Hacker News Reader

Comment by npunt | Hacker News Reader

Nobody's talking about the author's point at the end of the article, that we need better tools to reason about biology, and that the biology as a subject has a particular gap between reality and content that perhaps other subjects do not.

I couldn't agree more. Right now we're sipping the natural world's most complex ideas through the straws of simple language and static diagrams, and the constraints of all of these mediums (including school itself) in aggregate naturally lean toward making biology a rote memorization subject. Reasoning about biology in the way that it appears in nature is, in these mediums, going against the grain. It happens, but it's not the default. Your story about having a good biology teacher is this exception, not the default.

Complex things requires easy to use systems that reflect that complexity. That's the subject here - how do we build these methods of communication and understanding?

View on HN

Hey, neuroscientist here.

It's going to be a few hundred years before we get to that point, if we ever get there.

Bio is just really complicated, there may never be anything like a 'easy to use system.' We're still on the beach of it's ocean, counting the colors of stones.

For instance, in developmental bio (going from one cell to a functioning infant) we have three theories of how a cell determines what it should develop into: 1) The English model: The daughter cells get told what to be by the mother cells 2) The American model: the daughter cells take a look around themselves and determine what to be by taking a poll of the other nearby cells 3) The Las Vegas model: it's all random with lots of apoptosis and going broke.

They very fact that we think these models are right is very concerning to the field. We know deep down that none of this can be correct, but have not been able to disprove it all that well. To be clear here: dev biologists are nearly certain that their theories are crap, based nearly entirely on gut feelings. That's how gun-shy biologists are with any whiff of a 'grand theory'. That's how complicated things are.

It's not a given that bio can really ever be reduced back down to something understandable and simultaneously reflective of the 'real' state of things. That's not something nature is obliged to provide us.

Even more fundamental question is - why does a cell divide? All literature that I came across tells me "how" a cell division happens, but not "why"? What's the cause? What's the motive? For example, in Physics, the cause or motivation is reaching an equlibrium or minimality of energy transferred, entropy etc. For cell division - what's the nature's goal? Having more cells? Why?

Even searching for this is tricky, because you’ll see a ton of information on how the cell “knows” mitosis has begun and the chain reaction of what happens after that, but I found this article from 2015 [0], which refers to this Nature article [1].

[0] https://news.stanford.edu/2015/09/28/cell-division-skotheim-...

[1] https://www.nature.com/articles/nature14908

> Complex things requires easy to use systems that reflect that complexity...how do we build these methods of communication and understanding?

We've had the wetware for understanding complex systems for thousands of years. We added symbolics for communicating complex things later.

Nowadays, many people weak in symbolics are excellent in understanding, and many people expert in symbolics are novice in understanding.

I am not sure whether we are missing a symbolic form that is closer aligned to understanding, or whether we've simply overvalued symbolics at the expense of understanding.

Isn't this what we're all betting massive Transformer architectures are going to give us? Tools to explore and handle complex concepts. Reasoning may still be left to us, though.

Absolutely! Super exciting! But how precisely will it work?

Right now I don't know of an AI tool that can make a halfway decent biology diagram, let alone a complex 3d animation of a biological process you can talk to and ask questions of. The article was a call to action for this type of tooling.

I'd love to see this thread move from 'my experience in school was good/bad' to 'what if we made something that did X' or 'have you seen Y' :)

https://www.ladanuzhna.xyz/writing/deep-learning-for-biochem from previous HN topics

I have a hot take that biology is just too complex and complicated for humans to truly understand.

We have the cognitive ability to reason about the relationships between about 150 humans. So, if thats a plausible upper limit for how many genes we can hold in our head too, then we’re toast. Each cell has thousands of distinct proteins, and many more small molecules. When we knock out one gene, we regularly see hundreds change in response. We just can’t hold that large of a system in our head. Parts of it, maybe, but genes are so interconnected that it’s very hard to draw a sensible boundary between distinct “parts”. Also biology behaves in really unintuitive ways. Feedback loops, randomness, long tailed distributions. These are very important concepts for biological systems. Humans are also notoriously bad at thinking about all of them.

It’s just too big and too weird to try to think about a single cell. Forget tissues or organs.

So, I think computational modeling will be really important to teach to students early. We have to rely on computer models because its too complex for our brains.

Reply on news.ycombinator.com