4 ms·

Mathematics Shows How to Ensure Evolution

85 points by digital55 8y ago

alexbeloi 8y ago
These seems like a interesting model for analyzing which population structures are most conducive to the spread of memes or ideas. The obvious population structures are 'highly connected' but this suggests that you can have loosly structured populations as long as 'have certain features (sic), such as hubs and self-loops.'
- OnesimusUnbound 8y ago
  Hmm, for an example, in twitter, the popular accounts are one of the strongest "hubs". Their post can influence many "nodes". I'm not sure how I can relate "self-loops" in my example.
  - sleepyams 8y ago
    Perhaps a self-loop would represent time spent curating one's own profile (or crafting tweets) as opposed to consuming others?
  - amelius 8y ago
    Echo chambers?
- 7373737373 8y ago
  You might like this visual explanation by Nicky Case: https://ncase.me/crowds/ https://ncase.me/crowds/
  - alexbeloi 8y ago
    Very cool, I'm a fan of his stuff and hadn't seen that one yet. Thanks
smrk007 8y ago
Does anyone know if there is work similar to this being applied to machine learning? It seems like there could be a lot of potential.
- bdamm 8y ago
  That was my first thought too. Could it be applied to genetic algorithms? And from my quick review of the article, it seems definitely a yes. Genetic algorithms are asexual, so that's a check. The competitions just need to occur in pools with migration between the pools satisfying an amplifier pattern. Pretty exciting result, and I can't help but wonder if it will find its way into a number of fields.
  - im3w1l 8y ago
    There are both sexual and asexual genetic algorithms and people have experimented with constrained breeding.
carapace 8y ago
One interesting thing about evolution is that meta-evolution is evolution. In other words, we see "Darwinian" as opposed to "Lamarckian" evolution not because the latter is impossible, but because there is adaptive advantage to the former. Gregory Bateson works this out from cybernetic first principles in one of his books. "Steps to an Ecology of Mind" I think. An evolving system of either kind will evolve to evince mostly Darwinian evolution. Another way to understand this is to consider that you cannot distinguish "Darwinian" vs. "Lamarckian" in the evolution of single-celled microbes. Microbes don't grow old and die. Each Amoeba is billions of years old. (HeLa cells are called immortal.[1]) A change in the cytoplasm of an ancestor cell would be carried into both daughter cells. In "Selfish Gene" Dawkins works out how death and birth have adaptive value for genes in multicellular species. The meta-evolutionary "choice" to reproduce by sex and children (rather than, say, budding, or metamorphosis, or some kind of mass mitosis) is "made" by the normal process of evolution, not some additional meta-process. In short, there is no a priori reason to believe that the life of Earth has not evolved intelligence. (After all, we are intelligent.) What I mean is, the machinery of evolution itself will have evolved to be able to direct its own evolution, IF intelligently self-directed evolution is adaptively valuable. If you want to prove that Nature is not intelligent you have to show that there is adaptive value to remaining "dumb". Given what we now know about adaptive systems, that's the only way Nature would have evolved without ambient intelligence: if it pays to act dumb. [1] https://en.wikipedia.org/wiki/HeLa https://en.wikipedia.org/wiki/HeLa - - - - - edit to add: Trawling through the new links I see: https://news.ycombinator.com/item?id=17405096 https://news.ycombinator.com/item?id=17405096 > "Cells Talk and Help One Another via Tiny Tube Networks Long-overlooked “tunneling nanotubes” and other bridges between cells act as conduits for sharing RNA, proteins or even whole organelles " That's teh sort of thing I'm talking about: Life is one thing.
- olliej 8y ago
  What are you talking about - no amoeba is billions of treats old, not are hela cells immortal (the immortality of hela cells is a reference to it being the first set of human cells that could be reliably grown outside of a human. More specifically it is Henrietta lacks who is considered immortal). All unicellular life survives by growing (accumulating enough of the right chemical components) to replicate its genetic material (RNA or DNA), and then divide. Errors happen during duplication that provide mutations that natural selection can work on. In the general act of existing all these organism accrue broken proteins and waste that they cannot get rid of, these accumulate in one half of a dividing cell, which eventually results in one half of the separated cell failing. In addition to replication errors environmental stresses can also just directly damage the genetic material which again provides mutations than selection can be applied to. So yes we can distinguish between Darwinian selection and Lamarckism, as Lamarckism is not actually a thing - the evolution of single felled organisms is purely Darwinian, and can (and has) been extensively studied. I suspect the misunderstanding here is that meiosis (or is it meitosis? I can never remember which is which) is only notionally a clone. Once you acknowledge that one of the two cells after a split is not necessarily identical to the original you can see a lineage of genetic variation that is the hallmark of evolution. This is ignoring lateral gene transfer, but that’s essentially an effect of gene survival and statistics.
im3w1l 8y ago
They way you get rid of cancer is an evolutionary bottleneck. Distill a whole animal into a single sex cell that is then cloned and cloned again to make a whole infant. This forces the cell to follow the Categorical Imperative: however it acts will be universal law. Heads and that cell is healthy and the healthy cells win. Tails and that cell has cancer and is stillborn and cancer loses.
- civilian 8y ago
  I think your understanding of cancer is a little flawed. It's not just "the organism will get cancer" or not, it's horribly complex. Different cancers develop from different differentiated cells, and often it's a combination of genetics, environment and chance that causes cancer. It's awesome that you're thinking about this stuff though. I think you'd get a lot out of this: https://www.khanacademy.org/science/high-school-biology/hs-reproduction-and-cell-division/hs-the-cell-cycle-and-mitosis/v/cancer https://www.khanacademy.org/science/high-school-biology/hs-r...
  - im3w1l 8y ago
    The way humanity keeps its DNA clear of cancer over the long term is because people with cancer die and healthy people don't. A slow evolutionary dynamic keeping a fast evolutionary dynamic in check. This shouldn't be controversial. The mechanics of it, is like I said, the population bottleneck. Cancer cells are parasites and the bottleneck leaves them without a host. A fascinating cancer is "Devil facial tumour disease". It has learned to spread from animal to animal, thus bypassing this purifying mechanism. But as it no longer consists of the animals' own cells it may be argued that it has become its own species and should no longer be considered a cancer. Supporting the speciation argument is the fact that the cancer cells even have a different amount of chromosomes from the animals.
    - civilian 8y ago
      But cancer isn't a parasite. It's a flaw in an information system. We have several gates, oncogenes, that protect us against cancer forming. But then those genes take damage, then cancer can form. This isn't a determinist "cancer will form or it won't" situation-- there's a lot of variance and luck. Furthermore many cancers form later in life, after people have already reproduced, so natural selection doesn't have that strong of a hold on it. If cancer worked the way you think it did, don't you think that cancer would have been naturally eradicated during our evolution?
rdlecler1 8y ago
This is not unlike Peter Thiel’s Small-markets strategy. No one cares, you can monopolize it and then grow at the margin into new markets.

