Genetic Bottleneck or Evolutionary Breakthrough

Once again, I’m going to start with a disclaimer.  I’m not an anthropologist, I’m a computer geek.  Today I’ll be blogging about anthropology because I see in the population dynamics of my genetic algorithms a phenomenon that reminds me of something from anthropological history.

In other words, this happens in Genetic Algorithms, but to the extent that I talk about actual anthropology, I may be wrong.

A thing that happens in an artificial population of data genomes is that when a successful combination, or a successful mutation, emerges, it tends to sweep through the population at an exponential rate that depends on how advantageous it is relative to the average. What we hope for with genetic algorithms is for a fairly gradual expansion, where advantageous traits arising separately have many generations to find ways to get incrementally combined and genetic diversity remains high for as long as any variation in the genome can be exploited.  But if a trait is very highly advantageous, it can sweep through the population so fast that almost every genome not bearing it gets wiped out, and other advantageous traits are more likely to be lost entirely, with a drastic reduction in genetic diversity, than to be combined with the new one.

This is a mixed blessing at best.  The new trait represents a leap in fitness, but the reduction in genetic diversity usually means a long ‘plateau’ where fitness does not improve further until mutation very slowly starts reintroducing diversity, wherever it can be found in parts of the genome that don’t destroy existing advantages.

We twiddle parameters a lot to try to find the proper rates for these things.  The more well-known parameter is called the mutation rate (Or mutation rates plural, if you’re using several different rates to control different areas of the genome). It introduces genetic diversity, enabling progress to resume after those plateaus. But it limits the complexity of discovered solutions because, If it happens at too high a rate it can destabilize existing advantages and destroy progress.

The less well-known trait is called the exclusivity rate, and it controls the rate of expansion with which to reward a fitness advantage.  Exclusivity that’s 100%, where the most fit individual always gets the breeding opportunity, preserves absolutely nothing except the single most advantageous trait available at a given moment.  It means genetic diversity is always wiped out immediately when anything advantageous emerges. Nothing else is preserved, and the GA gets no traction from combining anything. This is equivalent to hill climbing; it can be faster than GA for small, simple problems, but doesn’t scale as well.

Exclusivity of 0% on the other hand, where the most fit individual has no reproductive advantage, of course doesn’t even climb any of the hills gradually, and advantageous traits are as likely to be preserved, or combine, or disappear, as any other traits.  So here is the conundrum; you estimate the scale and difficulty of your problem.  The bigger and more complicated it is, the lower the mutation and exclusivity rates you set.  Lower rates of both make preserving genetic material that may be combined in advantageous ways more likely – mutation by not altering it, and exclusivity by not replacing it.  But lower rates of both make progress slow.

You can have a higher rate of mutation with high exclusivity; you produce a lot of crap genomes, but are very reliable about throwing them away instead of preserving them.  That’s the hill climbing strategy.  With a low rate of mutation, you need either low exclusivity, or you have to limit the spread of advantageous genomes by other means.  Otherwise any new advantage discovered means quickly wiping out genetic diversity.

And every once in a while, if you set the exclusivity high enough to make progress at all, a trait will be discovered that is so extraordinarily advantageous that it will sweep through the population destroying genetic diversity and replacing the entire population with descendants of the very few individuals that got combined with it in the earliest generations. After that any outbreeding will put the offspring at a disadvantage relative to the inbred, and the GA will cull them.

Mama nature set her rates of mutation very low but compensated with huge populations (she can afford it; there’s a lot of parallel processing available with an Earth-sized CPU). She set her exclusivity rate fairly high, but compensated by limiting the rates of genetic destruction with geographic distance.  Nevertheless, every once in a while something so tremendously advantageous will come along that it sweeps across the available population wiping out genetic diversity.  And that brings us to my speculation about anthropology.

One of the stories that one occasionally reads when there’s a slow news day and they’re looking for filler, is about how, according to the genetics of modern populations, it seems likely that the entirety of modern humanity is derived from ~2000 or so breeding pairs who lived through a period of critically low population levels, at around the year -68000.

