Brain for a Snake. Part 3. Idiocracy
Вставка
- Опубліковано 25 сер 2024
- I continue to experiment with the simplest brain for a snake from a classic game.
What will happen to the descendants of a perfect organism under the influence of natural selection?
Let's test this in an experiment.
You can support the channel on Patreon:
/ simulifehub
More support, more opportunities for new projects.
I wonder if, although individual performance dropped, the entire species benefited from the skill drop due to a more even spread of food across the population
interesting idea
@@wallcraft-video thanks! Evolutionary game theory and its emergent properties are interesting indeed
Make a version where food starts to get more and more scarce after the yellow area is reached. To force evolutionary pressure. Or, you can do like the others, and take that "food per match" thing and use it as a fitness value.
Reminds me of a sim where the food spawning rate scales by (1-x) where x is (number of food + snakes)/(Max limit)
i think make mutation rare would(not wood) make the yellow region higher or make disappear? maybe if have the projects source code i try it but with out it you have to do it.
rarer
In the classic artificial life video "I programmed some creatures. They Evolved.", the guy showed how evolved organisms without random mutations were a better fit for their first scenario, but changing any parameters of the world (e.g. adding a single wall) was apocalyptic. So while most people imagine things evolving into a perfect form, you are very correct that "evolution has no end". It's a process that is about adapting to change more than finding the best possible solution. It's also why life on Earth has survived 5+ mass extinctions, albeit in forms that the previous generations would find bizarre. The first plants created oxygen through photosynthesis as a fluke of chemistry, and that oxygen killed nearly every other species. Now we mammals not only use that oxygen, we can't live without it.
That said I do have a pet peeve about people confusing societal change with evolutionary change. Society changes in similar ways to evolution, but evolution is wayyyyy slower, and cannot happen over just a couple generations. Now, someone enshrining a stupid thing they did as tradition and telling all their future descendants that it was good and they should do it too... yeah that can happen. One guy called ideas that spread LIKE genes "memes" but somehow that got co-opted to mean funny pictures of cats. The more you know!
The first time wood became a thing also had remarkable effects. Nothing was around to digest it yet, so it just piled up and got sealed up, creating much of our natural gas and oil reserves today while also sequestering metric Gigatons of carbon, causing a massive drop in temperature.
And each extinction was followed by a veritable explosion of new life, refilling the niches left in the aftermath.
As for the speed of evolution, this is extremely variable. Some changes take place over like millions of years, but in some rare instances you can get new species in like tens of years.
Certainly in the lab setting, where we've put various microbes under various selective pressures to see what sort of thing would form as a result.
But even out in nature this rarely happens. Namely when just very few mutations are sufficient to change an organism quite drastically, and there just so happens to be a neighbouring niche that can be filled by those mutations. The mutants at first will somewhat interact with the original population, but if the two niches are different enough, they will diverge surprisingly quickly, within just a few generations.
Afaik this happened in some fish that normally lives in pretty shallow waters, but some mutation occurred, making it much more suited to life in deeper waters or something. I forget the details. - At any rate, that happened fairly recently and very quickly.
Still, that's definitely a special outlier situation. The norm certainly is, that things take a loooong time.
Ultimately, biology has like three rules:
- you cannot break the laws of physics
- genes that improve their own reproduction rate will outcompete genes that worsen their own reproduction rate
- there are no rules
@@Kram1032*coal reserved. The wood became coal. But the rest is very true
what program / language did you use to simulate the interface and the states?
Technically, natural selection doesn't care about the success of the individual but rather about the success of a gene. If some gene makes the individual die sooner but also have more offspring which themselves survive to have offspring, that gene is going to get promoted despite seemingly being worse.
