299 | Michael Wong on Information, Function, and the Origin of Life
What this covers
Sean Carroll interviews Michael Wong, a planetary scientist at Carnegie Institution, about whether evolution—biological and otherwise—follows a universal law. Wong proposes that functional information, a quantifiable measure of how rare configurations are that can achieve a specific function, tends to increase in systems composed of many interacting parts, mechanisms for generating new configurations, and selection pressure for persistence. The conversation moves across the origin of life, planetary chemistry, urban scaling, and potential futures for technological civilization, all unified by the claim that evolution is not unique to biology but a broader physical process.
The episode distinguishes life from mere chemical complexity by requiring a full feedback cycle: systems must acquire information from their environment and then use that information causally to persist. Autocatalysis and homeostasis alone do not suffice; a living system must do information processing too. Wong grounds this abstractly but tests it concretely—in how minerals accumulate new forms over Earth's history, how RNA aptamers gained binding function in laboratory evolution, and how Saturn's moon Titan might harbor an atmospheric autocatalytic cycle sustained by surface chemistry. He addresses the RNA world versus metabolism-first divide in origin-of-life research, arguing both camps contain hidden assumptions that a clearer framework for function could dissolve. Notably, Wong extends the framework to cities and civilizations, drawing on Geoffrey West's work on super-linear urban scaling to explore whether human societies might face singularities avoidable only through accelerating innovation, or whether they might instead achieve "homeostatic awakening"—a conscious phase transition toward sustainability. The discussion remains anchored to what can be tested and what remains speculative, distinguishing between established observations in mineralogy and more tentative applications to social systems.
Wong argues that evolution—biological and non-biological alike—can be unified under a proposed law of increasing functional information, where systems composed of many interacting components, mechanisms for generating configurations, and selection for persistence-enhancing functions tend to accumulate functional information over time.
- Functional information (negative log base 2 of the fraction of configurations that achieve a function) is a quantifiable metric that increases in evolving systems
- What distinguishes life is a full cycle of acquiring environmental information and using it causally for persistence, beyond mere autocatalysis or homeostasis
- The framework applies across minerals, planetary atmospheres like Titan, biology, and potentially societies, all bounded by free energy availability
Functional information measures how constrained configuration space becomes when selection pressures act on systems.
- Functional information, introduced by Jack Szostak in 2003 and elaborated by Bob Hazen and colleagues in 2007, is the negative log base two of the fraction of a system's configurations that can achieve a specific function, measured in bits; the fewer configurations that can perform the function, the higher the functional information.
“what the functional information is, it's basically a measure of the fraction of configurations of a system that can perform or achieve a specific function. And so it's, the math is, it's the negative log base two of the fraction of configuration. So it's measured units of bits”
- Functional information has been observed to increase in test cases: RNA aptamers evolved on a lab bench by selecting and propagating the best binders show rising functional information, and minerals show increasing functional information over deep time—from the first minerals formed in stellar atmospheres through protoplanetary discs, the Hadean crust, plate tectonics, and ultimately the third of Earth's mineral diversity that is biologically mediated.
“you can evolve these things on the lab bench and select for the ones that can bind the best and then propagate them to the next tube... And so you see the functional information increase there”
- Rather than Kolmogorov complexity being the quantity that increases over time, the framework proposes functional information as the relevant measure, since a simpler system can function and persist better than a wildly complex one—so what matters is not traditional complexity but how systems interact and co-create one another's persistence.
“we're trying to maybe move away from the idea of complexity in a Kolmogorov complexity sense as sort of the thing that is increasing over time, but rather it's this functional information”
- Functional information is maximized when only one configuration achieves a function, but real function landscapes have some width—a protein or genetic strand tolerates some mutation before failing—so as evolving systems undergo combinatorial expansion while selection pressures increase, the sub-volume of functional configuration space shrinks and the system's functional information rises.
“you might imagine this function landscape that looks almost like a delta function, but there is a little bit of width to it”
Life is fundamentally a system doing autocatalysis, homeostasis, and information processing together in causal feedback loops.
- A function is any process emerging in a system that has causal efficacy over its continued persistence or viability; the core functions—autocatalysis, homeostasis, and information processing—directly contribute to a dynamic system's persistence, and having all of them constitutes a living system regardless of whether it resembles Earth life.
“a function is for us basically any process that might emerge in a system that has some kind of causal efficacy over its continued persistence or viability”
- What separates life from non-living systems is that life completes a full cycle in which information acquired from the environment is then used causally for the system's continued persistence—doing autocatalysis, homeostatic feedback, and information processing together—whereas minerals record environmental information in their structure but that information does not feed back into their persistence.
“in life, there is this full circle, this full cycle of the way that information is acquired from the environment and then utilized for the persistence of that dynamic system”
- In a hot box of atoms the possibility space is defined only by particle positions and momenta, but lowering the temperature so atoms can bind into molecules technically decreases the space of atomic positions while opening a richer, information-bearing possibility space of molecular configurations in which functions like autocatalysis can emerge that are impossible among free atoms.
