The wonder of evolution

Not arguing about that at all, since these kind of survival characteristics are very useful in helping individuals survive until they can reproduce - since, unlike tribbles, we aren’t born pregnant.
But notice how individuals beyond the age of reproduction start losing these survival characteristics. (I’m in that class, I know first hand!) We lose eyesight, we lose our ability to run and react quickly, etc. And to a certain extent there is a tradeoff where older humans are helpful in teaching younger ones, but harmful in using resources that those of reproductive age can make use of. Like my social security check.
Humans and other animals are programmed to die, not so all plants and fungus. I find this hard to explain except in the context that individuals dying contribute to the success of the species as a whole.

That’s perhaps an example of evolution’s lack of optimization. We could easily be as well off – and indeed I would argue much better off – if the whole concept of aging and death were replaced by indefinite regeneration at the cellular level instead of deterioration. This would have to be balanced by a very low reproduction rate, only enough to replace those lost through factors other than aging.

This is veering off topic into sci-fi, but the obvious win here is the endless incremental growth of knowledge we’d have in the elders. Every time an accomplished individual dies, it’s an almost incalculable loss of a lifetime of learning. What if Isaac Newton was still with us, still learning, still contributing? Or Einstein? Or Feynman? Or, for that matter, Shakespeare?

Mother Nature has never thought in those terms, the bitch. Her evolutionary plan seems to be based on much more primitive principles. :wink:

That is a common justification and has been rationalized in both gene centric view of evolution and in the context of kin selection as a game theory-based rationalization but I have yet to really see a defined mechanism that would actually cause senescence. I tend to subscribe more to the hypothesis of antagonistic pleiotropy; animals so so complex and their phenotypical expression is so diverse that it isn’t surprising that adverse characteristics crop up during ‘extended development’ (e.g. middle and later age) but aren’t selected out because of the limited fertility at that stage doesn’t really contribute to the collective genome anyway. For mammals, age tends to correlate inversely to average heartbeat but nobody really knows why or what it means.


Yeah. It would be interesting to see what happened with really large lists, for which the differences between bubble sort and faster alternatives mattered.

There’s a theorem that shows that in the most general case of all possible search spaces, no particular approach is better than any other in finding the best answer.

In search space the living creatures live in, though, genetic algorithms work better than at least some (if not all) other approaches

We know plenty of mechanisms for aging, but whether these are selected for or rather not selected against is the issue. That dying is pretty much universal and that nothing has evolved immortality- or re-evolved it since I think single cell creatures are pretty much immortal - would seem to argue it is disadvantageous. But that’s just a guess.

Heuristics have fads like everything else. Plus there is a bonus for novelty - better bees than the nth simulated annealing method.
In my field the reason that GAs are not used in real life is that they are slower and less reliable than methods employing algorithms with some heuristics. I nature if there were a master programmer, we’d probably have evolved in a hundred million years tops. Or six days like the fairy tale says.

Hydra magnipapillata isn’t immortal, but I’ll take an average of 1,400 years as a fair lifespan! It has amazing (though not complete) regenerative capabilities.

Yeah, we understand various biophysical mechanisms of senescence; I meant that we don’t have a evolutionary mechanism that would select for aging. Parents dying in defense of offspring is relatively easy to demonstrate, as are the ultimate advantage they gain by spending effort to rear offspring versus being more proliferate and just reproducing as often as physically possible. But the notion of dying off to make way for their offspring as a way of ensuring success of their genome, while perhaps philosophically appealing, just lacks any cause and effect that one could tie to an adaptive change. If some hypothetical creature were consuming enough resources to threaten the survival of its offspring it would also being consuming enough to threaten its non-offspring competition.

Animals, however, are really complex and go through various non-repeating phases of development through life that are more clearly defined that plants or fungi, and so it isn’t terribly surprising that a cumulation of developmental adaptations that are beneficial in early stages before or during prime reproductive phases might be detrimental in the post-reproductive stage of the animal. By analogy (always a slippery way of reasoning, I know) your car has a break-in phase where moving parts in the engine and transmission wear off roughness and flashing, then the car runs smoothly for the middle portion of the life, but eventually powertrain and other components build up enough cumulative damage that they start to malfunction and break down. Of course, a car or other machine is not self-repairing nor is it a product of evolution, so just take that as an illustration, not a demonstration of any actual evolutionary mechanism.


If death evolved it evolved long before parents took care of offspring, so that couldn’t be involved. I suspect the reason is that in times of scarce resources the death of individuals who can’t reproduce any more would be preferable to the death of those who can, since they are competing for the same resources. On a species level it wouldn’t matter how the individuals are related.

