An interesting question/problem

A few hours ago I set out to answer what I thought was going to be a relatively simple question. Let me explain what that question was by explaining how I arrived at this question. I've been quite interested in artificial or algorithmic evolution for a while now, and I often spend time thinking about how I could design a highly efficient evolutionary framework. There are many examples of this already out there; such as those virtual 3D creatures which evolve and learn to complete specified tasks on their own, simply by applying a natural selection process.

There are many useful applications for evolutionary algorithms, because you can train something like an artificial neural network to solve very difficult tasks, without explicitly defining how the task will be solved. This allows us to develop things such as complex game AI, advanced stock market trading bots, voice and face recognition, etc. All these technologies are actually now based on a lot of code which has not been written by men, but written through an unconscious process very similar to the way evolution solves problems through an entirely unconscious process.

Now imagine something like a self-learning virus, the prime goal of that virus being to spread and multiply. That is literally the closest thing to a virtual evolving creature as you can get. When we start thinking in terms of virtual evolution, a natural question arises, and this is: how long would it take before virtual creatures could reach a level of complexity equal to some of the creatures on Earth. Even the birds and the bees are highly complex creatures, they even have reasonably complex languages through which they can communicate considerably complex information.

One can actually compare virtual creatures and real biological creatures in an informational way. Our DNA contains all the information you need to build a human being, and you can record the chemical structure of DNA as digital information. I find it quite amazing that the entire human genome is only around 3 gigabytes in size. So how long would it take virtual creatures to develop language and develop 3 gigabytes of genetic code? Well it took life on Earth over 3 billion years to get where it is now, so perhaps it will take millions or billions of years.

Well actually it wont take that long, because computers can simulate evolution at a much faster speed. There's no waiting around for the virtual creatures to mature and reproduction is an almost instant process, thus they can replicate much faster than real creatures can, essentially meaning the speed at which they can evolve is only limited by the speed of our computers. But still I wanted to calculate how long this process might take... and that led me to another question which I find far more intriguing, the question I mentioned at the start of this thread.

As I was trying to solve this problem, I thought to myself, if it is possible to determine how long it takes for genetic code to reach a certain level of complexity, based on the rate at which the creature evolves and the DNA mutates... then it must also be possible to determine how fast natural biological life will evolve. The question is: how fast should the complexity of life increase according to the basic laws of probability? Again, we can picture this in a digital or virtual way.

Imagine if I have a virtual 3D creature, and I set a task for it, which is to travel as far as it can from its origin point within a set limit of time. These creatures start off with randomly generated brains and bodies, but as we pick the best from each new generation, they can develop very clever solutions for travelling large distances. Real experiments of this nature have shown that it only takes a few days for our virtual creatures to develop highly sophisticated mechanisms for solving any task we throw at them.

So it is possible to quantize the process and mathematically estimate how quickly the complexity of a system will increase, given you understand how the system evolves. Of course we understand how virtual creatures evolve, because we have designed them ourselves and thus it's relatively easy to state in mathematical terms, how fast a virtual creature will increase in complexity. However, real life evolution is a fairly different ball game. There are many more rules at play and I can't say we entirely understand them all.

But if we could understand those rules properly and define the expected rate of biological evolution, in terms of genetic complexity, then it opens up the exiting possibility that we can mathematically estimate the age of life on Earth by examining our current level of genetic complexity. Imagine the implications of that. It could either confirm or totally undermine the current timeline of evolution... what if life as complex as primates required much more than 3 billion years to evolve based on expected mathematical probabilities? What would that say about life on Earth?

It isn't a simple problem to solve when you really start thinking about it... my first attempt at reaching a mathematical solution was based on birth rates, population numbers, and the average genetic variance between any two random people. I figured it was possible to produce a solution if I knew how fast human DNA was changing (think speed at which each generation appears) and to what extent it was changing (think population of each generation). They say the amount of genetic variance between any two humans is about 0.1% on average.