“when you are able to make these fundamentally new structures, which are molecules, you may want to shift yourself into a completely different possibility space, maybe a different state space to describe the interactions of those molecules”
- When an autocatalytic cycle (A makes B, B makes C, C remakes A) is selected by the environment for dynamic persistence, the abundances of molecules shift away from what static persistence would predict—a short-lived molecule can become highly abundant because it is constantly recreated to complete the cycle—so detecting anomalous organic molecule distributions could be a biosignature revealing selection for function rather than static persistence.
“the abundances of molecules in your environment changes from the pure abundances that they would've had if they were just selected for their static persistence”
A law of increasing functional information applies to all complex systems with diverse components, configuration mechanisms, and persistence-based selection.
- The proposed law does not contradict the second law of thermodynamics but layers additional principles to explain why a universe driven from low entropy toward heat death produces complex functional systems at all, since one can imagine a universe that goes from low entropy to heat death producing none of these complex systems.
“what we're trying to do, I think, is just layer on principles that can help us explain the rise of complex functional systems in a universe that is driven fundamentally by the second law of thermodynamics”
- There may be a law-like description of complex evolving systems that expands evolutionary theory beyond Darwinian evolution, so that evolutionary processes—complexification, diversification, patterning—can be seen at play in isotopic, mineralogical, planetary, and biological evolution, and potentially in symbolic systems like language, technology, and culture.
“this paper which hypothesizes that there might be some kind of law-like description of complex evolving systems that expands evolutionary theory beyond just Darwinian evolution”
- The central proposal is a law of increasing functional information that applies to all systems possessing diverse interacting components, mechanisms for generating configurations, and selection for persistence-enhancing functions; it is a bounded, macroscopic law that emerges from large collections of interacting entities and that, like the second law, may not be derivable from first principles but must be tested incessantly across many systems.
“that's our proposal, that there is a law of increasing functional information, that applies to, again, all kinds of disparate physical chemical systems, including, but not limited to biology”
- Any complex evolving system has three attributes: (1) it is composed of many different interacting components, (2) there exist mechanisms to generate many different configurations of those components, and (3) there is selection for function that prunes the possibility space to a few select configurations.
“for Bob the attributes of any complex evolving system, number one, they have to be composed of many different interacting components, two, there exist mechanisms to generate many different configurations of those components, and three, there is selection for function that essentially prunes or widows the possibility space to a few select configurations”
Only one Earth life example exists, making astrobiology and information-theoretic frameworks essential for understanding life universally.
- Astrobiology is a field that isn't really a field yet—a pre-paradigmatic science that constantly moves in unexpected directions, partly driven by new observational capabilities like JWST broadening our view of exoplanets.
“Astrobiology is this field that isn't really a field yet, it's a pre-paradigmatic science. And so it's constantly moving in kind of unexpected directions.”
- Information may be the right 'Goldilocks' level of abstraction for understanding life—abstract enough to move beyond the specifics of Earth life but not so abstract that literally everything gets categorized as life—which motivates casting evolution as a broader process spanning living and non-living systems.
“Bob and I both talked about information as sort of that thing that might be the right sort of Goldilocks level of abstraction. So we want to abstract away from the specifics of life on earth, but we don't wanna add start too far that literally everything might accidentally be categorized as life.”
- The metabolism-first and genetics-first origin of life scenarios could both be right, both be wrong, or the correct geochemical system may not yet have been imagined, so researchers must remain very open-minded.
“they could both be right. There's no reason why they couldn't both be right. And they could also both be wrong, maybe we just haven't thought of the right geochemical system yet.”
- The origin of life field is split between metabolism-first and genetics/information-first camps, and the divide is driven so deeply because each paradigm favors a different prebiotic environment: RNA-first favors surface tide pools with wet-dry cycles, while metabolism-first favors deep submarine alkaline hydrothermal vents that provide redox and pH/proton gradients that life still harnesses today.
“the reason why it's driven such a hard wedge in the field is because each paradigm or expectation for what the first living system looked like results in different prebiotic environments that are favored”
- Traditional origin of life scenarios have a good deal of magic sprinkled into them—the RNA-world camp assumes Darwinian evolution will somehow build the required metabolic engine, while the hydrothermal camp assumes RNA will somehow eventually be an output—so each contains a 'miracle occurs' step that the big foundational questions about life must address.
“a lot of these traditional origin of life scenarios have a good deal of magic sprinkled into them”
- Because we have only a single example of life on Earth—where all organisms, however diverse, are related and share a universal biochemistry—it is hard to generalize about what life fundamentally is, and we won't truly know until we find more examples, which is why thinking about universal properties of life is valuable for guiding the search.
