This is a very interesting article/video, and well-worth reading to get a sense of how complex proteins are, but I’m not convinced by the principle argument that proteins don’t constitue “machinery”.
The core philosophical question is actually interesting: what is a machine?
The author suggests a machine is anything with solid parts of that have a specific function, and goes on to argue that the metabolic machinery or cells doesn’t correspond to any of these three criteria. I’m not sure it matters whether the parts are solid, I don’t think multifunctionality disqualifies something from being a machine, and despite the limitations presented, I am not even convinced that it is senseless to decompose a cell into components. On this last point, it bears emphasizing that his criticism bears the analytical method in general, and is not specific to biology. There are edge cases in any decomposition of a whole into parts.
So while it’s interesting to see how these philosophical questions are at play in cellular biology, the case remains to be made that “machinery” is a fundamentally bad analogy. Here it also bears emphasizing that confusion between the subject of study and the metaphor is a risk with all metaphors (the map is not the territory), not especially this metaphor.
In sum, I wish the author had focused on the philosophical substance of “what even is a machine?”. That’s bound to be a richer investigation.
I think the author covered it sufficiently, because the author is correct that if you just walk up to someone with the referenced videos, they are going to come away with the mistakes the video author discusses. They're going to think things have specific functions, in specific ways, in the super-orderly manners presented, etc.
Redefining machine to the point that it happens to cover what cells really are won't change those misconceptions, or at least, it won't change them for a very long time and until you get those definitions propagated far and wide. Which will be an uphill battle because they are frankly less useful than the current definition we have, which works in the real world we all interact with.
Nobody is hearing "the cell is a machine" and then goes off thinking "ah, what they clearly mean is that it is a whole lot of goop interacting in shocking complicated ways with each other all the time resembling literally nothing else a human calls a machine". Playing games with definitions is, well, exactly that: Playing games with definitions. It doesn't actually change anything in the world.
> Nobody is hearing "the cell is a machine" and then goes off thinking "ah, what they clearly mean is that it is a whole lot of goop interacting in shocking complicated ways with each other all the time resembling literally nothing else a human calls a machine".
Erm, a lot of the "nanomachine" people think exactly like this. Even simple things like "gears" do weird things an the nano level.
Nevertheless, the biggest problems with the conception of a "cell" would be solved if we quit drawing them in 2D. Draw some cells as spheres. Draw muscle cells as slightly elongated cylinders. etc.
If we drew the cells in 3D with chunks missing for labels, students would have a much better conception. Starting with: "Oh, hey, there isn't a huge amount of empty space--there's crap everywhere. These things are packed".
I suppose that's fair. I guess I'm lamenting the missed opportunity for an interesting philosophical discussion, and perhaps wishing the author had presented his (otherwise very interesting!) thoughts through a more nuanced lens. I don't think he would even need to change the title.
I still enjoyed the article, and would upvote it again if I could.
GEB explores a similar idea in investigating where a program is written
It's implied that the distinctions are somewhat arbitrary. For example, a song is "written" on a record with a bunch of grooves, and the record-player is the machine that plays it. But if the record-player spun backwards instead of forwards, you'd hear something completely different. So in some sense, the record-player also holds part of the program.
And all this runs parallel to the idea that a program is also data, IMO. I think a 'non-arbitrary' definition of a machine would be very general.
This was similar to my understanding (with my limited insight).
If we are researching a particular cell interaction, thinking it as a machine, we may walk away with wrong conclusions if we hypothesize, set up a test, and study results, thinking that this machine is expected to work the same way every time.
The author said specifically that he did not want to make the video about this philosophical point:
> I imagine I'll get some pushback on my definition of a machine. I didn't want to turn the video into a definitional debate so I glossed over it pretty quickly. That was intentional.
The specific definition is kind of besides the point, basically what he’s saying that if the cell is a “machine” then it is quite unlike any machine we are familiar with. It is far more complicated than anything humans have built or even imagined.
I’m not sure it’s entirely beside the point though, especially as his warnings and criticisms hinge on that very definition. To my point: there’s a nuanced (and, I would argue, more exact) definition of a machine that cellular systems meet.
That lay people (and the occasional person who really ought to know better) misinterpret the metaphor is not a terribly interesting point, and, I think, not a reason to throw out the metaphor.
