Evolution vs Iterative Development

7 posts / 0 new
Last post
TBRMInsanity TBRMInsanity's picture
Evolution vs Iterative Development
I was thinking about AGIs and biological organisms (mainly transhumans and uplifts) and how each improve over time (AGIs through the use of iterative development from one version (the parent) to the next (the child), and biological organisms through evolution), and what advantages and disadvantages these development models give. Evolution: The core concept of evolution is that the most fit organisms are able to survive and reproduce. Their offspring are a mix of the parents genes, plus usually some mutations and recurrences of repressed genetic expressions. Two of the main driving forces behind this model is creativity (sometimes expressed as instinct) and randomness. The greatest strength of evolution is the introduction of children that have unique genetics and unique solutions to problems in any given environment. The greatest disadvantage is the length of time needed to create a child (as children need to build up necessary experience before they are competent). Iterative Development: The core concept of iterative development is that the deficiencies of previous generation are fix in the current generation. This means all children have all the strengths of their parents, while fewer of their weaknesses. The two driving forces behind this model is logic and order (as represented by an AGI's learning algorithms). The greatest strength of iterative development is the constant improvement of children, while maintaining all the previous generation's strengths (no need to re-learn basic knowledge from generation to generation). The greatest disadvantage is the inability to think outside the box. The AGI is always going to implement its programming and as such if an opponent knows that programming they can predict how the AGI will act and take advantage of it (any expert gamer will know this). So what does this all mean? It means that as AGIs and organics (transhumans and uplifts) will have niche areas that they will always excel over each other. Organics will always do better in chaotic situations (like direct combat) where predictability can spell death, but AGIs will always do better in planning, organizing, and executing of plans (both short term and long) (like tactical command). The biggest threat I see to transhumanity now is their immortality. With immortality there is lest need to have children (beyond increasing the number of transhumans in the solar system) and as such the evolution process is effectively put on hold. My guess is that groups like the Exhumans and Ultimates (who are actively trying to increase the biological process to create the ultimate next evolutionary step) will be the true future of transhumanity and the evolutionary process.
Jovian Motto: Your mind is original. Preserve it. Your body is a temple. Maintain it. Immortality is an illusion. Forget it.
Decivre Decivre's picture
Re: Evolution vs Iterative Development
I see immortality as less of a weakness and more of an advantage. Rather than relying on the next generation to be an improvement on the current generation's weaknesses with all of its potential strengths, one has an infinite span of time to improve their own weaknesses while retaining strength... with the added advantage of lacking the training and growth time that raising a new generation would require. The end-result is exactly what you would expect from an immortal being: the potential to have an infinite number of years of experience behind your belt, and the ability to have a young enough body to utilize them. That said, iterative development will likely replace evolution in such a transhuman society like the one present in Eclipse Phase. Even if people are now immortal, and there is less incentive to raise a new generation, people who are capable of switching from body to body will still wish to have a cutting-edge for the capabilities of their bodies. "Next generation" will come to mean a new series of morphs with all of the strengths and none of the weaknesses of the last generation, but with the same minds that piloted the last generation's bodies.
Transhumans will one day be the Luddites of the posthuman age. [url=http://bit.ly/2p3wk7c]Help me get my gaming fix, if you want.[/url]
TBRMInsanity TBRMInsanity's picture
Re: Evolution vs Iterative Development
Decivre wrote:
I see immortality as less of a weakness and more of an advantage. Rather than relying on the next generation to be an improvement on the current generation's weaknesses with all of its potential strengths, one has an infinite span of time to improve their own weaknesses while retaining strength... with the added advantage of lacking the training and growth time that raising a new generation would require. The end-result is exactly what you would expect from an immortal being: the potential to have an infinite number of years of experience behind your belt, and the ability to have a young enough body to utilize them. That said, iterative development will likely replace evolution in such a transhuman society like the one present in Eclipse Phase. Even if people are now immortal, and there is less incentive to raise a new generation, people who are capable of switching from body to body will still wish to have a cutting-edge for the capabilities of their bodies. "Next generation" will come to mean a new series of morphs with all of the strengths and none of the weaknesses of the last generation, but with the same minds that piloted the last generation's bodies.
So in a sense, the evolutionary process will be replaced by the iterative process used by AGIs and while the advantages of years of evolutions will make the basis of organisms, we would eventually gain the weaknesses of the iterative process in the end (but also it's strengths). What about the exhumans and ultimates, the book is a bit vague on how they create biologically superior individuals, does anyone have an idea how you can maintain immortality but still biologically evolve (thus reducing the evolutionary disadvantage of needing to mature first)?
