The baffling intelligence of a single cell (2024) | Hacker News Reader

The baffling intelligence of a single cell (2024)

jsomers.net

336 pointsjsomers95 comments

I am always amazed how these complicated systems evolved through random variations.

There must be plenty of “versions” which don’t do anything but generations later are crucial to a function.

How does evolution not get stuck in a local optimum?

> How does evolution not get stuck in a local optimum?

Neutral drift and very large population sizes. Anything that preserves diversity in the population long enough to find the right basin. Most artificial evolution systems cannot support population sizes large enough to preserve diversity long enough to locate any solutions that matter.

Right now, evolution is running a live experiment with something like 10^30 candidates in one population alone. The computational requirements for this are effectively zero. The environment is the computer.

I thinnk that is how evolution works. Keep dying so the best genes reproduce for the environment, and if it stays the same you stay too, until the environment changes, then you must go with it. Rinse and repeat.

I assume it does get stuck, and creatures go extinct as their environment changes around them and evolutionary mechanisms don't by chance produce a change that enables survival either quickly enough, or because the required changes are not possible from the current state of the organism.

boggle randomization

Considering our eyes are wired inside-out (the nerve cables are on the same side light comes in, so they have to go through a hole in our eye, the blind-spot), I'd say that natural selection gets stuck in local optimums all the time. Cephalopods like octopuses have very similar eyes to us, but they evolved completely separately and their nerves are wired in "the right way".

So in our case the starting conditions just lined up the way it did and no single step (or series of steps) easily flipped our wiring, so we just rolled with it (and we get by quite well).

Random mutations I think

> How does evolution not get stuck in a local optimum?

Could be wrong, but my understanding is that it does? Plenty of things in nature are not really evolving much at all (as in, variations are not selected) even over longer periods of time. What can happen to it is getting removed by something that evolved from somewhere else (taking the evolutionary niche, or making it unsurvivable)

Gene duplication is actually quite common and contributes quite a bit to creating new "versions" of things.

Vertebrates for example had a at least 2 whole genome duplication events in the past giving us up to 4 copies of the same family of many genes. This enables mutations to alter one copy of a crucial gene to "invent" new functionality without killing the organism.

The most common version of this is probably duplication of G protein linked receptors for odor and taste. Mice and dogs have far more of these than we do.

Evolution does get stuck in a local optimum quite a bit. Think of how many vertebrates are now limited to 4 limbs compared to many insect or mollusk species.

Instead of having dexterous forelimbs around the mouth like crabs, elephants make due with prehensile noses.

Similarly, to get to running speeds while having opposable thumbs humans evolved to stand upright on the hindlimbs instead of simply becoming 6 limbed centaur like creatures. Humans can't outrun any major predators at short distances.

In some ways, a single cell is more like an entire artificial neural network.

Real neurons have some amazing properties that individual neurons in ANNs do not. It's still unclear how we can exploit those properties to improve AI.

Aren’t there efforts to simulate neurons? (Aka artificial brains)

yes, but I think they are egregiously wasteful of resources if your goal is to recapitulate the behavior of real neurons. Instead you can probably build a far simpler model that produces all the behaviors you want without directly simulating a neuron.

Indeed. And, one key insight here is, "artificial neuron" and "real neuron" don't have to map 1:1. You can use N artificial neurons to capture whole (relevant) behavior of a single real neuron, and still reproduce the larger real network, just with N times more nodes.

I think you might also want spiking (IE, the neurons are running semi-continuously and outputting pulse chains), and feedback (neurons form a cyclic graph), reweighting (neuron activation threshold changes), and rewiring (neuron topology changes).

in short, a continuously running graph that can rewire (instead of a feedforward network) with message passing; it would run training and inference simultaneously (inference would continuously retrain the network). Such a system running on the largest GPU clusters would be an interesting experiment in itself, although I don't think it would match the direct raw performance of today's AI models, and may never actually be better than just using feedforward networks trained by backpropagation, with access to external signals.

A single cell is a living and breathing (respirating) ecosystem.

DNA, the molecule and associated enzymes, is a computer in and of itself. Every dimension - from its hydration to dimensional folding - is a computational manifold.

The cell itself is a system of distributed supercomputers, a machine assembly factory, a dizzying highway and information network, and a universe. Replete with defenses, power factories, intricate and robust signal processing, and self destruct mechanisms.

DNA is not a computer. It's a passive storage mechanism. When you add the appropriate enzymes though it is turing complete.