To the authors credit, the explanation of how cells and proteins differ from everyday machines was a fascinating read.
Yeah I agree with you. I think the author’s main point that this is a harmful analogy is not correct. We need some model to make progress on the complexities.
machine (n.)
1540s, "structure of any kind," from Middle French machine "device, contrivance,"
contrive: to form or create in an artistic or ingenious manner
According to these definitions a machine is something that humans create.
We understand what we create. When trying to understand complex things, we form metaphors based on our actual understanding.
Whichever is the most advanced thinking of the age is used at a metaphor to describe life and nature. When it was steam, we used popularly understood concepts of steam engines to describe life. When it was electricity, we used electrical metaphors. Now that it is computers, we use those concepts.
We are also very overconfident regarding how well we understand and control the world. Having achieved smashing success in some areas, we then think we have total control. Chronological snobbery, if you will: https://en.wikipedia.org/wiki/Chronological_snobbery
The danger is then, at a popular level, we take the metaphor literally and turn it around to be something reductive: "Life is no more than a machine." That is frankly not true: this is where the metaphor falls apart & it's the danger the article describes. Life is more than a steam engine, it's more than electrical circuits, and it's also not a computer.
Life may be a purely material phenomenon (i.e. no more than atoms, void, and time), but that does not mean it is "just a steam engine / computer" etc.
> The core philosophical question is actually interesting: what is a machine? [...] I wish the author had focused on the philosophical substance of “what even is a machine?”
Indeed, and here we enter the land of metaphysics. Empirical science is not inherently mechanistic, though you cannot be blamed for making the mistake of thinking so given the crude metaphors we learn in school or how the methodological presumptions within scientific disciplines affect how we model and make predictions about phenomena. Even when not intended as literal characterizations of reality, against the backdrop of scientism, it is easy to reify metaphors into metaphysical truths if you don't know any better, when an authority figure compels you to do so, or when you fail to detect the whiff of contradiction. From experience, I will say that it is difficult to dislodge the mechanistic bias after a childhood spent having it reinforced in school and educational material. It might take a while to heal from the brain damage.
But where this article is concerned, I think the author missed what is the essential difference between a machine and a living thing, namely, that while the parts of a machine are arranged according to ends external to the arrangement (the artificer) and, more importantly, have no inherent tendency to operate the way they do, the "parts" of living things are inherently ordered toward an end. To put it in technical language, the parts of a machine (or artifact) are joined together by an accidental form, while what makes an organism is a substantial form. Of course, a mechanistic view of the world excludes substantial form and telos as a presupposition of its founding charter, not as the conclusion of a demonstration from observation or logical necessity, and so a "mechanist" is conceptually prevented from distinguishing machines from living things from the start.
Yes, the core philosophical question is very interesting. Unfortunately at this time we can't resolve that core philosophical question. When we can resolve it, I think we'll be able to move it into the part of philosophy that's also called "science."
The cell might be understandable as a giant machine, but I think the author is claiming that we either don't know that yet, or that the machine may be enormously more complex than some folks are suggesting.
Either it's a philosophical question, or it's some other kind of question. "Philosophical question" implies that we can't answer it at this time. (Or, equivalently, we can build equally convincing but conflicting answers.)
One claim in the article is that we can't answer the (much narrower) question, "Is the cell a machine?"
I don't see that saying "well that's a philosophical question" is so different from saying "as scientists, we don't know."
I'm curious, have you ever studied philosophy? I ask because I often hear this kind of rhetoric from people who have a distorted idea of philosophy, and don't grasp the impact that philosophical thinking has on the proper performance of scientific inquiry.
In the current case, we might not arrive at one definitive answer, but we can accomplish at least three things:
1. Eliminate bad answers
2. Understand the edge cases of the good answers
3. Ensure the scientific community is acting in accordance with 1&2
I think you may be surprised by how often the scientific community spins its wheels on nonsensical questions, due to a lack of philosophical grounding. One example is the mind-body problem. Gilbert Ryle shed light on the category error that underpins Cartesian dualism in the 40's. Until then, scientists were (quite frantically) searching for the locus of the soul.
How about Stephen Hawking, Lawrence Krauss, Neil Degrass Tyson, Richard Dawkins?
