Taking the 2nd law seriously - upper bounds on the integrity of biological organisms over time

Like Turing’s paper (on spatial patterns in biology), but for communication patterns in bacteria

• one to many

• things work together

The limits of computation in or between cells, due to the constraint of passive diffusion, vs how fast or efficient compute can be in computers with the capacity to tell information where to go

• To what degree did neurons solve this problem - and to what degree have other cells?

• Is the major innovation of the brain the capacity to communicate directly with another cell, vs the constraint of having to use passive diffusion? - to get feedback on

Did we evolve from a biofilm?

• Why don’t genes do more interesting stuff in bacteria, if they are in all of the bacteria? Why don’t they coordinate more interesting group-wide behavior?

  • Do the constraints of communication (don’t want to just through quorum sensing molecules out into the soup of the ocean) make this more feasible in multicellular organisms or a sessile community?

  • Does this mean evolution from a biofilm or multicellular community is more likely if one is later a multicellular organism?

• This would make so much more sense than evolving from a single eukaryotic cell, because then the cell would evolve in the context of a population and evolve corresponding mechanisms to work int eh context of this collection of cells

• What did biofilms evolve into?

Do a teardown of Dawkin’s Extended Phenotype, enumerating all the concepts covered and commenting on them

List of ideas to look for structure in latent spaces in biology

• Understand, for example, this lab’s work? https://goodarzilab.ucsf.edu/

  • Developing nn models for genomics since 2004

• I guess there is a difference between using predictive nns to, in some sense, replace or speed up experimentation with somehow interrogating the latent space of those nns to…extract insight? I guess this is why interpretability research is hard (and why it would be interesting to understand interpretability research). Well, then, I guess, there are probably clever ways of extracting insight from trained models (for example, creating canonical TFF binding sites by looking at activations).

• To what degree are LLMs different than previous models in this field? How is using them different than using any other models?

• What are the best next problems to tackle in biology?

• What does a base model look like in biology? Is there any reason it should be an LLM? Why do LLMs work well for protein structure prediction?

• Use cases

  • Use transcriptional data to characterize cell state (do you need ML for this though?)

• Is the next few decades just going to be like ‘learn the right latent space’ and then just predicting where things will go in a biological system

  • Why does that feel unsatisfying?

Understand why it is that (mostly physicists?) have a heuristic that power is correlated with simplicity and beauty - what is the actual evidence for this?

• What does it mean that the ‘laws of nature are written in the language of mathematics’ - how could it have been otherwise?

• Are some sciences more driven toward beauty than others?

• How much does what we find beautiful say about us, as humans? To what degree does it reflect some property of the universe, if it does?

  • Is this related to the anthropic principle?

• Physics explains what will happen given what is there today, but what explains what is there today - is that biology?

• What if understanding what we find beautiful in science, and how to relate to that when it becomes decoupled from progress in discovery is something at the core of interest?

• Terrence Tao on ‘What is Good Mathematics?’ - https://arxiv.org/pdf/math/0702396.pdf

• What does beauty in biology look like?

  • Scaling laws - things that work on a specific scale for a specific reason (for example, averaging over noise, physical laws converging at a given length scale)

  • How do you make something which gets better without breaking itself?

What will neural nets trained to best model the world find beautiful?

• For example, we find beauty in color distinctions and combinations, but the distinct cutoffs and color identities are completely arbitrary (are they?), our qualia - what will neural nets most closely trained to predict nature find beautiful?

A long list of all of the most beautiful things in biology

• Cells as nanofactories

• Energy scales corresponding to the same length scales you see in cells

  • Related, noise averaging out on higher scales

  • Schrodinger discussed quantized energy on the cell scale, and correspondingly less noise in a continuous context, vs noise on the continuous larger scale

  • The fact that you can use EM fields in the cell

• Life only arose spontaneously once

  • All body plans morphed onto each other

  • Come from one origin, for example, the way the brain built on the initial brain

• The bacteria is like a bag of DNA, the nucleus is a pretty specific constraint on size

• Cells are mostly run by passive diffusion, not sending information directly

  • Molecule can talk to every protein in the cell in ~1 s

  • Not sure this is a constraint - I’m curious what design principles this gives rise to?

• Cells don’t care about velocity magnitude and direction information (you can freeze and unfreeze a cell - turning off and on all of the velocities - and it will still work)

  • Is this a consequence of passive diffusion?

  • (Kind of related, might be separate), you can increase many whole-organism parameters (for example, ant speed, or increasing worm lifespan) by increasing temperature

• Some things in the cell literally look like waterwheels, or motors

• Modularity in biology

  • In protein domains

• Cascades in signaling?

  • Also in making proteins

• DNA as a 3D information structure

• Endosymbiosis - mitochondria come from bacteria

• How many engineered ways bacteria have to talk to each other

• To the extent to which we are patterns of information, DNA and one cell lineage have both stuck around for billions of years - this is, fundamentally, the nature of life (the nature of life is to stick around)

• (Possible)

  • We’ve completely mapped some things - for example, cell fate maps in C. Elegans

• (Maybe)

  • You’re never really sure at what level evolution is acting

    • Replicated elements in the genome

    • Cancer

• (Not sure)

  • Only neurons talk to unique cells?