85 points by digital55 8y ago

These seems like a interesting model for analyzing which population structures are most conducive to the spread of memes or ideas. The obvious population structures are 'highly connected' but this suggests that you can have loosly structured populations as long as 'have certain features (sic), such as hubs and self-loops.'

OnesimusUnbound 8y ago
Hmm, for an example, in twitter, the popular accounts are one of the strongest "hubs". Their post can influence many "nodes". I'm not sure how I can relate "self-loops" in my example.
- sleepyams 8y ago
  Perhaps a self-loop would represent time spent curating one's own profile (or crafting tweets) as opposed to consuming others?
- amelius 8y ago
  Echo chambers?
7373737373 8y ago
You might like this visual explanation by Nicky Case: https://ncase.me/crowds/ https://ncase.me/crowds/
- alexbeloi 8y ago
  Very cool, I'm a fan of his stuff and hadn't seen that one yet. Thanks

smrk007 8y ago

Does anyone know if there is work similar to this being applied to machine learning? It seems like there could be a lot of potential.

bdamm 8y ago
That was my first thought too. Could it be applied to genetic algorithms? And from my quick review of the article, it seems definitely a yes. Genetic algorithms are asexual, so that's a check. The competitions just need to occur in pools with migration between the pools satisfying an amplifier pattern. Pretty exciting result, and I can't help but wonder if it will find its way into a number of fields.
- im3w1l 8y ago
  There are both sexual and asexual genetic algorithms and people have experimented with constrained breeding.

