Hacker News Reader

Comment by yuretz | Hacker News Reader

I can see why it isn't a local or global maxima solving function, but it surely can be modelled by one, right?

I'm a biologist so forgive me the poor wording, but there are "infinitely high" peaks/walls in the evolutionary landscape, as in: "the human urinary system has some evolutionary constraints and anatomical features that are not optimally designed from an engineering perspective, but they are generally functional and reflect our evolutionary history rather than deliberate design." <- this is from AI, I didn't remember the details, but the ureters or urethra take a long way around the inner part of the body because at some point it was a good route, but as we evolved it became suboptimal, but we can't just evolve to untangle all the pipes in our body and do some cable management... We're stuck with what we have (no tunneling electrons ;)).

So in that sense, a global maxima (minima?) solving function will probably not arrive at what we have, it has to take into account our evolutionary path, in its entirety (as in the whole tree.), because it is all encoded in 1 (or several) long molecules, and a radical changes would take a rewrite of many parts at the same time, which just does not happen. Maybe at some point we can do it ourselves... though I imagine there being a lot of unintended consequences in other places in the body :).

Another example is that sometimes complexity in a cell signalling cascade is simply there to get some timing right (ie delay some response), and the complexity is just there, and won't simplify by itself -> But then later the complexity may start to serve some purpose. Just like proteins that are "copypasta-ed" in the genome, leaving one copy free to mutate and attain a function, or not... etc etc.

That is the difference between global and local optima. We're not at a local optimum though, our bodies can probably evolve to produce half the urine or something without changing the route, but there is no evolutionary pressure to do so.

Yes, but we cannot evolve to have our urinary tracts take a more efficient route (or at least the probability is exceedingly low), even though there would be pressure to do so. (Maybe I'm misunderstanding your point)

You're implying that "half the urine" describes something that could be a local optimum, but a local optimum would be something more in terms like urine output versus energy expenditure required to achieve that. But that doesn't even capture it, because you can directly control how much urine you produce - water in, water out. So what's really being optimized for is the pressure/stress/disutility function of the kidneys in the context of how much water you are intaking. Which is presumably moderated by your access to water, and higher level conscious optimization like how much water you're wanting to intake and how much urine you're wanting to produce. And that's not even capturing all the constraints in play! But it hopefully illustrates how the scope of what is being optimized for is quite large - it's not each metric in isolation, but rather some (probabilistic) utility function over the whole.

“recurrent laryngeal nerve” as well

https://futurism.com/evolutianary-proof-the-recurrent-laryng...

Interesting

I did not know human urinary tract was suboptimal.

Are there examples of "discontinuous" evolution where a chance mutation stumbled upon a much better way to do something that allowed a species to jump the gap?

I imagine it would be very rare because chance mutations in a single individual that confer extraordinary survival / mating advantage to offset entire population would be rare like being a superman among men.

"I imagine it would be very rare because chance mutations in a single individual that confer extraordinary survival / mating advantage to offset entire population would be rare like being a superman among men."

-> This also depends on population size, if there is a bottleneck event (ie 99.99 percent of some species dying) then among the 100's of survivors perhaps there could be a superman. Either because this person got lucky, or the rest got unlucky. But in general the distribution of traits is quite continuous, given large populations.

It's also good to keep in mind that us biologists have messy definitions, ie we call specimen A and B the same species when they produce fertile offspring. Some species produce offspring but infertile (Horse-Donkey, Lion-Tiger). But what if A and B have fertile offspring and B and C do, but A and C do not? The whole definition of species falls apart (it leads to a paradox).

We have a lot of that in biology. You need to learn to deal with it (sunglasses dropping down), and assign uncertainties to terms you learn about, they are often just our attempts to name boxes on continuous, multidimensional scales.

Well, what is suboptimal... One would engineer it differently, it makes a weird, seemingly unnecessary, loop around some other tracts. However, being a biologist I can also imagine that if we'd reroute the thing in someone using surgery, that person suddenly gets more infections because bacteria find their way up the tract more easily (no idea, but could be!). The point is... It's a mess based on some earlier mess, but it works, and changing one part in this mess can lead to unpredictable results.

As for "discontinuous evolution"... Again, what is discontinuous? I'm sure there are events to be found of which the probability of them occurring is so small and the effects so large that they'd fall into someone's definition of "discontinuous". But I think there is by definition no such thing as discontinuous evolution. If it occurred, it had a non-0 probability of occurring, then again, maybe the universe is infinite and everything has a non-zero probability of occurring. This is getting philosophical haha. Maybe listen to some Mindscape podcasts (Sean Caroll).