Jovian Motto: Your mind is original. Preserve it. Your body is a temple. Maintain it. Immortality is an illusion. Forget it.
nick012000 nick012000's picture
Re: Evolution vs Iterative Development
TBRMInsanity][quote=Decivre wrote:
What about the exhumans and ultimates, the book is a bit vague on how they create biologically superior individuals, does anyone have an idea how you can maintain immortality but still biologically evolve (thus reducing the evolutionary disadvantage of needing to mature first)?
They just jack their chosen Attributes as high as they can. Typically this involves Remade or Reaper morphs, or similar morphs that also have an Attribute max of 40 (like the two Exhuman morphs given in the GM chapter).

+1 r-Rep , +1 @-rep

Arenamontanus Arenamontanus's picture
Re: Evolution vs Iterative Development
nick012000][quote=TBRMInsanity wrote:
Decivre wrote:
What about the exhumans and ultimates, the book is a bit vague on how they create biologically superior individuals, does anyone have an idea how you can maintain immortality but still biologically evolve (thus reducing the evolutionary disadvantage of needing to mature first)?
They just jack their chosen Attributes as high as they can. ...
Which begs the question of how they actually do it. In my own work on human enhancement I often use the "bottleneck principle": a system is usually limited by one bottleneck, so if you fix that you can improve performance a lot (and you will get some other, new bottleneck somewhere else). Many people's intelligence seem to be limited by their working memory, so training, improving or circumventing it improves their effective intelligence. So one approach for designing a superior individual is to really fix key bottlenecks. Another approach is to look at evolution. The reason we have not evolved a proposed enhancement naturally are usually 1) evolution "wants" different things from us (hence no evolved contraceptive ability), 2) there were tradeoffs (e.g. food availability) in our evolutionary past that made the enhancement not happen (e.g. an even bigger brain is too costly), and 3) there are things evolution cannot evolve, either because they require nonbiological solutions (e.g. radio), fitness deserts have to be crossed (e.g. getting rid of the human appendix; small appendices are easily infected and their owners have lower fitness), lack of time (e.g. some likely intelligence-promoting genes are still diffusing) and a bunch of other possibilities. With technology we can do intelligent design: we can make different tradeoffs (many biomorphs probably burn calories at a rate that would starve them if they live in a hunter-gatherer society), we can include entirely artificial systems extending biology, and we can make long jumps in evolutionary space by designing biology. So I think the ultimate researchers (and of course their colleagues at Cognite and Skinthetics) are spending a lot of time analysing exactly what limits people. They might be buying recordings of human performance in bulk to run data mining, they subject people to severe testing (Aperture Science as an ultimate outfit?) and they run big simulations. When they identify bottlenecks they start looking for what causes them and ways around them. The easy bottlenecks are merely old trade-offs that can be patched with modern biology, but most of those have already been dealt with (this is largely the difference between Flats and modern biomorphs). The real challenge is to design around fundamental limitations of human design. This is where the ultimates have an advantage in that they care little for traditional biology and hence can consider radical new ways of doing the same thing. But coming up with entirely new ways of making bodies stronger (fullerene fibres instead of actin, with diamond myosin-analogues?) is much easier than finding *good* ways of doing it. This is where the ultimates have perhaps their main advantage: they ruthlessly test radical enhancements on themselves or on indentured infugees. The lessons learned go back into the research database and get incorporated in future designs. This is a mixture of evolution and iterative development. Rational planning can do amazing things (long jumps in phenotype space) but is bad at foreseeing consequences. This is where evolutionary approaches excel (local optimization). In fact, if you have the computing power and no ethics you can do amazing stuff in simspaces. Add randomly varied versions of your enhancement to forks of one or more people, run them fast to see who does best in a number of tests and keep the versions that did well. Erase the forks and seed next generation with versions of the modified enhancements. Over time you will get better and better enhancements. At the simplest, this is just a genetic algorithm using random variation. But you can have AIs or forked researchers in the system doing the enhancement updating too, rather than just normal evolution. In fact, you can run the same evolution on them: enhancer designers that do well get more forks (maybe tweaked in different ways). "Congratulations, Dr Soitre, your design reached 75.2% score. That means that you go on to next iteration. Your personal fitness is 0.45, 12% above the median. Good luck with your next assignment." I think Xiphos is an exciting but scary place. Can't wait to have my players go there.
Extropian
Decivre Decivre's picture
Re: Evolution vs Iterative Development
TBRMInsanity wrote:
So in a sense, the evolutionary process will be replaced by the iterative process used by AGIs and while the advantages of years of evolutions will make the basis of organisms, we would eventually gain the weaknesses of the iterative process in the end (but also it's strengths).