• (Not sure)

  • The Hardy-Weinberg equilibrium is true after one generation of random mating, and is a good approximation empirically in a lot of situations, possibly also interesting to mention the binomial theorem

• (Cool but not sure beautiful)

  • At Low Reynold’s number, if you push a bacteria at 30 um, it will stop moving after ~0.1 Angstrom if you stop pushing it

• (Cool but not sure beautiful)

  • Your normal intuitions for things can be very off - for example, the size of a protein vs the size of RNA

  • An ion channel holds ~4 ions, but 10^7 can pass though it per second

  • Molecules can move at around the speed of sound

• (Cool but not sure beautiful)

  • We use bacteria as factories to make drugs

• More functional, while interesting

  • The flywheel of biology - we can now use biology to make discoveries

  • Related, we can use increasingly complex biological entities as drugs or tools

• An interesting fact

  • That Schrodinger predicted the existence of an aperiodic information-containing crystal

• What makes something beautiful?

  • The existence of a constraint, the ability to get to the same thing from multiple different ways, or see how something in nature must be true?

• Fun questions

  • What are the limits on what humans can do, for example, the amount of force hypothetically possible to exert in a punch?

Do we live at lengthscales at which physical laws become tractable

• This could mean ‘simple’ or something like that, it would be weird if it was understandable and I’m not sure I’d claim that

• Biology - cell-level (interesting stuff can happen with quantization) + higher order (noise cancels out)

  • Does thermal noise being additive, and typically at a specific scale, cause or mean something here? What did Schrodinger say about this?

• Physics - Newton’s laws in every day life as simplified assumptions

I’m not sure I actually want to write this but I could write - why is evolution by natural selection not tautological

Why is physics so interested in the anthropic principle, and why is quantum mechanics so interested in measurement?

Physical principles and limits of communication in cell populations

• Is there a physical limit to the number of cells which can ‘usefully’ communicate - is this somehow related to diffusion timescales?

  • How do different types of cells (for example, neurons, immune cells, cells in a bacterial population) communicate - on what timescales, and what mechanisms support this communication?

• Do cells which can communicate with each other (share this mechanism) tend to be more genetically related than cells which can not? How many different populations of cells which can communicate with each other are there typically in an overall population? What are the different methods of communication this is relevant to?

Genome structure

• Do some genes try really hard to segregate with other genes? How would we know this was happening if so? What are ways for this to happen other than them being linearly close to each other? Is there meaning to which genes are linearly close to each other? How would we know if so? What does the 3D structure of the genome mean?

What things tend to have tradeoffs to fitness, and what things do not?

• Is there a general principle of why certain things affect fitness? Is it, if you really examine the cases in laboratory evolution carefully, typically a tradeoff of the total energy the cells has available? Or is there some discrete tradeoff in a resource which normally does one thing then needs to do something else? How much can one directly link fitness increases or decreases to the levels of some available metabolite, or another variable?

• What scale of metabolic fluctuations are relevant to the cell? In what ways is it equipped to deal with this, what principles guide this?

What types of ideas is statistical thermodynamics good at describing?

• What are the characteristics of its system components? Amongst what types of particles is statistical mechanics most powerful at predicting overall behaviors?

Continuous vs discrete in biology

• For example, species as discretely things which can or can’t reproduce w each other vs continuously things which are connected by variation

• Mildly related, but lack of binaries in disease definition (continuous grouping of systems, more discrete if there’s a distinct category of cause) - I’m not sure this one actually fits though, ugh

Strategies of either sticking around more or replicating to dilute/segregate damage in an organism

• Weird that there’s been one continuous line of cells for ~billion years

• Is replication required to dilute or segregate damage?

• Why can’t one thing stick around without replication?

How does time link to different things, like energy usage or temperature?

• What would it feel like to live in worlds where those parameters were changed (for example, if society had some cap on how quickly it could use energy - it might feel like time was moving a lot slower, although our metabolic rates would be the same)

• Is all life just getting rates to run at compatible times, aka controlling the relative experience of time of different components of the system

Additive vs multiplicative noise (maybe)

• What things in biology have additive vs multiplicative noise?

• Do neural spikes and gene expression have multiplicative noise?

• Is thermal noise mostly additive?

What is beauty?

• Physics and then biology

• Math

• What is beauty in writing?

  • Strunk and White? Adhering to some structure, following some principle of conciseness? Why is conciseness relevant to beauty?

  • What is it about language to do with the sounds? How does that map onto anything else? Music vs meaning - but surely the two are more connected?

  • What is coherent vs incoherent writing? When is the latter beautiful? Show, not tell - what does that mean, and when ought one to deploy it?

• Essay on the similarities between a beautiful proof and a beautiful essay?

What are all the examples of mathematical theories which, by their structure, got more out than you put in?

• i.e. examples from Wigner’s unreasonable effectiveness of mathematics

Structure of truths in biology, truths in (?), truths about structures that a simulation could encode

Kind of a blog post - list of important papers from universities

New kind of thinking - enabled by BCI?

• Able to truly understand dimension, do something similar to what the Greeks did with the independent reality of objects (abstract math)

• Analogy to what happened with the Greeks discovering the conscious reality (?) of abstract mathematical objects, an important jump there

• A true understanding of dimension

• An increase in capacity (i.e. working memory and stamina) vs an increase in imagination (what is dimension, another meta similar to Greeks finding abstract reality to mathematical concepts - conscious experiencing of them?)

Game theory

• What if there was a game theory where agents could choose what other agents to play with (instead of being randomly matched)

How do you tell whether something is evolution by natural selection or not

• ways we use the term, and ways that is confusing

• how would we tell?

• for what things is it valid to use this? what do we mean if not this when we use the term otherwise? what can we say about that?

• does the anthropic principle have anything to do with this?

Would c want to use c?

I could add a set of pictures to my website