“with a single example of life here on Earth, it's hard to do that kind of generalization from our so-called N equals one”
- Titan's atmosphere is mostly nitrogen with 3-5% methane, yet methane is irreversibly converted to higher hydrocarbons by photochemistry on a ~30-million-year timescale—far shorter than Titan's 4.5-billion-year age—so its persistence is a conundrum; an intriguing abiotic hypothesis is that surface organic mineralogy reconstitutes methane by combining atmospheric hydrogen with surface carbon, making the whole atmosphere-plus-surface system an autocatalytic cycle selected for dynamic persistence.
“it is irreversibly converted to higher order hydrocarbons through photochemistry... And the chemical timescale for this process is roughly 30 million years, so vastly shorter than the 4.5 billion years that Titan has been around”
Defining function contextually by what enhances system persistence avoids teleology while remaining objective and measurable.
- Whether a feature counts as a function is deeply contextual—being slimy doesn't matter most of the time, but for a snail generating mucus to move toward its next meal it becomes relevant to ongoing persistence—and the framework's challenge is to define functions contextually while still tying them objectively to system persistence.
“most of the time being slimy doesn't matter, but if you're a snail generating mucus so that you can move this ancillary function of movement so that you can go and get your next meal does become very relevant for your ongoing persistence”
- Because persistence-enhancing functions create pressure to discover new functions, novelty generation itself can be selected for—especially in environments that are not just spatially heterogeneous but also temporally changing, like the early and modern Earth—since a currently functional configuration may become less functional as the function landscape shifts, forcing systems to keep exploring parameter space to maintain viability.
“novelty generation might itself be selected for, especially when you have an environment that is very complex, not just spatially heterogeneous, but temporarily changing as well”
- The relevant kinds of functions are those whose removal would impair the system's ability to persist—if you removed the function, the system would not persist as well—which lets persistence-serving functions be discussed without overly teleological language about purpose.
“is it safe, or is it fair to say that specifically the kinds of functions we have in mind are those that if you removed them, the system would not be able to persist as well?”
- Functional information is expected to peak and then decline like a coffee cup's complexity that rises as milk forms tendrils then fades to uniformity, because the framework requires free energy gradients to generate configurations, and deep into the universe's history that low-entropy source runs out so the second criterion (configuration generation) is no longer satisfied.
“We would expect the functional information of systems in the universe to peak at some point and then return down”
Quantifying functional information remains tractable for mineralogy and simple systems but combinatorially explosive for biological systems.
- Functional information is difficult to quantify for complex evolving systems because you need the entire combinatorial configuration space and the functionality of every configuration; it is tractable for mineralogy and artificial life systems like Avida but the combinatorics explode wildly for biology, though shortcuts may exist—just as we cannot exactly solve the solar system's gravitational potential yet still send spacecraft to Jupiter.
“the functional information is difficult to quantify for many complex evolving systems because you need to know the entire sort of like combinatorial parameter space, and you also need to know the functionality of all the different configurations”
Complex systems from cities to civilizations follow singularity trajectories escapable only through innovation or homeostatic phase transitions.
- A globally connected planetary civilization may head toward a singularity it can skirt repeatedly through innovation, but eventually the innovation timescale shrinks so much that a fluctuation knocks it into 'asymptotic burnout' from overusing energy; alternatively, on gaining planetary-level sapience, a civilization may recognize the burnout trajectory and undergo a 'homeostatic awakening'—a phase transition to a sustainable mode prioritizing societal and planetary well-being.
“eventually that timescale of innovation becomes so small that some kind of internal or external fluctuation will knock you off and you will then enter what we call a burnout, which is that you use too much energy and you're no longer able to be viable”
- The symbiotic relationship between science and science fiction may be an autocatalytic cycle in culture that is selected for because it inspires scientists to discover new truths that help society persist.
“maybe that symbiotic relationship is an autocatalytic cycle in our culture that is selected for because it inspires scientists to go and do cool things and discover new truths about the universe that help us as a society persist”
- It may be naive to expect alien civilizations to follow an expansionistic, colonialistic, extractive trajectory like becoming galaxy-colonizing Type III civilizations; instead, civilizations that 'made it' may have learned to live in harmony with their planetary environment, so SETI should look for signs of planetary wisdom rather than just outward expansion into space.
“Maybe what we should expect is civilizations learning to live more in harmony with their planetary environment and look for signs of that kind of planetary wisdom, so to speak, rather than just the continued expansion into outer space”
- Drawing on Geoffrey West's work, cities act as social reactors whose energy consumption scales super-linearly, putting them on trajectories toward singularities where they would use infinite energy in finite time, which they only escape by innovating at an ever-faster pace.
“cities as social reactors are super linear in scaling, especially their energy consumption. And that the cities tend to go on these trajectories towards some kind of singularity where they will utilize infinite amount of energy in a finite amount of time. And how we skirt our way past those terrible singularities by having innovations, but these innovations must come at a faster and faster and faster pace”