Known as the bathtub curve. I’m quite familiar with this in semiconductors. If high reliability is needed, we burn in parts to move them along the curve and force early life failures before the part is shipped.
However, it is not always reliability. I have a paper that shows that much early life failure comes from a part being in a new environment, and therefore gets tested in a way it can’t be in the factory. For a car it would like some having defects where something breaks when it goes over a bump. Since factory test drives don’t go over bumps, some percentage of cars will fail in the field. That’s not a reliability issue, it is a lack of testing. Just putting the example back to the car analogy.
Our bodies have all sorts of features for enhancing reliability, and some species can grow new parts. I suspect evolution would have found a way somehow for at least some species if this did not cause a reproductive didadvantage.

The difficulty with this line of reasoning is that it implies that something is making a rational decision about why one organism to die to allow others to flourish. But three is no mechanism within modern evolutionary synthesis or any other accepted framework of evolution that really produces a convincing thesis for what would actually drive this. It would seem that a successful organism would survive and reproduce indefinitely until overcome by a more successful organism; of course, this might cause the problem that potentially successful lines would be outcompeted by inferior but already proliferate ones, just as a large manufacturer might squeeze an upstart with an innovative product by dint of just dominating the market and undercutting profitability. And in fact individual human and animal cell lines can continue growing and dividing indefinitely, and some multicellular lifeforms can grow or divide indefinitely, either being effectively immortal or at least for which ‘death’ of an individual has no particular meaning.

So death isn’t just a theological conundrum, but also one within the philosophy of evolutionary biology. There is probably a good and really obvious answer to this that nobody has quite expressed; for complex animals, it may be that a collection of mutations just makes it progressively more difficult to regenerate, or again that we just continue to develop into post-reproductive stages for which evolution has not pressured for useful repair and survival, but there isn’t any kind of consensus or even a falsifiable hypothesis for why we die.

Sure, once we are post-reproductive, the only value that is provided is in the social and familial work that grandparents (or childless aunts and uncles) provide to ensure viability of still develoing children, and at some point the value of that diminishes as bodies become less agile and minds less flexible, at which point there is no evolutionary pressure to select for some kind of continued reliability. But that begs the question that considering the ability of biological systems to continuously repair and modify themselves according to a developmental plan, why do they just start giving up at some point? When children or even young adults twist an ankle or break an arm, they heal in a few days because the body is easily able to make repairs, and it is unclear why that starts to break down over time. Is it something to do with genetic errors building up in mitochondria, and if so why is it so consistent throughout the body and across individuals? Nutrition and general fitness certainly play a role but even the fittest person in the world is still only going to live to maybe a couple of dimes past a century, and even that is mostly luck of the draw.


I might have fallen into the common trap of seeming to impute agency to any of this, but I certainly didn’t intend to. Mutations that cause more deaths for older organisms leading to increased reproductive success of younger organisms has nothing to do with rational decisions.
As anyone who has had any exposure to reliability knows, the more components the more opportunity for failure, not counting components meant to deal with failures. However as you mentioned larger organisms with lower heart rates seem to have longer lives. We certainly don’t know the answer, but it doesn’t seem like a total mystery.
On the other hand, if genes are really controlling everything, they’re basically immortal and don’t give a crap about the body that carries them - so long as it reproduces and spreads them further. (And any intentionality in that sentence is a metaphor, not a hypothesis.)

I found the best way of explaining evolution is the concept of the ‘adjacent possible’.

Imagine you are an organism that sits and waits for food to come by. Your ‘adjacent possible’ is all the food near you that might come your way. But then a random mutation causes you to be able to move a bit. Your adjacent possible just expanded. When food gets scarce, all the organisms that can 't move, die. So your mutation survives. Now all your species can move.

This now creates a new adjacent possible for other creatures. For example, a parasite that rides on you might have an advantage. So they expand into that adjacent possible. But now in the new world of moving creatures carrying around parasites, new adjacent posdibles open for other creatures. And so it goes. This is why Darwin, when he found an orchid with a 30cm deep nectary, knew there would be a moth around with a 30cm proboscis. 150 years later, tye moth was discovered. It didn’t just randomly evolve a long proboscis thatnhappened to match the flower - they evolved together a bit at a time.

So evolution isn’t random: it’s a random walk through the ever-growing adjacent possible.

Human social and technological evolution does the same thing. When we learned to control fire, new adjacent possibles opened for different ways to organize society. We tried a bunch that failed, and found some winning combinations. That organization allowed us to build resources, opening new adjacent possibles.

Complex evolved systems can have crazy relationships you would never expect because they evolve along woth the rest of the ecosystem into new adjacent possibles. For example, there is a parasite that has a lifecycle where it gets into an ant, and changes the ant’s brain such that it climbs the nearest stalk of grass and waits to be eaten by a bird. The bird then craps out the parasite some distance away, allowing it to spread.

The parasite doesn’t kmow birds exist. It doesn’t kmow anything. It just evolved its way through the adjacent possible to a stable reproduction/population spread technoque.

When you look at evolution as a walk through the adjacent possible, the number of trials needed goes way down because at every step of evolution there are limited successful adaptations.

