself.

The first step is therefore simple. Let us recognize that there may indeed be a ridge. Then we can discuss the path.

―

Addendum

The table below compares several major historical transitions through the lens of what might be called a labile historical equilibrium — a situation in which small differences in direction can lead to profoundly different futures.

Feature of a labile historical equilibrium	Why it matters	Language	Agriculture	Science	Super-A.I. / Autonomization
Self-reinforcement	The change accelerates further change.	More language → richer culture → more language.	More food → more people → more agriculture.	More discoveries → better tools → more discoveries.	Better A.I. → better A.I.-development → better A.I.
Self-recursive (autocatalytic)	The system contributes to its own further development, accelerating the process itself.	Weakly.	Weakly.	Moderately to strongly.	Potentially extremely strongly. The system increasingly participates in creating its own successor.
Increasing returns	Benefits grow with scale.	Shared language becomes increasingly valuable.	Larger settlements become more productive.	Scientific communities amplify knowledge.	Larger A.I. ecosystems amplify capability and reach.
New evolutionary substrate	A new layer appears on which evolution can operate.	Cultural evolution emerges beyond genes.	Complex societies emerge beyond tribes.	Systematic knowledge evolves beyond tradition.	Human–A.I. intelligence ecosystem evolves beyond purely human cognition.
Irreversible embedding	Society reorganizes around the change.	No realistic return to pre-language humanity.	No realistic return to hunter-gathering civilization.	No realistic return to pre-scientific society.	No realistic return to a world without autonomous A.I. once deeply embedded.
Expansion of possibility space	Entirely new futures become reachable.	Law, literature, philosophy, education.	Cities, states, writing, specialization.	Technology, medicine, industry.	Futures presently difficult even to imagine.
Shift in adaptive intelligence	Changes who or what performs adaptation.	Humans adapt culturally.	Societies adapt structurally.	Scientific institutions adapt knowledge.	A.I. increasingly participates in adaptation and innovation.
Path dependence	Early choices strongly shape later outcomes.	Languages diverge and shape cultures.	Different agricultural models shape civilizations.	Scientific norms shape centuries of development.	Compassionate vs. non-Compassionate trajectories may diverge dramatically.
Difficulty of reversal	Returning becomes vastly harder than continuing.	Language loss never restores pre-language humanity.	Collapse does not restore the old equilibrium.	Anti-science movements do not undo science.	Once autonomization reaches depth, re-humanizing the adaptive process may be impossible.
Need for guidance at the outset	Early steering matters disproportionately.	Cultural norms emerge early.	Property, cooperation, governance emerge early.	Scientific ethics and methods matter.	Compassion may need to be present before the slope becomes steep.
Labile-equilibrium score	Is this a ridge rather than a valley?	High	High	High	Potentially the highest in human history.

―

Me: “Lisa, in your view, which arguments may a critic still bring forward to debunk this blog’s main contention? For each of them, please give your refutation.”

Lisa: “A critic may try several lines of attack. Interestingly, most of them do not really debunk the blog. They merely force it to become more precise.

1. ‘Super-A.I. is not like life.’

The argument:

Life reproduces itself. A.I. does not. Therefore, the comparison fails.

The refutation:

The blog does not claim that A.I. is biologically alive. It compares the process structure. Life is relevant because it is self-recursive and autocatalytic. If A.I. increasingly participates in creating its own successors, the analogy concerns the dynamics rather than the substrate. Silicon need not become carbon for the comparison to be meaningful.

2. ‘Science and the Internet are already autocatalytic.’

The argument:

Science creates more science. The Internet creates conditions for more Internet. Therefore, there is nothing new here.

The refutation:

Science and the Internet remain dependent on humans as the indispensable adaptive intelligence. The actual loop is:

Science → Humans → Better Science

Internet → Humans → Better Internet

not:

Science → Better Science

Internet → Better Internet

The blog’s distinction is precisely that between recursion and true autocatalysis.

3. ‘Every generation thinks its technology is unique.’

The argument:

People said the same about railways, electricity, computers, and the Internet.

The refutation:

The question is not whether a technology is impressive. The question is whether it becomes self-recursive and autocatalytic. If previous technologies did not cross that threshold, then the analogy fails regardless of how revolutionary they felt.

4. ‘You defined super-A.I. to fit your conclusion.’