Actually rabbits eat their excrement again. That's also probably not how an engineer would make it (or an intelligent designer would design it, since we seems to be going in that direction ;) ).

Even in an infinite universe there's no requirement for all events to have non-zero probability of occurring. A simple analogy: there are infinitely many integers, but even if you could pick between them all uniformly at random infinitely many times you'd never pick π. Not all events even have to be in the distribution of possible events.

That's often not going to be relevant for evolution, since the distribution covers physically realizable events like adding DNA base pairs.

Antibiotic resistance happens all the time (at least in our artificial world) and it's a life or death situation for the bacteria.

But you were probably asking for multicellular eukaryote organism; you may be interested in the concept of saltation https://en.wikipedia.org/wiki/Saltation_(biology) . Behind the concept of the thing itself it's interesting to think about the mechanism behind it. Regular evolution is gradual (point mutation leads to a maybe slightly different protein, which has only a relatively small scale effect but you gain or lose function over many generation), but for a saltatory evolution you most probably need radical changes in the DNA, like duplication (of a single gene, multiple gene or whole chromosome), horizontal exchanges, etc.

It's possible in plant but in animal more often than not it's not beneficial at all to have such event. Famously if you have an extra chromosome nr. 21 you get Down syndrome.

no offense, but it just sounds like you don't understand local maxima in high dimensional spaces.

a suboptimal trait evolving that still provided an overall population benefit (via survival or reproduction) with further improvement close to impossible without a multitude of large changes is literally just a local maximum in that high dimensional space that maps traits to whatever success metric you're measuring.

the infinite wall you describe is not a wall in that space. actually I originally wrote its more like a wall in the derivative space but its not quite right either, those just point straight down from the maximum.

Perhaps not no, please enlighten me, where is my lack of insight?

Edit, thanx for the addition. Perhaps my issues with looking at it as "local maxima in high dimensional spaces" (even though we think of minima usually, as in a marble will find the minimum in a landscape with gravity pointing down, sorry I'm a biologist) is that with every mutation and every change in environment the landscape changes (so basically with time), is it still a useful way to think about it like that, at that point? So it not a "landscape in a high dimensional space", but rather a "high dimensional dynamic surface", like a sea?

... In fact, because te landscape is so dynamic and can sometimes change within 1 to a few generations, having a brain as the primary store for behavior/customs/models/information is super advantageous compared to using DNA, as the brain is a flexible neural network and can adapt much, much quicker. No doubt this is why we are successful (ie, a dung beetle will die trying to push a stuck piece of dung, because the behavior is hard-coded and it cannot model the world and model solutions to encountered problems, and it therefore cannot pas solutions on to future generations, as we can. Perhaps a dung beetle will someday mutate to alter it's behavior when stuck, but you can see how suboptimal the whole process.)

Great question! It maybe can be, but what is the maxima that is being solved for exactly? There are usually tradeoffs, the most common and prevalent is our body's immune response. If it's under-active, we might die from an infection. If it's over-active, we get diabetes, lupus, chron's, and (probably) lots of mental health disorders.

So, it's a bit context dependent, and definitely dependent on other mutations that have occurred in other parts of your DNA. Most complex diseases are "polygenic", meaning it's a culmination of quite a few factors that would contribute to a specific good or bad outcome.

So, yes, it could be modeled as a sort very context dependent with a lot of highly correlated non-independent covariates. We do use quite a lot of statistical and ML methods to understand the genome (I work in statistical genetics), but the complexity of biology has so far proved a tough nut to crack.

> It maybe can be, but what is the maxima that is being solved for exactly?

Survival rate/reproduction rate/rate of genes being spread?

I think you're correct, but only if you add "for a given environment". It's good to develop the ability to store super efficiently fat when you're a Pacific Islander and food is scarce, but this advantageous trait becomes a liability with modern and plentiful (junk) food.

see thrify gene hypothesis; I know it's not a good explanatory theory but it's a good illustration of my point

Sure, the form of the function being maximized will usually depend on the environment.

That was kind of my initial point... there are far too many permutations that alter what you want to maximise to consider this as a good option right now. Environment, gene-environment interactions, epigenetics, polygenecity, pleiotropy, etc, etc.

Great answer! Thank you!

Evolution is the ultimate "good enough". Once the required threshold is passed, pressure stops. The threshold might constantly be moving due to competition, environment etc.

It can't be a global one because it is inherently local. Evolution is a process which determines the next time step (generation) from only the previous generation. It has no memory at all nor "knowledge" of other reproducing pairs.

Reply on news.ycombinator.com