carapace 8y ago

One interesting thing about evolution is that meta-evolution is evolution. In other words, we see "Darwinian" as opposed to "Lamarckian" evolution not because the latter is impossible, but because there is adaptive advantage to the former. Gregory Bateson works this out from cybernetic first principles in one of his books. "Steps to an Ecology of Mind" I think. An evolving system of either kind will evolve to evince mostly Darwinian evolution. Another way to understand this is to consider that you cannot distinguish "Darwinian" vs. "Lamarckian" in the evolution of single-celled microbes. Microbes don't grow old and die. Each Amoeba is billions of years old. (HeLa cells are called immortal.[1]) A change in the cytoplasm of an ancestor cell would be carried into both daughter cells. In "Selfish Gene" Dawkins works out how death and birth have adaptive value for genes in multicellular species. The meta-evolutionary "choice" to reproduce by sex and children (rather than, say, budding, or metamorphosis, or some kind of mass mitosis) is "made" by the normal process of evolution, not some additional meta-process. In short, there is no a priori reason to believe that the life of Earth has not evolved intelligence. (After all, we are intelligent.) What I mean is, the machinery of evolution itself will have evolved to be able to direct its own evolution, IF intelligently self-directed evolution is adaptively valuable. If you want to prove that Nature is not intelligent you have to show that there is adaptive value to remaining "dumb". Given what we now know about adaptive systems, that's the only way Nature would have evolved without ambient intelligence: if it pays to act dumb. [1] https://en.wikipedia.org/wiki/HeLa https://en.wikipedia.org/wiki/HeLa - - - - - edit to add: Trawling through the new links I see: https://news.ycombinator.com/item?id=17405096 https://news.ycombinator.com/item?id=17405096 > "Cells Talk and Help One Another via Tiny Tube Networks Long-overlooked “tunneling nanotubes” and other bridges between cells act as conduits for sharing RNA, proteins or even whole organelles " That's teh sort of thing I'm talking about: Life is one thing.

olliej 8y ago
What are you talking about - no amoeba is billions of treats old, not are hela cells immortal (the immortality of hela cells is a reference to it being the first set of human cells that could be reliably grown outside of a human. More specifically it is Henrietta lacks who is considered immortal). All unicellular life survives by growing (accumulating enough of the right chemical components) to replicate its genetic material (RNA or DNA), and then divide. Errors happen during duplication that provide mutations that natural selection can work on. In the general act of existing all these organism accrue broken proteins and waste that they cannot get rid of, these accumulate in one half of a dividing cell, which eventually results in one half of the separated cell failing. In addition to replication errors environmental stresses can also just directly damage the genetic material which again provides mutations than selection can be applied to. So yes we can distinguish between Darwinian selection and Lamarckism, as Lamarckism is not actually a thing - the evolution of single felled organisms is purely Darwinian, and can (and has) been extensively studied. I suspect the misunderstanding here is that meiosis (or is it meitosis? I can never remember which is which) is only notionally a clone. Once you acknowledge that one of the two cells after a split is not necessarily identical to the original you can see a lineage of genetic variation that is the hallmark of evolution. This is ignoring lateral gene transfer, but that’s essentially an effect of gene survival and statistics.

im3w1l 8y ago

They way you get rid of cancer is an evolutionary bottleneck. Distill a whole animal into a single sex cell that is then cloned and cloned again to make a whole infant. This forces the cell to follow the Categorical Imperative: however it acts will be universal law. Heads and that cell is healthy and the healthy cells win. Tails and that cell has cancer and is stillborn and cancer loses.

civilian 8y ago
I think your understanding of cancer is a little flawed. It's not just "the organism will get cancer" or not, it's horribly complex. Different cancers develop from different differentiated cells, and often it's a combination of genetics, environment and chance that causes cancer. It's awesome that you're thinking about this stuff though. I think you'd get a lot out of this: https://www.khanacademy.org/science/high-school-biology/hs-reproduction-and-cell-division/hs-the-cell-cycle-and-mitosis/v/cancer https://www.khanacademy.org/science/high-school-biology/hs-r...
- im3w1l 8y ago
  The way humanity keeps its DNA clear of cancer over the long term is because people with cancer die and healthy people don't. A slow evolutionary dynamic keeping a fast evolutionary dynamic in check. This shouldn't be controversial. The mechanics of it, is like I said, the population bottleneck. Cancer cells are parasites and the bottleneck leaves them without a host. A fascinating cancer is "Devil facial tumour disease". It has learned to spread from animal to animal, thus bypassing this purifying mechanism. But as it no longer consists of the animals' own cells it may be argued that it has become its own species and should no longer be considered a cancer. Supporting the speciation argument is the fact that the cancer cells even have a different amount of chromosomes from the animals.
  - civilian 8y ago
    But cancer isn't a parasite. It's a flaw in an information system. We have several gates, oncogenes, that protect us against cancer forming. But then those genes take damage, then cancer can form. This isn't a determinist "cancer will form or it won't" situation-- there's a lot of variance and luck. Furthermore many cancers form later in life, after people have already reproduced, so natural selection doesn't have that strong of a hold on it. If cancer worked the way you think it did, don't you think that cancer would have been naturally eradicated during our evolution?

rdlecler1 8y ago

This is not unlike Peter Thiel’s Small-markets strategy. No one cares, you can monopolize it and then grow at the margin into new markets.