One thing about AGIs: a lot of people seem to be under the assumption that they work like very complex modern AIs, when they don't. AGIs are functionally artificial human minds, fully capable of all the unpredictability and irrationality that humans are. Outsmarting one is nowhere near as simple as outsmarting an FPS bot. For the most part, there are no real disadvantages to the iterative process. Human (and uplift, and AGI) individuality ensures that new ideas will continue to come forth, so long as we don't propogate our species in a way that threatens to reduce individuality (en masse forking, for instance)
TBRMInsanity wrote:
What about the exhumans and ultimates, the book is a bit vague on how they create biologically superior individuals, does anyone have an idea how you can maintain immortality but still biologically evolve (thus reducing the evolutionary disadvantage of needing to mature first)?
They are definitely proponents of the iterative process. The Remade morph was manufactured by the Ultimates as a replacement to the standard human form, and is a perfect example of their aims and desires.
Transhumans will one day be the Luddites of the posthuman age. [url=http://bit.ly/2p3wk7c]Help me get my gaming fix, if you want.[/url]
Decivre Decivre's picture
Re: Evolution vs Iterative Development
Arenamontanus wrote:
Which begs the question of how they actually do it. In my own work on human enhancement I often use the "bottleneck principle": a system is usually limited by one bottleneck, so if you fix that you can improve performance a lot (and you will get some other, new bottleneck somewhere else). Many people's intelligence seem to be limited by their working memory, so training, improving or circumventing it improves their effective intelligence. So one approach for designing a superior individual is to really fix key bottlenecks. Another approach is to look at evolution. The reason we have not evolved a proposed enhancement naturally are usually 1) evolution "wants" different things from us (hence no evolved contraceptive ability), 2) there were tradeoffs (e.g. food availability) in our evolutionary past that made the enhancement not happen (e.g. an even bigger brain is too costly), and 3) there are things evolution cannot evolve, either because they require nonbiological solutions (e.g. radio), fitness deserts have to be crossed (e.g. getting rid of the human appendix; small appendices are easily infected and their owners have lower fitness), lack of time (e.g. some likely intelligence-promoting genes are still diffusing) and a bunch of other possibilities. With technology we can do intelligent design: we can make different tradeoffs (many biomorphs probably burn calories at a rate that would starve them if they live in a hunter-gatherer society), we can include entirely artificial systems extending biology, and we can make long jumps in evolutionary space by designing biology. So I think the ultimate researchers (and of course their colleagues at Cognite and Skinthetics) are spending a lot of time analysing exactly what limits people. They might be buying recordings of human performance in bulk to run data mining, they subject people to severe testing (Aperture Science as an ultimate outfit?) and they run big simulations. When they identify bottlenecks they start looking for what causes them and ways around them. The easy bottlenecks are merely old trade-offs that can be patched with modern biology, but most of those have already been dealt with (this is largely the difference between Flats and modern biomorphs). The real challenge is to design around fundamental limitations of human design. This is where the ultimates have an advantage in that they care little for traditional biology and hence can consider radical new ways of doing the same thing. But coming up with entirely new ways of making bodies stronger (fullerene fibres instead of actin, with diamond myosin-analogues?) is much easier than finding *good* ways of doing it. This is where the ultimates have perhaps their main advantage: they ruthlessly test radical enhancements on themselves or on indentured infugees. The lessons learned go back into the research database and get incorporated in future designs. This is a mixture of evolution and iterative development. Rational planning can do amazing things (long jumps in phenotype space) but is bad at foreseeing consequences. This is where evolutionary approaches excel (local optimization).
This really doesn't involve evolution at all. It's a combination of iterative development and natural selection (a very small aspect of the overlying concept of evolution). Natural selection will always exist... but in the world of Eclipse Phase where bodies are bought and sold, it now also comes in the form of profit and loss. Poor-selling morphs will disappear and make room for good selling morphs, which will eventually be replaced by better-selling morphs.
TBRMInsanity wrote:
In fact, if you have the computing power and no ethics you can do amazing stuff in simspaces. Add randomly varied versions of your enhancement to forks of one or more people, run them fast to see who does best in a number of tests and keep the versions that did well. Erase the forks and seed next generation with versions of the modified enhancements. Over time you will get better and better enhancements. At the simplest, this is just a genetic algorithm using random variation. But you can have AIs or forked researchers in the system doing the enhancement updating too, rather than just normal evolution. In fact, you can run the same evolution on them: enhancer designers that do well get more forks (maybe tweaked in different ways). "Congratulations, Dr Soitre, your design reached 75.2% score. That means that you go on to next iteration. Your personal fitness is 0.45, 12% above the median. Good luck with your next assignment." I think Xiphos is an exciting but scary place. Can't wait to have my players go there.
Thanks to the invention of the cortical stack, you may not even need multiple forks. You can probably use the same person over and over in a never-ending series of tests, and put him in a better body when the last one fails him (and he dies). Make sure you pay extra for his therapy sessions. He'll need them.
Transhumans will one day be the Luddites of the posthuman age. [url=http://bit.ly/2p3wk7c]Help me get my gaming fix, if you want.[/url]