You started here with a *single* Adam, giving little to no opportunity to retain genes. What happens if you fill the board with Adams right away, and then wait for the same number of moves? - Each originally identical copy of Adam gets to have its own set of offspring and way of modifying its performance. Do all of them converge to the same rough area on the chart, or are some significantly higher or lower (while still pulling through)?
if I immediately fill the board with Adams and then wait the same number of moves, it will be the same. I think so. There are more unsuccessful mutations than successful ones. With artificial selection, they can be severely cut off. With natural selection, this happens more gently. Therefore, I tend to think that the ideal with natural selection is unattainable.
@@wallcraft-video Unless you introduce some notion of elitism (explicitly keeping "the currently best" in the gene pool), yes, natural selection will not tend exactly to The best one.
However, by way of your setup here, you are also selecting for a population rather than an individual. I wonder if an evolved population (with diversity) can do a bit better than clones of the same "perfect" individual
And I think trying it with a full board of Adams could still be interesting to figure out which specific motivations tend to happen, as well as whether the noise range is a different one.
Mostly I'm interested in how much the "very first mutation" matters to the further trajectory. At a guess, there are a lot of early mutations that are "not great but still good enough", but different such early mutations imply different further mutations that are still "good enough"
if you want to increase the limit you can reach by natural selection, you must have much harsher condition (less food, more competition). Because for now, the bad mutation are not punished enough, and survive too much. It is similar to humans right now, a life too easy leads to devolution.
Yes and no. Needs to be a gradual change. If you go too harsh, you just make any mutation that *could* eventually lead to something better but currently is slightly worse unsurvivable.
Slowly ramping up the difficulty, or perhaps directly tying it to the number of survivors should certainly help though
Interesting, what if you use Evolution Strategies or other "mathematical" optimisation instead?
I think you should also set up tests, that have food in fixed positions, and you measure the time for them to eat everything alone
the smartest ones, you should be able to lure them out of a maze by placing food along the path out ;)
Hold up, what's the system in the outro?
It looks like physarum but somehow more discrete
Fascinating
This is a continuation of the project "The battle of clans"
The mode is turned on here, where the amount of energy moving inside the body is displayed.
hi dude ive been following your channel for a bit and im really interested in this type of things and would like to specialise in this in uni. But there is one problem, what is this type of programming called? Also can i get a job for doing these things all day?
In short, evolution sacrifices a little bit of perfection for a little bit of flexibility. In a dinamic environment, flexibility is important so that the organism can adapt quicker to changes. A perfectly fit organism is much more vulnerable to changes and takes longer time to adapt.
Diversity in a population is also important. What was thought to be a bad mutation could save a population from extinction in a drastic environmental change.
@@wallcraft-video Diversity is a key ingredient indeed.
Very interesting
what will it do if you put an ant, a snake and another kind of simulation and make them compete ?
Possibly with a much larger population size you would see evolution reach a higher threshold. Large population sizes should mean that random bad luck is less important and those very small benefits will not be swamped by the noise in the environment.
Maybe you're right, but I'm not sure
The real champion is the friends we made along the way.
1:10 fast foward*
Yet evolution (macro) doesn't happen.
Maybe a new form of evolution.
Start with 6 copies of Adams and put the in match simulation.
The two snake with the lowest score at the end go extinct. The 2 snake with highest score mutate and replace the 2 dead, the 2 middle snake just stay unchanged.
Automate this and let it run for a few hundred thousand matches to select ever fitter individuals hopefully
The environment you are performing the natural selection in and the way in which you are testing fitness via matches are not the same. It may be that optimizing for survival in the environment does not give the best strategy in the matches, which simply measure food collected and do not measure survivability or chance of having offspring. Although your theory is interesting, I don't think your test is sufficient to show that it is correct.
Good point, I have to agree with you.
With gene editing and cybernetics humanity might be going to intelligent design instead of natural selection
Nice
sad that you changed voice generator to a more common one :(
If you have links to good generators, you can give a link. I've tried recording my voice, but it sounds even worse. Hiring an announcer is too expensive.
@@wallcraft-video Time to train a snake who can read scripts!
day 3 of telling you to bring back the old voice