That's a fair distinction, though calling DNA completely passive feels a bit too strong. How it's folded and which proteins can access it also affect what happens in the cell

That's because like in abstract algorithmic descriptions, there is no distinction between code and data, in the physical world, there is no distinction between interpreted code and interpreter, it's just a chemical reaction between the two.

(That's true for computers (on semiconductor basis) as well, is the behaviour of an addition in the ALU or in the bit pattern to select the instruction? It's both, the ALU alone would exhibit all kind of operations (including unintended/not-meaningful ones), the bit pattern alone could trigger any kind of operation depending on the circuit it occurs in. But if both happen to occur at the same point in spacetime, the physical laws apply and the bit pattern emerges, that humans would call the addition.)

The DNA doesn't fold itself. There is an entire collection of proteins and enzymes that maintain it (example: histones and proteins that adjust histones). Some DNA sequences do affect that folding, but not in an "active" way.

There are a very small number of DNA enzymes produced in labs, but there are no endogenous DNA enzymes (that anybody has found so far). Instead, it seems like life is much more based on RNA enzymes and proteins (the core of the ribosome is an RNA machine, decorated by proteins that enhance its speed and reliability).

Cells are amazing. Even mitochondria are amazing, BUT let's not get carried away with the sci-fi nonsense. You're projecting computer science on to biology because "we're all programmers here" and to the HN crowd (myself included) we like to see the world through comp-sci metaphors. But that's just hyperbolic and inaccurate.

I think the other article from Js Sommers on the awe inspiring complexity of Biology is also worth a read through

https://jsomers.net/i-should-have-loved-biology/

It's funny because the poster he shows (the Roche Biochemical Pathways)- I had that on my wall in college in 1992 and by the end of my biochemistry class, I could explain nearly all of it. The elucidation of the metabolic cycles and photosynthesis stand as one of the world's most impressive achievements (yet, with all that known, engineering biology remains devilishly complex).

Fascinating.

I totally see the CheY as an intermediary layer that learned how the changes in attractant must affect mechanical activation.

I would agree, but I don't think they're that much more interesting than the neurons of many "biologically plausible"-type neuronal nets, which allow for inhibition/exhibition, etc, but are obviously still limited in variation between individuals.

Where I think the really interesting parts missing are those in real neural circuits do, especially since the connections are more heterogenous, and the dynamics are much more fluid and time-dependent, whereas typical at-scale computer neurons obviously operate at fixed time steps, and don't exactly have bursters etc.

I wrote a very large scale distributed simulations for real neurons (straight from imaging datasets) when I was working in a TMS lab, and a longtime side project I always wanted to get going was some kind of pipeline for spitting out neuromechanical/cable-theoretic RTL tailored to specific morphologies and networking them according to their connectomes. I don't even think I would see any performance improvements, it just feels fun.

I think the timing aspect is especially interesting here. We tend to focus on making individual artificial neurons more sophisticated, but a lot of the interesting behavior might come from how the whole circuit evolves over time

Absolutely! Just look at the gastric mill circuits that have been studied for decades. Single-digit neuron counts and still has incredibly complex functionality.

Cell membranes are fascinating in regards to your ANN nod.

As I understand it, AI 'neurons' have nothing to do with IRL neurons other than a passing nod to the name loosely based on the idea that 'they are connected to each other'. A neural net is nothing other than an (very) approximate model loosely based on our understanding of IRL neurons. There are many many ways a real neuron could be resonding to it's environment to do what it does, whereas our artificial 'neurons' are just a mathematical function.

I would image then, trying to imbue our neural networks with the properties of real neurons is a futile ambition?

Not so much futile (well, perhaps it is considering there’s much we still don’t know about biological neurons) as slow - simulating all that in software won’t be fast.

I suspect the first ANN researchers called them neurons to get more funding.

In essence an ANN is just a complicated mathematical function that you can tweak to fit the training set. Originally the activation function of a neuron in an ANN resembled real neurons, but later it was found that other, very different, functions are much better, like RLE.

I agree- neural networks were called like that to be sexy. It's certainly what attracted me to them in the 1980s (because I was intersted in both computers and biology). I got in a lot of arguments in the early 2000s before I understood just how fundamentally different neural networks are from brains.

I asked ChatGPT to estimate memory requirements of a brain simulation, with neurons from a paper linked in OP post ("Single cortical neurons as deep artificial neural networks"). This paper uses neuron models that are supposed to simulate real ones. The answer was 8-50 PB, with 10^7 weights shared between neurons, or 3.4 EB if each neuron has its own 10^7 weights. So it's not out of the realm of possibility.