I have no dog in the fight, I just remember my favorite authors not been too keen about philosophy.
There really is no great mystery question - because "machine" is a word not defined by the universe and is controlled by humans, who can use whatever definition they want. So whatever definition you use, this word has no external objective constraints, there will be a common ground between people or not and there is no objective way to make everybody agree on one definition, it's all voluntary. (Even if they do, the agreement is never complete because it's only about the aspects talked about, and then words are not reality and come about indirectly through many processes in the brain and never completely match between different people, or even the same person over time)
"What is a machine" is a great way to waste time doing nothing, unless you can beforehand agree on constraints, e.g. by looking for a definition for one clearly defined purpose. But that's just moving the goal post to a little outside the box that then this definition is tied to.
There is this strange movement, or maybe it isn't strange, when I was younger I followed it too, to take words way too seriously. There's entire fantasy stories about "true names", some kind of absolute comprehension about something expressed as a word. There also are practical examples, for example I once read about a father whose daughter had some rare disease about which nothing really was known. The only thing that had happened, at least by the time of the story, was that somebody had encountered this disease previously and attached a word to it. But without any content, there was no knowledge about how it occurred or how to treat it. But when the father was told that name he said something like "Finally we know what the problem is, I can breath much easier" (words partially made up because I don't remember, but that meaning was there).
Just look at unrestraint, non-concrete discussions about words and there arbitrary definitions, looking for something "philosophical" behind those words. In this context, a blog post I once bookmarked (hey, I finally get to use my bookmarks!): https://www.lesswrong.com/posts/7X2j8HAkWdmMoS8PE/disputing-... - with some practical advice, not an attempt to solve anything "philosophically".
I see sooo many comments here basing their agreement or disagreement on some definition of the word "machine". The default these days is to try to find something wrong with somebody's statement(s), by exploring the edges or even beyond of definitions of some word or words, making discussions very tiring. When instead one could just try to find the idea the person wanted to express and not get hung up about some words. Because deep inside the brain there are no words, the ideas one tries to express are much more fuzzier and flexible and trying to turn them into words is a very inexact process. Our brains don't think in words, except for a tiny part under our direct attention.
The concepts leading to the use of this or that word, and the many words accompanying it, are never the same in different brains, and not the same in the same brain over time. To make the best use of our very limited communication tools that serialized extremely intricate four-dimensional patterns (brain structure plus the dynamic behavior) into rigid words you have to want to communicate and to understand and be positive, at least when something is a little off-center from the usual things that everybody involved in the communication has already established a good synchronization on.
When discussing if we need to buy more milk because there is none in the fridge we have a much easier time with the communication and with agreements, because we are not going anywhere near the edges of well-used concepts (what is milk, what is a fridge, what does "empty" mean, etc. etc. - all of those can lead to deep disagreements at the edges). For a text like the one here some more allowances would be better to get something useful out of the communication.
IMHO the best way to make something out of this article is if you already have a pretty good and somewhat deep comprehension of organic chemistry at least. The many stages of going deeper and deeper down the rabbit hole from the beginning teachings of atoms and chemical bonds to more and more complex understanding that you get from university chemistry, where it gets more like physics, coupled with seeing it "in use" for large and complex organic molecules, is I think a good preparation to understand the article's point.
This is a great point. Maybe the focus should be, "how not to think about machines." If we build more complex functional systems (with the complexity, reliability, and efficiency of a cell) then we will certainly have to rethink what a "machine" is.
A cell is not a machine in what most people would understand under this word. It's almost like it's full of agents that each have their own goal and through endless evolutionary selection these goals between those agents became interconnected in such way a complicated way that they play this "orchestra" that is a cell. But when you look inside it's a huge biological mess. There is no central part that controls everything - cells and multicellular organisms are hugely distributed systems.
Intuituvely, a machine is a thing with an internal state and behaviour that predictably depends only on that state. However, what if the internal state exists, but cannot be completely known or reproduced?
> gives us a false sense of confidence in how much we know about the biological world
A false sense of confidence could make people vulnerable to bold false claims, which might lead to poor consumer choices and poor policy choices.