I soon realized it would not be that simple. Think about the problem again; we want to know how long it takes life to evolve from its most simple form, simple celled organisms in the case of biological life, into the highly complex life we see on Earth today, with huge banks of genetic programming. That 0.1% variance we see between humans is nothing more than a reshuffling of genetic information... it doesn't help much with describing how fast our genetic code actually increases in size. This is one of those annoying processes we don't fully understand.

Also keep in mind that the size of a system doesn't necessarily depict the complexity of a system, however as something becomes more complex it inherently requires more space to store more information, since there is a limit to the efficiency of any system. I think our genetic evolution actually accounts for this by having large volumes of unused space (junk DNA) which is used to extend the genetic code without needing to extend the total size of the code just as it is "needed" (it would be random obviously).

What seems to happen in my opinion, and I'm no expert, is that mother nature always maintains a genetic "buffer" (the junk DNA) so that our genetic code can be easily reprogrammed and actually extended so that some junk DNA can become new functioning DNA and thus increase the overall complexity of the organism. As the size of the genetic code grows, so does the buffer space, as to ensure there is always some free space for extending the active code. This would be my basic mathematical framework for finding a solution to the question I have posed.

I'm not going to dive into any mathematical estimations just yet, because I have a lot more to think about. In fact I'm going to stop at this point because I've written enough for one thread and I want to get some thoughts and input on what I have said so far. I definitely think what I am proposing here is 100% possible, we just need to fully understand how natural biological evolution works so we can understand the mathematical rules underpinning that system. Maybe someone has even attempted this before, I have no idea.

You make some interesting points and it is clear you have given this quite a bit of thought.

One thing that comes to mind that may complicate things even further is that in a lot of ways, environment affects evolution as well. You can't directly determine if an evolution in biology is a step forward or step backward until you know how it affects its day to day life.

For example, say an organism has a sudden leap in evolution and it's eyesight is now twice as good as its previous ancestor. This organism may then be able to hunt prey much better, become stronger, appeal to the opposite sex and procreate more. Thus increasing its enhanced population to become the more dominant subset of it's species.

If we take that same concept and apply this step in evolution to an organism that lives deep underground where there is no light and the enhanced eyesight makes the organism no better than its ancestor, then its giant evolutionary step is sorted of 'wasted'.

You probably could create a variable in your equations that mimic the chances that a step in evolution is successful or not. It would just further complicate an already complicated equation!

reply to post by fenceSitter

If we take that same concept and apply this step in evolution to an organism that lives deep underground where there is no light and the enhanced eyesight makes the organism no better than its ancestor, then its giant evolutionary step is sorted of 'wasted'.

I think you're missing the bigger picture here. Natural selection ensures the bad or pointless mutations are phased out, while useful changes are adopted. That's why the natural direction of evolution is in the direction of ever-increasing complexity. The problem is mathematically quantifying that growth trend, which requires understanding the system. And obviously the whole thing is based in probability, you can't say for sure how long it will take an organism to reach a certain level of complexity, or even if it will reach that level of complexity... but just like the toss of a coin, the probable outcomes can be estimated and averaged.

reply to post by ChaoticOrder

True. Natural selection does phase out useless evolution and majority of the time is does result in more complexity, but maybe not always. What if a human was born without an appendix? It is not needed, possibly the body could expend it's resources in a more efficient manner and be a better design overall, but it would also be simpler and not more complex.

Maybe this is just a rarity that may not even need to be considered, just throwing ideas out there for you to consider. I am a programmer and do have an interest in producing algorithms that mimic natural systems. I've even toyed around with my own AI design. Some interesting results but my computer runs out of memory long before I can really determine how well it is working! Still got lots of ideas to incorporate though.