On the other hand it shows how utterly tiny and primitive state-of-the-art LLMs are compared to real-world brains. By the way our brain only runs on on a max clock speed of around 100-1000 Hz.

Yet LLMs can achive miraculous results with so little. It really begs the question, where is the limit? How smart would a artificial human brain be if it would clock in gigahertz and had the somehow orders of magnitudes more efficient reasoning and detail recall of LLMs?

I feel like it's also that these single cells are much more like deterministic computing machines (sure there are error bars in chemical processes, but that exists in semiconductors too), unlike the stochastic process that we know use to bruteforce behaviour.

here's more captivating illustrations of cells and cellular mechanisms from David Goodsell (some of his pictures are used in the OP's article). i simply love his style and sense of color.

https://ccsb.scripps.edu/goodsell

i've read and can highly recommend his book The Machinery of Life for a patient and visual explanation of these baffling machines!

I would like yo argue thatbthis is not intelligence. A system regulated with a nutrient sensor that increases concentration of said nutrient and changes the turning direction of the flagella when the concetration lower bellow a threshold is not intelligent. It's a fully automatic mechanism, which is the opposite of what it is understood by itelligence.

It is cool, but not intelligence.

What is understood by intelligence?

Isn’t intelligence a fully automatic mechanism? You go down that rabbit hole enough and I always find it hard to find the differences. Same thing with LLMs, people say it’s just a word predictor. But that’s how I wrote this comment and how I think.

https://maxleiter.com/blog/weights

Love this

I think there's a point as you increase the comolexity of the "information processing" system where you get emergent intelligence.

A mechanism that makes a rotor go one way or the other as a function of the time derivative of a variable would not qualify as such.

Now, if you have a trillion such mecanisms atuned and feeding one anothers inputs in a network... Well you just got yourself a brain and then we can start talking about intelligence.

I would argue you don't know what intelligence is.

Are you suggesting that the commenter is less intelligent than a single cell?! How dare you, sir, how dare you!

It's a very smart cell, all be said.

Then do. Enlighten us and save the ad hominems.

He didn't say, that he knows what intelligence is.

What the...

It is well known that the only definition of intelligence is the ability to solve Raven's progressive matrices. /s

What neurons do is also quite automatic.

What are we if not a system of sensors directing our muscles towards nutrients and reproduction

Nah. Lets be honest - we don’t know.

Not much more, tbh, but the way we act to maximize our intakes are so much more sophisticated than a simple control regulator that I think we deserve to be called intelligent.

What is meant here is (of course) INTELLIGENTLY DESIGNED.

(...and since HN is mostly of evolutionist Faith, they can't stand that term...)

The irony of misspelling intelligence on that comment. And yes you can argue this is not "intelligence" but then you can extend the same argument to human beings (assuming that is what you understand as itelligence) and say they are not intelligent as well.

It is more than just "cool". Furthermore, I would probe and inquiry what you actually mean by a positive definition of intelligence so we can get the claim better instead of a negative defintion.

I think it depends on your definition for 'intelligence'. IIRC, Wikipedia lists 15 different definitions for 'intelligence'. One common ground is 'the ability to solve problems', which covers some aspects of how the term 'intelligence' is usually used, but not necessarily everything.

If your 'problem' as a cell is to feed yourself, it is a clever apparatus that solves that problem, and can be depicted as intelligent.

If you understand the 'fully automatic mechanism' completely, you agree that it solves a problem, but you deny that this is intelligent. However, the denial depends on your understanding of the mechanism. If you assume that the world is deterministic (that means, every effect has a definite cause, to put it simply), we cannot be something other than a 'fully automatic mechanism', however a very, very complicated one that we do not fully understand. Indeed, in a deterministic world nothing can be different from a 'fully automatic mechanism', and, by your implied definition of intelligence, nothing can be intelligent. This conclusion leaves the definition of intelligence as something 'other than fully automatic mechanism' rather useless, as it does not distinguish anything. (If you do assume that the world is not deterministic, a similar argument still holds, as long as you do not imply that a person can determine the outcome of the non-determinism, which sounds strange to me to imply.)

I think what you want for something to call 'intelligent', is that you don't immediately understand how the mechanism works and the assumption of some kind of thought process and/or planning. The first property unfortunately depends on what you might consider your own intelligence, which makes it difficult as an objective measure of intelligence so that you can argue about it in a forum. The second property (thinking/planning) might be present in the cell in a way that you don't see (eg. see storage of the current value of 'now') and, similary, it might not be present in us humans in the way self-perceive it.