Beyond that, the approach seems really wrong-headed. Popular science education has a successful approach when you want people to understand the limitations of a model: you hook people with a fun headline of "COOL NEWS, THINGS ARE MUCH WEIRDER THAN YOU THINK!" You teach forward from the oversimplified model.
You aren't going to reach many people with a negative, moralizing message that people need to go backwards and give up a model that has given them insight. "Oh, we should never have given you this, you're not smart enough for it, you're going to make a mess."
The sad thing is that the information conveyed here is really cool, and it could easily be framed in a fun way! Instead, it's presented in a way that sends the message, "Look, you need to understand how dumb and dangerous it was for you to think you had any insight into this."
> It's finally dawning on us that the cell is not a machine
I really, really, really don't understand the sense of moralistic fear infusing this whole presentation. Reading the piece and watching the video, I feel like I'm reading and watching a religious sermon on the spiritual dangers of materialism instead of a piece of popular education to help the public have a better-informed understanding of science.
you are thinking about individual understanding, but science writers must be concerned with population understanding. People die and get born, and this is a force in the population making it know less over time. This force and education must reach some sort of equilibrium, and the author is concerned about the shape of the resulting steady state of knowledge.
It is far more important, typically, to not lie to people than it is to tell them the truth, for if you are lying to them in service of later telling them the truth, and people don't seek out further education, then most people will just believe lies, and then eventually die, never having contributed their knowing the truth to the population knowledge distribution.
On the other hand, if everything is always presented in its full complexity with a million caveats and you lose people because they don't have the prerequisite knowledge to be able to digest it, I guarantee this will result in many, many more potential contributions being lost than from those who were just "satisfied that it was all worked out already".
Your "guarantee" is insufficient evidence that lying to the population is worthwhile. The burden of evidence is on you, I think.
You're also making unreasonable assumtpions about the burden associated with not lying. You don't have to completely solve the miscommunication problem in order to do something about it, and even if you were to try to completely solve the miscommunication problem, it would not require you to include a million caveats. Basically you could solve the whole problem by not making analogies.
Yeah I mean, we could just never use real structural data and just communicate purely in symbols that way users would know it's "not real", or inherit a billion dollars and come up with a perfect simulator and high-end computers for all of humanity to view it. This stuff is HARD to make, and nearly impossible to satisfy everyone. Drew Barry's animations are "lying" because they're not a perfect representation of reality? In my opinion, it's better for his animations to exist and be "lying" than have nothing.
You're falling into your own trap of what is and isn't a lie.
Lets go with the common aphorism 'all models are wrong, but some are useful'. As it puts forth there are a large number of useful models, but wouldn't that fall under your term of a lie since it's not the truth of a situation.
The problem with reality is it feels no need to be explainable.
I think I meant to post this to the user above you....
Exactly? I wasn't the one who classified Drew Barry's work, widely acknowledged as some of the very best and most accurate in molecular animation, as "lies" ... so, not really sure what you're saying? In reality, humans can't see down to the nanoscale with their eyes, yet there are real 3D structures down there that have a physical form. If we display them as being a couple of inches large so that humans can see the shape, is that misleading? Should we never try to visualize anything for fear of it not accurately representing reality?
I guess my point is, as someone who has done a ton of work in this area, there are a million choices one can make when choosing what to visualize, and the lines between truth/reality/lies/accuracy can get muddled. A lot of it depends on context, use cases etc. For example, a star high school teacher may show one of Drew Barry's animations to their class and use it as an example of the limitations of visualization. A creationist could use the same animation as evidence of "God's perfect design".
What I'm understanding from the article and from your arguments, is that unless you can guarantee that a piece of work 100% represents reality in all its complexity, it does more harm than good. I'm politely disagreeing, I think it's better for these works to exist: "Perfection is the enemy of done".
People like this are one of the reasons biology visualization is so difficult: Nothing less than a perfect simulation is good enough, and any editorial/design choices to make certain processes perceptible to humans are problematic in some way. It's all true really, but there are many valid reasons to create visualizations that are less-than-perfect. But yeah it's a fine balance between accuracy and the appearance of accuracy (e.g presenting a process as linear when it's really stochastic)
I enjoy when the nice abstract models we work with are given a course-correction, and we are reminded of the profound, gargantuan complexity of the real world.
