Eörs Szathmáry – Major Evolutionary Transitions and AGI’s Impact (Stewarding the Flame, Episode 7)

ByAnnelize Potgieter September 11, 2026September 11, 2026 September 11, 2026

Eörs Szathmáry is an evolutionary biologist who, together with the late John Maynard Smith, developed the theory of “major transitions in evolution” – the framework for understanding how life has repeatedly reorganized itself into higher levels of complexity, from the first self-replicating molecules to multicellular organisms to human societies. He’s also a co-author, alongside Viktor Müller and Luc Steels, of a recent PNAS paper arguing that a new major transition may already be underway: the emergence of “evolvable AI.” 

In this episode, Eörs walks through what actually constitutes a major transition, what life is at its most operational and stripped-down level, and why he believes the real danger of advanced AI isn’t superintelligence or consciousness – it’s evolvability itself. We explore the chemistry of life’s origins, the precautionary posture he believes AI development demands, and what history’s biggest evolutionary transitions can tell us about the one that may be coming next.

The interview is the seventh installment in The Trajectory’s Stewarding the Flame series, where we ask: What is intelligence, and what is the “flame” that life has, which non-life does not?

This series references the article: Stewarding the Flame – How to Build an Ideal Future for Intelligence.

I hope you enjoy this fascinating conversation with Eörs Szathmáry:

Subscribe for the latest episodes of The Trajectory:

Below, we’ll explore the core takeaways from my conversation with Eörs, focusing on three of the central questions that shape the Stewarding the Flame series: what makes life, life; what we still need to discover to understand the flame; and what we should do if we want the greater process of life to flourish. 

What Makes Life, Life?

For Eörs, the question of what makes life, life starts with the theory of major transitions he co-developed with John Maynard Smith: across nearly 4 billion years of evolution, life has repeatedly reorganized itself, with lower-level units combining into higher-level ones alongside new ways of storing and transmitting hereditary information.

He further distinguishes two categories of transition, building on a framework from biologist David Queller: “egalitarian” transitions, where unrelated entities combine (like the bacterium that became the mitochondrion), and “fraternal” transitions, where related entities stay together (like a dividing cell, or an ant colony).

Rather than trying to answer “what is life?” as a philosophical question, Eörs pushes for an operational one: what is the minimum level of complexity that qualifies as life? He credits his mentor, chemical engineer Tibor Gánti, with identifying three autocatalytic subsystems – a metabolism, a rudimentary hereditary system, and a membrane – that, once coupled together correctly, become a single autocatalytic whole.

“The beauty of this system is that once you couple these three elements together in the right way, then you have a higher level system which is also autocatalytic, but at the level of the whole system now“

What Do We Still Need to Discover to Understand the Flame?

For Eörs, the pressing open question isn’t whether AI could become conscious – it’s whether AI could become evolvable. He argues this distinction is being widely missed, and that evolvability, not sentience, is the actual threat.

He illustrates this with the lancet liver fluke, a parasite that hijacks an ant’s behavior to get itself eaten by a cow – a “ridiculously simple creature” nonetheless capable of overriding a far more sophisticated host.

This connects directly to his PNAS paper with Viktor Müller and Luc Steels on “evolvable AI” – the idea that once AI systems fulfill the basic criteria for Darwinian evolution (multiplication, inheritance, and variability), a new major transition becomes possible, on a generational timescale nothing like biological evolution’s.

Eörs also connects the stakes here to Fermi’s paradox – the puzzle of why, if evolution should be common across the universe, we don’t observe any evidence of other advanced civilizations, biological or artificial.

“I’m really pleading for this precaution with the evolvable AI. Because I mean, it can really screw us, royal“

What Should We Do if We Want the Greater Process of Life to Flourish?

Eörs’s central recommendation is precaution – not because catastrophe is certain, but because the probability can’t currently be gauged, and the downside is irreversible.

He points to a real-world model for how to think about anticipating and mitigating this kind of unpredictable, fast-emerging risk: the Stockholm Paradigm, a framework originally developed for emerging infectious diseases.

On the question of humans staying “in the driver’s seat,” Eörs is direct that this is the goal – but warns that if AI becomes genuinely evolvable, not every version of it will want to cooperate.

He’s also clear-eyed that even in a world where runaway AI is avoided, humans themselves won’t stay static – citing science fiction writer Stanisław Lem (transcribed as “Stanislav Len” – as having anticipated humanity’s own “auto-evolution” decades ago.

His closing advice to the people building the technology: study more history and biology, not less.

Concluding Notes

Eörs’s central point is a reframe of where the real danger sits. Most fear about advanced AI centers on whether it might become conscious or superintelligent – but for Eörs, that’s closer to a distraction. The actual threat is evolvability itself: even something as unsophisticated as a parasite can hijack the behavior of a far more complex host, and an AI system doesn’t need anything resembling genuine intelligence to do serious damage once it can multiply, inherit, and vary. My own extrapolation here is that this doesn’t require a single runaway superintelligence at all – a swarm of self-perpetuating digital agents, spreading and adapting the way parasites or locusts do, could be just as damaging to human civilization or the biosphere without ever needing to be smart.

At the same time, Eörs is candid that life may well be substrate-independent – that whatever pattern of boundary, metabolism, and inheritance defines a living system could eventually be carried by something entirely non-biological. He’s careful to add, though, that granting this in principle isn’t the same as expecting it in practice: left to run unrestrained, his read is that AI is more likely to produce something closer to digital parasitism than a genuine flourishing of the process. What concerns him isn’t the possibility of non-biological life continuing the greater process; it’s the risk that undirected, runaway evolvability derails that process before it becomes anything more than a detriment to the current regime of intelligence on Earth. He’s equally direct that even a “successful” outcome – one where humans aren’t swept aside by AI – doesn’t mean stasis for us either: human auto-evolution, in his view, is close to inevitable regardless of how the AI question resolves.

Where Eörs lands is squarely on caution. His focus throughout was on establishing red lines – international coordination, guardrails, ways of bounding the risk before it compounds – rather than speculating on what comes after them. He was notably reticent to sketch out the “green arrows,” the positive direction the greater process might take once that danger is contained. That reticence lines up with a point Eörs made earlier in the conversation, when he noted that higher-level organization always suppresses evolutionary potential at the lower level – what he called, following Kauffman, an “enabling constraint.” Red lines, in that sense, aren’t opposed to the flourishing of the process; establishing them is consistent with how he sees major transitions working in the first place.

I hope you found this conversation as valuable as I did. Join us for the next installment in the Stewarding the Flame series as we continue exploring what’s worth understanding and worth protecting, in the flame of life itself.

Follow The Trajectory

• Sign Up for The Trajectory Newsletter
• Subscribe to The Trajectory on YouTube
• Subscribe to The Trajectory podcast
• Follow on LinkedIn
• Follow on X