Back to the watch analogy. Imagine if every part only fit in exactly one way - i.e. each gear only fits onto the watch after the previous parts have found their way together. So when you first shake the bag of loose parts, they’ll try to go together randomly but will only succeed if part 1 randomly attaches to part 2. Now that they are together, a new adjacent possible is opened for part three, which can now attach to part 2. Keep going until the watch is assembled.

It would still take a hell of a long time, but if you shake long enough, eventually you’ll have a watch. The analogy is flawed in that evolution isn’t trying for some specific outcome, but it illustrates how the restricted nature of random selection at each stage of evolution makes evolution work.

I’d count them. Components meant to deal with failures can induce failures. That’s one of the things that makes reliability engineering so difficult.

Agreed. We now have software that can automate tests for other software. The problem is that the test software can also have bugs.

Of course they can fail, but there should be a net gain in reliability, which is why I excluded them.

Dave Patterson (the popularizer or RISC) did his dissertation on the verification of microcode, and I saw his talk at the Micro-8 workshop. He verified low level microcode against a higher level description. He found bugs in the microcode, the description, and his verifier. So the test cases can have bugs too.

Well you’d think so, But it isn’t always the case.

For example, in 1966 a Soyuz launch escape system failed and fired prematurely on the pad, ruining the mission.

RAID systems, if not properly designed and deployed can actually reduce system reliablility.

Software exception handling can be both a source of bugs and a vector for system attacks.

Software and hardware designed to prevent accidents can sometimes cause them. See the 737 MAX’s MCAS system.

Failover systems, when done inocrrectly can add a single point of failure to a redundant aystem.

And so it goes. When designing for maximum availabilty and safety, you have to consider every part of the system - including the stuff added to improve reliability and safety.

Then there’s the human factors. Design something to be toleeant of human error, and over time you’ll get more human error because the behaviour is unpunished. Put airbags in cars, and people drive more recklessly. Put in emergency auto-braking or lane holding software, and drivers will start to lose situational awareness or become more distracted, counting on the system to protect them. And so it goes.

Safety and reliability engineering is full of ‘wicked’ problems where adding something to fix a problem in one dimension introduces problems on others.

But this is really a hijack of the thread. If you want to keep talking about it, we coild open another…

Yeah, I get what you are saying. It just becomes a delicate semantic dance about intentionality. Evolutionary biologists and zoologists often talk metaphorically in terms of intentionality even though they understand that “genes” don’t actually control things, and the result of favorable selection is a statistical effect, not a deliberate cause. Richard Dawkins has received no small amount of criticism over the years for the title of The Selfish Gene, as well as misunderstanding by people insisting that gene-centric theory is a justification for Social Darwinist-type thinking, when Dawkins actually doesn’t suggest such a notion at all.

It still leaves the question of why animals (and plants, most other living organisms) die, when there is theoretically no reason absent of trauma, adverse environment, or lack of nutrients, that they shouldn’t be able to continue on forever with active repair and error-correction. And the answer is that nobody really knows, although it seems that aging and death is built directly into the genome and epigenome, to the point that an ‘aging clock’ has been identified in neural function that quite precisely correlates to somatic age. We’re programmed to die, and it is probably a good thing given how ossified and reactionary the thinking of many people becomes with age, but the specific mechanisms that cause that are as yet unexplained to any accepted degree.

The launch escape system isn’t a reliability feature, it is a safety/crew recovery system. Such systems, because they are not part of nominal function, can certainly compromise overall system reliability by functioning when they shouldn’t, or failing to function correctly when they are activated exacerbating a prloblem. With such systems (like airbags in cars) there is a tension between whether the system should “fail-safe” (compromise reliability in favor of not functioning) or “fail-deadly” (in favor of functioning even when not activated, or function outside of parameters).


Reliability in this sense depends on goals and context. From the standpoint of someone who wants to stay alive, an airbag or an escape system is more of an emergency failover, just as in the context of someone who doesn’t want to lose data a RAID array is an emergency failover. It’s all in how you look at it.

The principle remains: It’s possible to introduce new failure modes while attempting to engineer more reliability into the system. I could probably come up with examples all day.

A good example is exception handling, which is designed to make software more robust and to fail in predictable ways or recover gracefully from exceptional conditions when something is wrong. Bad exception handling code is a common source of bugs, and a common security risk. Bad exception handlers can give the user stack traces, or other information that gives away stuff about the system that can be used against it, or swallow errors that should shut down the system to prevent damage.

Good exception handling can make a system more secure, but screw it up and you make it more vulnerable than if you had done nothing at all.

How is a species supposed to evolve if individuals don’t die? If one species is eternal and a competitor isn’t, won’t the competitor have many more generations in the same amount of time, allowing it to eventually find ways to out-compete the other? I guess you could say that maybe a species just keeps reproducing even though its immortal, but then you would rapidly run into overcrowding.

Death seems like a very necessary part of evolution to me. And the fact that so far as we know it is universal means to me that it’s fundamental to our existence.