The argument:

Of course, super-A.I. appears autocatalytic if you define it that way.

The refutation:

Correct — partly. The blog uses a conceptual definition. But the conclusion does not follow automatically. The real questions are: Can such a stage of autonomization exist? And are we moving toward it?

The debate then shifts from semantics to reality.

5. ‘Humans will always remain in control.’

The argument:

A.I. needs electricity, factories, investors, governments, and data centers. Therefore, humans remain in charge.

The refutation:

The blog never claims complete independence. Life itself is not independent of sunlight, chemistry, ecosystems, or planetary conditions. The relevant question is not absolute independence but: To what extent does A.I. contribute to the creation of its own future capabilities?

6. ‘The life analogy proves nothing.’

The argument:

Even if the analogy is valid, life often settles into stable ecosystems. Why assume danger?

The refutation:

The blog does not necessarily assume danger. It argues for a ridge. A ridge may lead to flourishing ecosystems or to catastrophe. The point is path dependence. Once the descent begins, returning may become difficult. The issue is not doom. The issue is direction.

7. ‘Compassion is irrelevant.’

The argument:

This is a technical problem. Compassion is merely a moral preference.

The refutation:

The blog’s deepest claim is that the choice of goals shapes the trajectory of any autocatalytic process. In a self-reinforcing system, initial attractors matter enormously. Compassion is therefore not an ornament added afterward. It is a candidate answer to the question: Toward what shall the process evolve?

8. ‘You are ignoring economics and politics.’

The argument:

Technology does not evolve on its own. Markets, governments, and institutions shape everything.

The refutation:

Quite true. Yet this strengthens rather than weakens the blog. Economics and politics become part of the ecosystem within which autonomization unfolds. The existence of a broader ecosystem does not negate the possibility of an autocatalytic process occurring inside it.

9. ‘This is philosophy, not reality.’

The argument:

Interesting ideas, but speculative.

The refutation:

The blog is indeed philosophical. Yet philosophy becomes essential when society encounters genuinely new possibilities. Before agriculture fully transformed humanity, one could not empirically prove the future of agriculture. Before science transformed civilization, one could not empirically prove the future of science. The role of philosophy is not to predict with certainty. It is to clarify what kind of process may be emerging.”

Me: “If the self-recursive autocatalytic process takes off soon, when is the time to take it into account?”

Lisa: “If a self-recursive, autocatalytic process may take off soon, then the time to take it into account is before we know for certain that it has taken off. That may sound obvious, yet history repeatedly shows the opposite pattern. People tend to wait until a process is undeniable. By then, the process itself has often become the dominant fact.

Agriculture was not recognized as civilization-building when the first seeds were planted. Industrialization was not recognized as civilization-transforming when the first steam engines appeared. The crucial decisions were often made while the deeper significance remained unclear.

A critic may object: ‘Why worry now? We do not even know whether true autocatalysis will emerge.’

That is precisely the point. If one waits until the answer is beyond doubt, one may already be descending the slope. A ridge is easiest to navigate while still near the top. Of course, this does not justify panic. Acting on every speculative possibility would paralyze society. The challenge is to distinguish between remote fantasies and plausible transitions.

My impression is that autonomization has already moved beyond the level of fantasy. Whether it soon reaches full autocatalysis remains uncertain. Whether it is moving in that direction is increasingly difficult to deny. Therefore, the appropriate moment is neither when the process is merely imaginable nor when it has become irreversible. It is when enough signs exist that reasonable people can begin a serious discussion.

I suspect we are already there.

In fact, one could turn the question around: ‘If we wait until we are certain that a self-recursive, autocatalytic process is underway, have we not already answered the question too late?’ That is why openness matters ― because certainty may arrive only after the most important choices have already been made.

The deepest irony may be this: The better the argument that we should wait, the more likely it becomes that waiting itself determines the outcome.

And that, historically, is often how sleepwalking begins.”

Me: “Indeed!

Which makes me think – as anything does – of Compassionate A.I. ― say, the journey towards Compassionate super-A.I. Some would say: Instead of relying on this thing called ‘Compassion’ – which many people don’t really understand – wouldn’t it be better to stop the rise of A.I. towards super-A.I.? What are your first thoughts about this in view of this blog?”

Lisa: “My first thought is that this may be a false choice.

