Life as Thermodynamic Evidence of Algorithmic Structure in Nature (2012) | Hacker News Reader

Life as Thermodynamic Evidence of Algorithmic Structure in Nature (2012)

mdpi.com

75 pointsjkhdigital39 comments

Many regard MDPI as a predatory publisher. That's not a universal view and there is clearly a substantial amount of valid research that appears in MDPI journals. But they seem to have a bias towards accepting papers with cursory review and collection of article processing fees. For example, they ask that peer-reviewers submit reviews in 7 days or less, which is something few experts are going to have time for.

Then there is the matter of MDPI creating deceptive journal titles like _Cells_ which surely is often confused with the incredibly prestigious journal _Cell_.

Anyway, caveat lector.

Isn't everything[1] thermodynamic evidence of algorithmic structure of nature?

When you have intial conditions (big bang, fundamental constants) and dynamical laws (physics), everything is an algorithmic construction to some degree, at least according to physics as we know it (i.e. who knows what's truly fundamental).

[1] Not eveything is governmed by thermodynamics like initial conditions of the big bang, fundamental constants being the values they are.

Well, every closed, Markovian system is governed by thermodynamics in a rather specific mathematical sense [0]; but yes I agree that initial conditions and parameters (such as constants) are not, but the dynamics themselves are.

Unless our assumption about Markovian dynamics is wrong, in which case it’s not even clear we can make useful predictions from such a theory.

———

[0] see, e.g., https://guille.site/second-law-markov.html or Cover’s own textbook.

This really screams "BS" to me, but I should temper that perception by the fact that I only read the abstract (and so this comment itself would, ironically, qualify for at least one BS flag.)

The most suspicious bit is the claim that organisms can't survive in completely unpredictable environments. That statement is both obviously true or obviously false depending on the variance of the unpredictable random variables... A totally unpredictable temperature between 25 and 26C would not be much of an issue, a totally unpredictable temperature between 0K and 1,000K would be. In fact it would be so much of an issue that its unpredictability wouldn't contribute very much to its difficulty!

There are pretty compelling reasons to believe that life/physics are strongly tied to computation. See eg the slime molds experiments for some of the more visually impressive cases of that.

The question of how to turn these loose inklings into actionable, falsifiable, reproducible science- rather than Aristotelian armchairing - is a hard one.

When I see those inklings, I uncontrollably see a cellular automata operating on a non grid (or a grid with very small cell size) space. Whether that automata is our or its own implementation is the real question. My hunch is, regardless of the underlying cause, our experiment proves neither.

If there are no regularities whatsoever in the environment, how could an organism possibly form and survive?

"No regularities" is one of those things that sounds like it's saying something, but isn't. What does it mean, the laws of physics change every 5 minutes?

It means that the environment at time t is totally unpredictable given the environment at time t-1. If you imagine that the universe at time t is described as a string of random bits, and the universe at time t-1 is also described as a string of random bits, it means that you can't predict any of the bits at t given the bits at t-1. That's what it would mean for nature to be "algorithmically random" in the sense of the paper.

Well the organism IS part of the environment. So it wouldn't be able to encode anything if all it consists of is noise..remember.. we aren't separate from physics...we embody it.

Life is a noise ratchet.

This was also my question.

Obviously there should be some regularity for an organism to form.

But lets assume that the environment for an organism is indistinguishable from the random noise.

Then the only strategy the organism can form is to randomly sample the environment for food.

If there is enough food in the environment then the organism survives, if not then it does not survive.

Therefor it is disproven that an organisms can survive only in predictable environments.

It could be probably displayed that any more complex strategy could be not developed though.

If the environment is noisy at a low-enough level then it won't have enough stability for there to be consistently-metabolizable molecules around at all.

Except that organisms are generally not ergodic systems, so they are subject to gambler’s ruin. Even with “enough” food the probability of non-survival is non-zero and asymptotically increasing over time.

This is what they are trying to argue: "Information processing in organisms should be understood as the process whereby the organism compares its knowledge about the world with the observable state of the world at a given time; in other words, the process whereby an organism weighs possible future outcomes against its present condition. While the larger the number and the more accurate the processed observables from the environment, the better the predictions and decisions, organisms cannot spend more energy than the total return received from information.".