Sometimes, we get lost in our models and start thinking that they are reality, rather than just poor approximations of the real mess that's out there.
So, nice article (:
For a shorter version, that also gives some of that sense of the real complexity of cells, I like "Cells are very fast and crowded places"[1].
One good quote:
molecules move unimaginably quickly due to thermal motion. A small
molecule such as glucose is cruising around a cell at about 250 miles per
hour, while a large protein molecule is moving at 20 miles per hour. Note
that these are actual speeds inside the cell, not scaled-up speeds. I'm
not talking about driving through a crowded Times Square at 20 miles per
hour; to scale this would be more like driving through Times Square at
20 million miles per hour!
A machine to me is a combination of elements that interact to perform some function.
Wether those elements are multifunctional doesn’t matter.
Also solid is not a requirement: ropes and pullies, expansion elements, self altering software: etc are all found in human made machines. And they are not solid.
I disagree with this video: life is based on molecular machines. It just happens that these machines are really complex and dynamic. But machines nonetheless.
Sure, and depending on how you define a computer program hamlet is a computer program, and depending on your definition of Christmas movie, Planet of the Apes is Christmas movie.
The point is that for typical and sane definitions cells aren’t machines, hamlet isn’t a program and planet of the apes isn’t a Christmas movie (But Die Hard is).
There's some good stuff in this but while I agree with many of his points, I don't agree with his fundamental conclusion. I see several examples of "machines" in biology which satisfy all my requirements. In particular DNA and RNA transcriptase, the ribosome, and motor proteins are all machines. They act with very high fidelity, have a specific and necessary purpose (to the extent that an evolved system can have a well-defined purpose), consume energy to achieve their goals, and eventually break down and make mistakes.
The current belief about alphafold is that its confidence is actually a "disorder predictor": when it makes low confidence predictions, it's really that the protein itself doesn't adopt a well-defined structure.
But further than this, we actually do protein design and create new proteins that have well-defined functions. FOr example there is a class of proteins known as proteases that "cut proteins" (conceptually like scissors, but physically using enzymatic activity). A well-known protein designer at Genentech took an existing protease and made it more heat-stable (IE, didn't stop working when you raised the temperature). This protein was then sold licensed to laundry manufacturers who added it to their detergent (it does a great job cleaning up protein stains, hence the advertisement "protein gets out protein"). It's literally a machine you put in your machine to make your machine machine better.
I just think that the definition of “machine” is too restrictive. To me a machine is a very broad word and no matter how complex we find our bodies to be, it’s still ok to think of them as machines. That doesn’t mean it’s ok to oversimplify how they work or to presume we know everything about them.
Reading this on mobile, I feel really lost about what the article is trying to say. The problem is all the sources are in between the essay-bits. I would prefer it if the author just used footnotes like Wikipedia when citing a claim and stuck all the sources on at the end.
> I like when humans finally caught up to this advanced biological process and had to compare it to machines that humans created that led to this discovery just to understand it.
The solar system is like a clock. Cells are like machines. Brains are like computer networks.
It's interesting to realize how regularly we map our discoveries about the physical world to the technologies of the day, "just to understand it," and also interesting to be reminded of the limitations of such analogies.
The best and most literal parallel I can think of is DNA is like code. Yet I find it interesting, almost uncanny, that we discovered that right around the same time we started developing prgoramming languages.
> The best and most literal parallel I can think of is DNA is like code. Yet I find it interesting, almost uncanny, that we discovered that right around the same time we started developing prgoramming languages.
Eh... codons were only discovered in 1961 (Crick, Brenner et al. experiment), the Jacquard Loom was invented in 1804.
I think the better parallel is information theory and DNA coding, which happened around the same time, and there was a lot of flow from the IT folks to the DNA folks when people were trying to deduce the "code". See https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3220916/ and other things written by Schneider
I have been wondering if a better representation of a cell would actually be a large centralised message “queue”, which represents the state of the cell somehow (one entry for each metabolite molecule, or for each proteoform).
The enzymes/proteins would then sample this queue, and return the message, process the message, or add new state to the queue. So this is essentially a large decentralised message processing mechanism, and what you would expect is that over time the state of the queue would shift, representing a shift in the state of the cell.