Back to your topic though - I think the first step is to determine all of the factors that come into play that determine the results of evolution. Like you said you need to understand these processes before you can estimate them.

reply to post by fenceSitter


What I was trying to say is that those odd events don't matter, because the underlying trend is always the same, ever-increasing complexity. We want generic probabilities, we don't need to simulate every possible side-road evolution could take.

reply to post by ChaoticOrder


I find your ideas intriguing and I'm sure some science organization somewhere is likely doing exactly what you proposed although I don't really know. I do find your concept of junk dna to be interesting. I never thought of junk dna as buffer space, almost like extra room on a hard drive, that may contain information about how the hard drive operates but is otherwise unused space. I don't know how to feel about that thought yet, but it is an interesting concept none the less.

Originally posted by ChaoticOrder
How long would it take before virtual creatures could reach a level of complexity equal to some of the creatures on Earth?

It is impossible to answer that question, since we really don't know complex the creatures on earth actually are, why they exist, why they feel pain or pleasure or other motivations. It is an unanswered mystery. Any attempt to actually frame that complexity will necessarily requires a lot of assumptions and over-simplifications.

Originally posted by ChaoticOrder
One can actually compare virtual creatures and real biological creatures in an informational way. Our DNA contains all the information you need to build a human being, and you can record the chemical structure of DNA as digital information.

We can record the DNA sequence, but we have no idea how it all works together. It is probably not correct to rigorously compare the complexities of DNA with the modest instruction set of a computer at this time, except in the crudest and most general terms. You can do a binary dump of an executable and learn completely what the program does. However, you can obtain a complete description of a person's DNA, and have no idea what all that information actually means.

I think the basic problem I have with your conjectures is that your ideas confuse metaphor with reality: A wind-up toy can mimic a mouse, but it isn't a mouse -- A perfect wax figure looks like a person, but is not really alive -- The reflection in the mirror is not really a portal to an identical world -- A computer simulation is just a process that performs a defined task.

We speak of computer viruses, genetic algorithms, neural networks. They are cool names, but just metaphors for real-world systems. Real viruses are more complex than computer viruses. We still don't know how to cure someone who has rabies or HIV, or completely understand how immunization really works. We can learn the complete genetic code of a simple virus. But that hasn't helped us cure any viral infection so far, and maybe never will.

There are forces at work in the world that are a complete and utter mystery to humans It is hard to model something that you don't completely understand. Any results obtained from that model can't be trusted, and it is potentially dangerous to think otherwise.

All that said, your thread definitely made me think for a while.

I imagine your variables would grow in number exponentially as you progress. Definitely a very daunting task!

Not one I'd like to undertake after working all week!!

reply to post by Axial Leader

It is impossible to answer that question, since we really don't know complex the creatures on earth actually are

The level of complexity is simply derived from the size of the genetic code. As I stated, the size of a system doesn't necessarily depict the complexity of a system, since different systems can have different complexities, but overall the system will slowly increase in size as it becomes more complex, since there's a limit on the efficiency of any system. And that gives us a basic way to measure the complexity of life on Earth and then estimate how long it would take, according to the expected probabilities, to reach that level of complexity.


reply to post by PurpleChiten


Yes it will be a very hard problem to solve, but certainly not impossible. And computers will be able to handle the calculations required quite easily.

Here's a good piece of information concerning functional DNA vs pointless buffer DNA:

The Encyclopedia of DNA Elements (ENCODE) project[1] reported in September 2012 that over 80% of DNA in the human genome "serves some purpose, biochemically speaking".

en.wikipedia.org...

It turns out a lot of junk DNA is not just junk. I would throw out a wild guess and say that more like 90% of our DNA is functional and provides some sort of useful purpose, while a healthy 10% is pointless code maintained as a buffer zone for future development.

Originally posted by ChaoticOrder
reply to post by Axial Leader

 

Overall the system will slowly increase in size as it becomes more complex, since there's a limit on the efficiency of any system. And that gives us a basic way to measure the complexity of life on Earth and then estimate how long it would take, according to the expected probabilities, to reach that level of complexity.