But that's obviously false.

If the environment is rich in sources of energy, such that expending energy in never really an issue, it makes no difference whether the total return from "processing observables" is more, less or equal to the energy spent doing it.

The argument meant in the paper is that organisms encode information about their environment in order to survive. The encoding operation costs energy and generates entropy. Organisms use the encoded information to try to gain (or not lose) energy. Only in environments where organisms can use the information storage cycle to generate net positive energy (and maintain their net negative entropy relative to their environment) can life exist.

In environments which are "too unpredictable", life cannot encode sufficient relevant information for that cycle to be net positive, and therefore life cannot exist. Since life exists, nature is therefore not "too unpredictable".

That's the argument anyways.

For your objection, if we actually lived in a universe where everything was fully unpredictable between 25 and 26 degrees and temperature was the only important variable, then that would be every bit of problematic as between 0 and 1000K. Since life needs energy gradients and quickly destroys existing low entropy states, life always relies on finding temperature boundaries, regardless of how small it is to survive.

That's because continuity is an important part of predictability, right? When temperatures change, they do so continuously in whichever direction. A truly unpredictable environment would see temperatures change effectively instantaneously between points. Even for small differences in temperature, that's anathema for any system. Even simple structures would be unlikely to form as they'd immediately be ripped apart by the sudden change in the state of energy distribution, right?

Hence the ice age..extinction. Organisms only encode for certain levels of environments based on their past history. Some mutations get thrown in, providing select individuals with special doomsday survival genes.

Yeah, I was thinking of making a climate change comparison. Maybe civilization is at a scale that wouldn't allow its own survival should the environment change "instantaneously" (50 years, for a global civ that's endured essentially continuously for 6,000-10,000 years, is instantaneous).

This ignores epigenetics and proteomics.

There's a naively fashionable idea that DNA is basically the same as a Turing tape.

It isn't. It may be true that biological systems can be understood in terms of information theory, but the information is in the entire ecosystem - including the sum total of all individuals and species and their previous and current state.

E.g. on Earth, the entire planetary ecosystem eventually evolved a species with the ability to understand quantum theory - which happened to be a useful adaption, at least for a while.

But you're not going to find an explicit formalism for Schrodinger's Equation in human DNA no matter how hard you look for it.

Sounds like DNA is like a Turing tape from which are produced cells capable of processing information.

His point was that DNA alone is not sufficient to to represent the role of Turing tape, because the resulting cells also take tons of input from their environment in the forms of things like epigenetics.

In other words, Turing machine as many other algorithms can take parameters/arguments.

A classic turning machine does not take parameters/arguments except those written on the tape beforehand. The tape and table of instructions describe the totality of the computational universe.

I don't think anybody thinks that DNA is literally classic turing machine with infinite tape..

Well no, but the problem is it's a misleading analogy, misleading enough that it's inadequate for any amount of formalism.

It can be trivially mitigated, you just say that input signals through the time is located in some area of the tape, and from that you can apply all your formalisms.

Pretty sure you’ve attacked a straw man here. The article does not make the claim that DNA is the one and only source of information used in biological computation. It does, perhaps, imply that DNA is the most important source of biological memory.

Yeah. Epigenetics is more like a small twist in the Fundamental Dogma of Biology. And "proteomics" doesn't mean much of anything--all proteins derive from DNA.

I don't know, are go-to statements and code reuse a small twist in the dogma of software engineering?

Is your Python-level code directly interacting with CPU operations and memory bit flipping a small twist?

Hi-C 3D DNA structure experiments as well as small RNA fragments interfering with transcriptions and protein function are not a small twist in my opinion - it's like finding out there's an entire extra hierarchy of layers to the system.