My TL;DR take on this: thinking of proteins like rigid little single-purpose parts might fall short of the mark. Proteins are kind of wiggly, and nature can easily multi-purpose a given molecule.
Its also, the situation deforms the outcome. Plants deform and grow different in zero gravity, so its carbon nano-machinery, but machinery that uses the circumstances to shape itself. Like a liquid metal terminator, that becomes a bell when meating the casting mold.
Take allergies, here is a hyper active defense system that wants to fight for its life and looks for a challenge.
And it will find its fight and find its challenge, shaping itself accordingly, and if there is no opponent, it will take on the things close enough to it.
Which is why its such a folly to take a creature completely out of its natural environment, like taking away half of what defines the creature. And instead of accepting the mistake, some even dig in further, developing phobias against dirt, parasites and "nature contamination".
I wonder how this relates to the "mystical realm" - the yogic concept that a certain "life-force" (prana / chi) exists that coordinates, at a higher level, the behavior of cells.
It seems you are taking the notion of "the cell not being a machine" a little too far. Molecular biology has not shown the need for the hypothesis of a "higher level life-force" to explain what is going on inside a cell.
I couldn't figure out if the author is trying to make a point about the definition of machine, or if they're trying to make a point that the animations are misleading. Because they say the animations are misleading, but then they go into a discussion about what makes a machine a machine.
> the animations are misleading.
Ok, why are the animations misleading? Are they representing a process that does not occur?
> Let's go from the start. Why isn't the cell run by molecular machines like the animations suggest? Well, first we need to come up with some criteria to distinguish machines from non-machines.
Wait, what? No, we don't first need to define what we mean by machine. I just want to know why the animation is misleading as an approximate model. Does the animation represent the interactions occurring in the cell or not?
> I also know that many machines have bendy parts, and that some machines have multi-functional parts.
Ok, further digging into what people think about when they talk about a machine. I still just want to know if the thing in the video is misleading, as accused.
> the structure of proteins isn't hard and rigid like we see in the animations. They're actually more like "dense liquids" that constantly jiggle around inside the cell.
But do they do what is shown in the video? Maybe just in a less rigid manner? It's just a model! In the ted talk there's another video of some process and you can see it's all liquid. Can we focus on the animation and why it's misleading?
> This is already strike one for the machine metaphor because, if my bike jiggled this much I wouldn't be able to go very far.
Well yeah, it wouldn't work very well as a bicycle, but a jiggly machine wouldn't be a bicycle, it would be something else. Doesn't mean it's not a machine. And we're still not talking about the accusation of misleading videos.
> Proteins almost never have one shape, they have a bunch of different configurations that they shift between.
But for the model in the video, for that particular protein shape, does it represent the interactions that are occurring? Does the Helicase process the DNA strands? Does the DNA wrap around Histones?
> It is finally dawning on us that the cell is not a machine.
No, there was an accusation made that was not backed up aside from an elaborate side argument about whether the cell is machine-like. Obviously the animations don't completely represent reality. They're just models after all. Simplified for presentation and communication of a particular concept. I still don't know if the interactions in the animation are false or not. Or if the author is really calling them misleading because they're more rigid than in real life, and they don't represent the full complexity of every possible protein type.
I agree with you. It seems like the author has a grudge with whoever calls cells "machines" and doesn't answer the questions we actually have here about cells.
The core philosophical question is actually interesting: what is a machine?
The author suggests a machine is anything with solid parts of that have a specific function, and goes on to argue that the metabolic machinery or cells doesn’t correspond to any of these three criteria. I’m not sure it matters whether the parts are solid, I don’t think multifunctionality disqualifies something from being a machine, and despite the limitations presented, I am not even convinced that it is senseless to decompose a cell into components. On this last point, it bears emphasizing that his criticism bears the analytical method in general, and is not specific to biology. There are edge cases in any decomposition of a whole into parts.
So while it’s interesting to see how these philosophical questions are at play in cellular biology, the case remains to be made that “machinery” is a fundamentally bad analogy. Here it also bears emphasizing that confusion between the subject of study and the metaphor is a risk with all metaphors (the map is not the territory), not especially this metaphor.
In sum, I wish the author had focused on the philosophical substance of “what even is a machine?”. That’s bound to be a richer investigation.