I see where you are going with this, but I am not sure if we can really measure the complexity of something we don't fully see, and there are a lot of unknowns at play -- the main unknown being how it all works! (Edit: By "all" I mean a substantial part of the "system" that constitutes "living forms".)

Err -- now, we are kind of caught now in the definition of "complexity" itself, so the whole conversation is becoming pretty complex, if you know what I mean

Anyway, I don't want to discourage you. Sounds like you have some good ideas. Please keep us posted.

reply to post by Axial Leader


Perhaps I can explain it in another way which makes more sense. Any system can be expressed in mathematical terms, and usually the underlying rules which form the basis of the whole system are lot simpler than you may first assume.

Imagine I have a simple evolutionary algorithm... the digital DNA of my virtual creatures is just a string of characters, kind of like how we store human DNA as a string of characters (G, A, T and C). At the start I have a bunch of creature with DNA consisting of 32 random characters.

Now I give my creatures a task, the task in question is irrelevant, it could be anything. First I test the "fitness" of my creatures by measuring how well they perform this task. Then I pick perhaps 50% of the best performing subject and allow them to "breed".

This process allows some of their DNA to become fused and there is also a small possibility random mutations in the DNA could occur,. Some of those fusions and mutations may also act to extend the size of the DNA (increase the number of characters). You simply keep repeating that process.

The result will be a clear increase in complexity of the DNA over time. As it becomes more and more complex, the DNA will get larger and larger in size to account for this increase in complexity. And by running this experiment we can easily measure how quickly the complexity of such a system will increase.

The complexity of the DNA will always be proportional to the size of the DNA within some probable threshold... human DNA is no different. And if we can confidently say at least 80% of our DNA is functional, that gives us a pretty good idea of the complexity relative to size.

And now we can ask the question... mathematically speaking, how many genetic mutations would be expected for our human DNA to have reached the level of complexity we see now, and how many generations, with what size populations, over how long, would explain the level of complexity we see.

reply to post by ChaoticOrder


You've a keen mind.

what if life as complex as primates required much more than 3 billion years to evolve based on expected mathematical probabilities?

Ok, but it is really hard to factor in when the next extinction event is going to take place. We've had five already, and that would need to be factored in. Evolution would have progressed much more quickly without periodic interference of global proportions. To be sure, the best innovations in the evolutionary process stemmed from adaptations to the environment.

The virtual equivalent of "evolution" wouldn't need to parley with environmental set-backs, requiring millions of years to recover from. However, I think you would need to program in a "virtual selection" process, one to simulate natural disasters, just to make the "virtual" counterpart more robust.

A virtual process would run rampant rather quickly, but "causing" random extinction events would probably cause very interesting "virtual adaptations".

I think perhaps the best way to resolve the questions would be to set up such a virtual environment, and watch it execute. That, I would think, is already taking place somewhere. It would be nice to find said simulation software somewhere, and run it on an isolated personal network.

reply to post by Druid42


Great post. Massive extinction level events are something I hadn't considered until now. And I'm sure simulating extinction level events would produce some very interesting results. It also means we haven't really had more than 3 billion years of continual development as I assumed, the time available to reach out current level of complexity is shortened by extinction level events, because they can cause massive set-backs... but at the same time, as you mentioned, they could also result in interesting spontaneous adaptations.

reply to post by ChaoticOrder


Yes, I'd think evolution occurred in 500 million year spurts. Kinda forces the hand of nature to adapt.

Also, you would have to deal with environmental issues, poor air quality, lack of sunlight, saline water, etc, to make a truly accurate simulation. Call it evolution under duress. Such factors would need modeled as well.

So... a useless adaptation today might suddenly turn out to be critical in the future -but- since it's useless today that line gets wiped out?

Cool thread, I'll definitely be following it.

AAGGH! AI! KILL IT WITH FIRE!