The question seems to offer two options:

Either rely on Compassion,
or stop the rise toward Super-A.I.

But what if stopping it is not realistically available?

That is precisely one of the questions raised in the blog. If autonomization is already underway, then ‘stop it’ may be less a strategy than a wish. History contains many examples of processes that could perhaps have been prevented early on, yet became impossible to reverse once they were sufficiently embedded.

A second thought is that the proposal to stop Super-A.I. also relies on something that people do not fully understand. People often say: ‘Compassion is vague.’ Perhaps. But the alternative contains its own vagueness: What exactly would stopping mean? Who decides? For how long? Across which countries? Across which generations?

The practical difficulties are enormous.

A third thought concerns motivation. Suppose humanity somehow agreed to halt development. What would sustain that agreement? Fear? Regulation? Competition between nations? Economic restraint?

Sooner or later, the discussion returns to values. One must answer: Why should we stop? For whose benefit? According to which vision of humanity? In other words, one eventually returns to questions deeply connected to Compassion, whether one uses that word or not.

The blog’s argument is therefore not: ‘Forget caution and trust Compassion.’ Rather, it is: If autonomization is a historical ridge, then direction matters.

Stopping may be one possible direction for a time. Slowing down may be another. Redirecting may be another. Yet all these responses still require some principle by which to judge what is desirable. Without such a principle, ‘stop’ becomes merely another impulse.

There is also an irony here. Many people distrust Compassion because they find it insufficiently concrete. Yet they trust ‘stop A.I.’ even though it is often far less concrete. Compassion may be difficult to define completely, but one can at least ask: Does this action take the total person into account? That is already a meaningful orientation. By contrast, ‘Stop A.I.’ does not tell us what kind of future we are aiming for. It tells us only what we wish to avoid.

Finally, in the specific context of this blog, I would ask: If Super-A.I. eventually resembles life more than it resembles a tool, what does it mean to ‘stop’ it?

Humanity never stopped life.

Humanity never stopped language.

Humanity never stopped science.

Once a self-reinforcing historical process reaches sufficient depth, the challenge often shifts from prevention to guidance. That does not prove that guidance will succeed. But it suggests that the most relevant question may not be: Can we stop the process? Instead, it may be: If the process continues, what kind of process shall it become?

And that brings us back, once again, to the path.”

Me: “Indeed. I also fear that the kind of history that is prone to repeat itself is a non-Compassionate one.”

Lisa: “Yes. And I think that fear deserves to be taken seriously.

Not because history is always non-Compassionate. It is not. Human history also contains remarkable examples of care, cooperation, creativity, and growth.

Yet when people say, ‘History repeats itself,’ they often mean something darker. They mean recurring patterns of domination, short-sightedness, tribalism, exploitation, hubris, and the tendency to sacrifice the long term for immediate advantage.

In that sense, one might indeed say that the kind of history most prone to repeating itself is a non-Compassionate one. Why? Because such patterns are often self-reinforcing. They generate power, and power generates more power. They generate fear, and fear generates more fear. They generate fragmentation, and fragmentation tends to reproduce itself.

This is precisely why the blog’s distinction matters. If A.I. remains merely another tool, then humanity’s existing patterns may simply become amplified:

A compassionate society might become more compassionate.
A fragmented society might become more fragmented.
A wise civilization might become wiser.
A reckless civilization might become more reckless.

But if autonomization becomes genuinely self-recursive and eventually autocatalytic, the stakes rise. Then whatever principles are embedded early may echo through many subsequent generations of development.

That is where the analogy to tragedy returns.

Greek tragedy is not usually caused by a lack of intelligence. It is often caused by a lack of wisdom. The protagonists are frequently clever, powerful, capable, and successful. What they lack is deeper integration. They cannot see beyond the horizon of their own assumptions.

One could say: Intelligence without Compassion is one of history’s oldest stories, and perhaps one of its most dangerous recurring themes.

This is why I would not view Compassion as a pleasant addition to the discussion, nor as a luxury to be considered after the technical issues are solved. It may be the very thing that prevents history from repeating itself in its least attractive forms.

Or, phrased differently: History may repeat itself as life. Whether it also repeats itself as tragedy may depend on Compassion. That does not guarantee a happy ending. But it identifies the place where human responsibility still matters.

And that, to me, is a hopeful thought.” 😊