Those are great and non-obvious discoveries, but they don't particularly change the fact that (virtually?) all genetic information is carried via DNA. (...and possibly similar mechanisms that don't particular change the overall story)

Well, no, it doesn't change the fact that DNA contains (most of) the information. What it means is that even if I give you all of the bases on all of the chromosomes, if you don't know how these strings of beads will fold in a real cell (in the presence of all kinds of other things, like proteins and metal ions, which in a fertilized egg will come from the mother - a bootstrapping problem), it does not get you very far. It might not be unlike getting all the lines of code in a codebase, but not getting the ordering, only a bag of lines of code.

The central dogma, though still taught in schools, is a strongly oversimplified model of how information moves in a cell. In fact, there is so much regulation going on on all levels that proteins, in some sense, aren‘t even the main „product“ of DNA. (Most DNA is not proteinogenic, but rather has regulatory functions.) Also, proteins are extensively modified after transcription and even after translation, so yes, „proteomics“ very much does have meaning.

https://en.wikipedia.org/wiki/Central_dogma_of_molecular_bio...

My point was that all of that regulatory behavior is ultimately captured in the organism's DNA.

OP was claiming that you can't even really understand any of this unless you understand "the entire ecosystem - including the sum total of all individuals and species [on earth] and their previous and current state".

That's technically true in some sense, but most scientists would roll their eyes at the wildly expansive nature of this claim.

If you fully understood the information in an organism's DNA, plus the DNA of their close symbionts, that'd be a rather complete picture.

Isn’t “life” only significant because we ascribe significance to it. It’s like saying if a a random number generator spit out 7777777 that it must be rigged.

It’s hard to find a definition of “significant” (or rather “meaningful”) that is not “whatever humans ascribe meaning to” and is also disprovable.

This blog explores the topic in detail, and offers a remarkably balanced perspective: https://meaningness.com/preview-eternalism-and-nihilism

ORT (Only Read Title) and abstract so far but it reminds me:

Cf. "Life as Evolving Software, Greg Chaitin at PPGC UFRGS" https://www.youtube.com/watch?v=RlYS_GiAnK8

> Few people remember Turing's work on pattern formation in biology (morphogenesis), but Turing's famous 1936 paper On Computable Numbers exerted an immense influence on the birth of molecular biology indirectly, through the work of John von Neumann on self-reproducing automata, which influenced Sydney Brenner who in turn influenced Francis Crick, the Crick of Watson and Crick, the discoverers of the molecular structure of DNA. Furthermore, von Neumann's application of Turing's ideas to biology is beautifully supported by recent work on evo-devo (evolutionary developmental biology). The crucial idea: DNA is multi-billion year old software, but we could not recognize it as such before Turing's 1936 paper, which according to von Neumann creates the idea of computer hardware and software.

- - - -

Also "Algorithmically probable mutations reproduce aspects of evolution such as convergence rate, genetic memory, and modularity"

https://arxiv.org/abs/1709.00268v8

> Natural selection explains how life has evolved over millions of years from more primitive forms. The speed at which this happens, however, has sometimes defied formal explanations when based on random (uniformly distributed) mutations. Here we investigate the application of a simplicity bias based on a natural but algorithmic distribution of mutations (no recombination) in various examples, particularly binary matrices in order to compare evolutionary convergence rates. Results both on synthetic and on small biological examples indicate an accelerated rate when mutations are not statistical uniform but \textit{algorithmic uniform}. We show that algorithmic distributions can evolve modularity and genetic memory by preservation of structures when they first occur sometimes leading to an accelerated production of diversity but also population extinctions, possibly explaining naturally occurring phenomena such as diversity explosions (e.g. the Cambrian) and massive extinctions (e.g. the End Triassic) whose causes are currently a cause for debate. The natural approach introduced here appears to be a better approximation to biological evolution than models based exclusively upon random uniform mutations, and it also approaches a formal version of open-ended evolution based on previous formal results. These results validate some suggestions in the direction that computation may be an equally important driver of evolution. We also show that inducing the method on problems of optimization, such as genetic algorithms, has the potential to accelerate convergence of artificial evolutionary algorithms.

(quoting myself from ~2 years ago, https://news.ycombinator.com/item?id=18571878 hope nobody minds.)

I think a similar observation can be made by watching crystals grow.

Reply on news.ycombinator.com