This is usually presented as the ‘Toba Catastrophe Theory’, on the grounds that the crisis may have been precipitated by the eruption of the Toba Supervolcano in Indonesia.  This happened in about the year -73000, and was definitely a significant crisis for terrestrial life, just as a Yellowstone eruption would be today.  It’s estimated to have coughed up some 2500 cubic kilometers of ejecta, and laid down volcanic ash about 15 cm deep over pretty much all of Europe and most of Asia.  It accounts for a significant deposit of Iridium (more common in volcanic ash than in other soil) at about the same strata worldwide.  Pollen analysis and the presence of secondary ash deposits indicates that there was widespread deforestation and that most forests burned in southeast Asia. It’s likely to have destabilized the climate pretty dramatically worldwide with a ‘Winter’ caused by stratospheric volcanic ash blocking out sunlight, lasting 6 or 8 years, followed by a century of reduced sunlight as the last of the ash filtered out of the atmosphere, altered weather patterns lasting at least hundreds of years, and may even have triggered a ten-century cooling period sometimes called a mini-ice age.

But there’s a lot of controversy about this.  There are a lot of places where the same kind of stone tools are found above and below the ash strata, indicating local populations that weren’t wiped out by the dust clouds and weather changes.  And there is a fair amount of evidence that a lot of populations survived because the weather patterns just plain didn’t subject them to the worst of the ash and fallout and deforestation effects.

The possibility I’m blogging about in this post is that this critically low population level may not in fact have been a population crisis as we understand it.  It may not have been so much a ‘chokepoint’ where humans almost went extinct and were cut back to a dangerously small population, as it was a speciation event where a particularly beneficial trait arose within a small subpopulation.

What I’m talking about is the emergence of what we have called ‘Cro-Magnon Man’, which is a term for the earliest examples of modern humans or Homo Sapiens Sapiens.  Cro-Magnon man is widely considered to have arisen around the year -48000, which is significantly after the genetic bottleneck, but which may be consistent with the gap between the emergence of the very first tiny population of Homo Sapiens from which we are all descended, and the establishment of the species in sufficiently wide range and large numbers that it becomes visible in the fossil record.

It seems plausible to me that human (for pre-modern values of ‘human’) populations survived the Toba event worldwide, but that at around this time, in just one small population of humans, the modern traits we associate with true Homo Sapiens Sapiens arose.  The reason we don’t see evidence for a larger population at that time in our genetics would be because we are descended, almost exclusively, from that small population. If they followed the patterns I observe in GA when highly advantageous traits develop and exclusivity is quite high, then they wound up with a large evolutionary advantage and spread without much outbreeding. The descendants of outbreeding were at a disadvantage relative to the descendants of inbreeding, and their offspring did not survive as reliably.  So the traits in the genetic pool from outside that small population did not have a chance to get mixed with those of the expanding population and its new traits, and we don’t see them (or at any rate not much of them) in humanity at the present time.

We have evidence of some outbreeding, some admixture of Neanderthal genes with Cro-Magnon genes, here and there in the gene pool – and to the extent that it conferred advantages, it became part of the modern species.  But except for the very best of such traits, or those that readily combined without loss to the advantages of either side, such heritage is more likely to’ve been a hindrance than a help, and identifiable Neanderthal genetics are correspondingly rare.

It’s tempting to speculate that a tremendously advantageous trait that emerged at that time has something to do with the way we use our brains, what kind of intelligence we have, the emergence of language, or some other pet theory.  But  my own speculation should be understood to be biased.  I’m tempted to speculate about such things specifically because I spend almost all my time thinking about them anyway.

The fact is we have evidence that modern humans are descended from a small population that existed around the year -68000.  We don’t have any really good evidence of why.

All attribution of reasons is somewhat speculative, and the only thing I can really point out is that the emergence of a hugely advantageous trait – without regard to what trait specifically – explains the currently known genetic evidence as well as the usually postulated global crisis in which the whole population was cut down to very few individuals.  Even this understanding is suspect; it is based on my observation of population dynamics in GA populations implemented in bits, which are not really the same as population dynamics in living populations implemented in DNA.

So – once again.  I’m not an anthropologist, I’m an AI geek.  I don’t have evidence about what happened to humans.  I have information about what happened to humans from pop-science articles, and a pet theory.  My theory is informed by population dynamics from purely artificial simulations that aren’t even attempting to model DNA genetics, and probably qualifies, at best, as more of my ‘Mad Science.’

But I think it’s interesting and plausible.

Leave a Reply