Menu‣
How to solve problems Solving Aging
Fiction The Hike
Other Projects Arcadia DEEPWAVE
You can comment on everything. Or give me (harsh) personal feedback.
Made with 💙 by me (how to).
Dedicated to my family.
Status: Draft Epistemic status: Extremely weak, These are my 2023 believes, entering the field of aging biology. I’ve always been fascinated by it, so I have some preexisting models, but my study of the details extends only to reading roughly half of Molecular Biology of the Cell by Alberts. So expect major revisions.
Biology has to be simple
You've probably heard this: biology is the ultimate legacy code, written by millions of years of evolution, adding fixes on top of fixes. No wonder we can't understand it! And we don't even have a good debugger! I think this is largely wrong, you will need the details to build the debugger - the thing that helps you figure out what’s going on - but the actual solutions to aging will be roughly as complex as washing hands or pacemakers.
Complex systems are powerful, complicated systems fragile
This is just simple intuition, but I don’t think a living being run by legacy code would survive very long. If you’ve ever worked with old software, you know that they are super fragile and break all the time. No, you need simple, but intelligent architecture. You need captialism over bureaucracy: simple systems that if working together lead to strong emergent behavior. This is how you get resilience: Conways Game of Life will always run longer (and produce more beautiful things) than any written top-down code.
Prediction: we should be able to find simple mechanisms in biology that if changed lead to very strong effects.
Genes/Metabolism are the wrong level of abstraction
So why do we see so few strong effects from simple systems (examples would include washing hands, variolation or my hypothesis that this ant lives 10x longer by just smelling like the queen)?
In [[Understanding = Economics + Tech Tree]] I argue that everything is just "things falling into shape", whereby the shape is how likely possible things are (or the ways in which things prefer to be put together, organic chemistry, technological feasibility, material science, training data) and the fall is the process with which things are tried out (gravity, evolution, SDG).
This is relevant for biology: I think most scientists have lost the forest for the trees
You have to understand the details to know what's possible: Organic Chemistry / Small Molecules / Proteins / DNA / Epigenetics only tell you something about the Tech Tree, they are single points of data about what is possible.
You have to understand the whole to know what exists: The actual driving forces (economics, gravity) look more like evolution, systems biology, information / game theory, swarm intelligence.
The simplest example of what I mean is cell metabolism. You could spend literally trillions of dollars exploring this:
Courtesy of expasy.org
But what I’m seeing is this:
Specifically, I think metabolism (the changing of one chemical to another) and gene-regulatory networks (changing how things change) are dynamically changing to meet the cells goals. It’s taking in environmental concentrations as inputs and outputs decisions like “should I kill myself?”, “should I replicate?” or “should I just keep being this specific skin cell?”.
If this is roughly accurate, it has huge implications. It makes me extremely skeptical of small molecules, proteins, and even gene editing!
First, it makes every intervention like trying to change an image models output by influencing individual nodes and weights. This is crazy, because even if we had good “Bio Interpretability” (where we understand a single nodes influence on the final output), it would still be very costly, because we need to tweak different nodes in different cells and somehow compute which node in what cell locally and, and, and …
The rebuttal to that would be the claim that somehow that network is not as complicated, that it has master switches, or at least nodes with a large influence on the final output. This seems logical, information has to be integrated somewhere, and p53 or the Yamanaka factors are pointing in that direction. Even if this is true to the extreme, where aging is entirely programmed, we’d have to find those switches, and then activate them non-uniformly across the whole organism.
A different critique would be that we can use AI to map other models. If we just screen all possible drugs, build a perfect simulator, etc. we will finally be able to cure all disease. To this I have my second comment: what if the network doesn’t want to be changed? What if we’re dealing with a system that has been fine-tuned over millenia to resist “value drift”?
Exposing too much power to a single node or pathway makes you easier to attack. Whether it’s internal shifts in goalstates (like cancer), or external goals like from pathogens, you represent vast ressources that could be free lunch for someone else. If you had only one information aggegator (master switch in the metabolic network) you would die soon (and indeed it seems that whales have more copies of p53), so I hypothesize that cells try to be as decentralized as they can get away with (too much becomes inefficient).
Again, this is bad news for supplements and drugs: in most cases I think homeostasis works against you, not for you. Yes, you can probably find ways to break the system (their are many toxins out there), but making the system more efficient / resilient / better, is going to be very hard, because what you’re trying to do looks dangerous from the perspective of the cell.
Evolution might not like us
This is controversial, but I basically think Joshua Mitteldorfs theory of “aging as group-selected adaption” is pointing in an interesting direction. I know everyone hates group selection, but I think his specific version - where aging is optimized to prevent population collapse - actually makes a lot of sense. I don’t know how this will lead to practical interventions