More than once did I experience a seemingly 'intelligent system' that I found fasciniating, which, after I studied it, I did not think of as intelligent anymore (eg. my first impressions of Weizenbaum's Eliza), which for the aforementioned reasons I think is kind-of silly of me.

I agree with you in that it is very subjective to define intelligence.

I personally define it as more of an emergent property where you increase the complexity of the problem solving machine to the point where it begins to be more than just interconnected automatisms.

Maybe it could be said that processing information is not intelligence by itself. Intelligence, for me, requires the ability to adapt the way one processes information to the task at hand, either by changing it entirely of adapting and tuning to get more out of it.

A fixed mechanism that controls a variable can't be intelligent.

Everyday I read something new about life on earth and my will to exist amongst the ecosystem grows stronger. The life that surrounds us is so vast that we cannot possibly fathom the extent to which it is working at.

Speaking of single cells, here is a detailed model of the eukaryotic cell that includes animations showing Transcription and Translation (DNA -> RNA, mRNA -> protein): https://grimoire.bio/eukaryotic-cell

Of possible interest is John T. Bonner's book The Evolution of Culture in Animals

previous comments, https://news.ycombinator.com/item?id=39777229

Thanks! Macroexpanded:

The baffling intelligence of a single cell: The story of E. coli chemotaxis - https://news.ycombinator.com/item?id=39777229 - March 2024 (213 comments)

Beautiful website.

Is a single cell, storing state in the form of chemical modifications and using it to compare "now" with "a few seconds ago," the same computational primitive that a neuron uses, only slower and simpler? Curious

That's the thing; neurons in ANNs are really just processing the inputs they receive from other neurons. But natural neurons are not necessarily restricted to that.

This reminded me of the work of Maturana and the concept of autopoiesis

Anyone read Signature in the Cell by Dr Stephen Meyer

No, could you share the connection you're making to it?

It's about the complexity of a single cell - the number of mechanisms it has in place on how it works. He sets up the argument for Intelligent Design using the Cell.

Didn't expect to spend so much time on that article. Interesting

It's an amazing design. Did we look at this to create the electric motor?

No. They were observed in the 1800s, but the structure and mechanism were not elucidated until around 1970 https://www.nature.com/articles/245380a0 It was a big surprise to see elements that greatly resembled existing motors (stator, rotor, etc) and in fact was one of the things that inspired the concept of intelligent design (even though, we believe instead that it evolved spontaneously)

The problem with saying that God created life is that it doesn't actually explain anything because then you have to explain where a vastly more complex God came from. If the universe or life requires an explanation, adding an eternal God doesn't simplify the problem so we might as well cut out the middleman and view the physical world itself as the eternal starting point.

I'd be careful about calling the physical world eternal, since we don't actually know that

Have you seen Attack on titan. Watching that whole show, half the time you dont know why something is happening, except that it is cool. But once you know the whole story, you can piece things together no problem.

God (any god) is a comforting mythology to help explain the things in life that are hard to come to terms with, but it's a mythology, it's no different to believing in Harry Potter, or Dionysus.

Bringing the concept of God into a biology discussion is sort of embarrassing. "Intelligent design" is for the unimaginative mind. We must always be open to being wrong, evolution fits what we know now, but I'm sure in a century we'll have a better understanding and it'll be a more complete understanding, but what it won't need is a divine eternal.

The world's not flat, there is no god. Can we get back to science now.

> “God (any god) is a comforting mythology to help explain the things in life that are hard to come to terms with”

An Alien God (as a way to view evolution) from 2007 on lesswrong, isn’t comforting:

https://www.lesswrong.com/posts/pLRogvJLPPg6Mrvg4/an-alien-g...

"""the Shaper of Life is not itself a creature. Evolution is bodiless, like the Judeo-Christian deity. Omnipresent in Nature, immanent in the fall of every leaf. Vast as a planet's surface. Billions of years old. Itself unmade, arising naturally from the structure of physics. Doesn't that all sound like something that might have been said about God?

And yet the Maker has no mind, as well as no body. In some ways, its handiwork is incredibly poor design by human standards. It is internally divided. Most of all, it isn't nice.

In a way, Darwin discovered God—a God that failed to match the preconceptions of theology, and so passed unheralded. If Darwin had discovered that life was created by an intelligent agent—a bodiless mind that loves us, and will smite us with lightning if we dare say otherwise—people would have said "My gosh! That's God!"

But instead Darwin discovered a strange alien God—not comfortably "ineffable", but really genuinely different from us. Evolution is not a God, but if it were, it wouldn't be Jehovah. It would be H. P. Lovecraft's Azathoth, the blind idiot God burbling chaotically at the center of everything, surrounded by the thin monotonous piping of flutes."""

> Evolution is bodiless, like the Judeo-Christian deity

It sounds like you actually refer to pantheism, which is very much different from Christianity.

It would be logical to first explain all the phenomena of the universe down to the lowest quantum level, where it all breaks down into pure math for which no one currently has any non-contradicting theory, before even attempting to try explaining where the vastly more complex God came from.

It also could very well be the case that there's a limit not only to the complexity of things a human mind can grasp, but also to the simplicity thereof, and the concept of God may be located beyond the latter.

That depends entirely on your definition of the term God. In fact Christianity argues that anything non-transcendental[1] can in fact not be God, precisely because of the point you make. This was one of the distinctive properties of Christianity in the first centuries, which is why it was perceived (correctly) to be the enemy of all the entities the Romans called God, which resulted in the persecution.

(So to state it differently, Christianity says that the God which you are talking about, does indeed not exist.)

[1] Transcendental meaning that it is not only the origin of specific things, but of space, time and causality itself. God is not before the big bang, because the meaning of before is in fact something that only exists in the universe. God is as much before, after and next to the big bang as he is to now.

This is why to me it seems pretty logical, that modern atheism occurred in a christian dominated culture. It's not that the atheists invented a new idea. They simply "forgot"(more like didn't agree) what Christianity says a God is, returned to the definition of God from the Antiquity and "simply" rephrased the criticism Christianity already voiced for millennia.

While biology has inspired many inventions, the electric motor is not one of them. Electric motors were invented in the mid 19th century, before cells were known to be anything more than simple blobs of protoplasm, barely distinguishable under a microscope. It wasn't until high powered microscopes were invented that biologists had any clue to the previously unthinkable complexity of cellular life.

Insane site

Wow you weren't kidding. It is really well done.

It would be mistake to assume that something which appears to be a single cell is a separate from its environment. Space connects all things and if Space happens to be awareness / consciousness then intelligence would be everywhere affecting all things.

The line between quackery and atomic science is so thin it is a bit cuckoo.

A single cell is quite literally defined by its membrane that… separates it from its environment.

And that is our definition. Read Kant

> [...] if Space happens to be awareness / consciousness then intelligence would be everywhere affecting all things.

The name for this kind of random connection of disparate things, implying meaning, without providing any, should be "'if' abuse".

You are reinventing pantheism.

>> if you understand the signaling pathways of bacterial chemotaxis you can disrupt them;

You don't need to do that. You are pursuing a goal that is not assigned to you by nature. This goal is assigned by the business and a perceived "human benefit".

This end-less understanding and influencing of the things should stop. Even if you say there are medical benefits to such investigations, you are going against nature and trying to control a few things while ignoring the full context that turns against you.

This is just some form of druidism.

Wel, here is the cycle. Tech and food, opulence gives you free time, free of struggle. The idle people indulge in more science and produce more opulence and more idle people. Tye climate and bio ecosystem turns against the excessiive idle population. Struggle is again reinstated. The new generations give control back to nature.

Right the goal is to have opulence, free time, and also not to become decadent. There is a process by which this gets better over time (generations, cycles), each successive cycle has the potential for retaining memory/historical consciousnesses/psycho-social development. Just intentionally doing nothing/halting progress is not the answer and also it just trivially leads you to being wiped out (like via military means) by people who actually do things/progress. This is the basic flaw in Kaczynskian thinking, it requires all of humanity/nations to 'collude', it's a basic game theoretic issue.

>> There is a process by which this gets better over time.

It's a zero sum game. It appears to get better only because you are only looking at the corner that was wiped clean, but not the other corners where the garbage went to. And for some reason the "wiped clean" state is considered a good thing. Other misconcieved "good" things are straight edges, flat surfaces, smooth curves, uniform colors and shapes, artificial uniformity in general, thinking hard, smartness, precision, excessive hardness or strength, extreme skills, any characteristics that reflects a lot of human/macine effort etc.

Human progress doesn't require too much of human agency to the extent of causing irreversible effects or a great risks. Right now the agency is completely out of pace with the nature, arrogantly driving things and crumbling down when nature stares back at.

It's simply the case that material conditions are better than they were 50,000 years ago and also better than they were 2,000 years ago. It's also the case that overall the general level of refinement of the average person is higher, you can just read the things people used to do to each other and how they lived.

Again this recourse to nature is just druidism. You have a particular metaphysics and morality, it's just wrong.

You are literally posting on the internet, likely from an air conditioned room. You can and should stop using these things.

Reply on news.ycombinator.com