The dragging tail seemed very important to those creatures. It was helping to push the "spine" nodes upwards, which pulls one of the "hind legs" up so that the other hind leg can make contact with the ground. I thought it was really awesome.
Have you tried putting an energy restriction on the creatures? One major reason real-life organisms can't solve problems just by spazzing out is that doing so is very energy inefficient. Perhaps that is what is missing.
I think this is the right answer for a couple reasons: all that movement across all those limbs is expensive like you say. The other consideration is that it costs a lot for organism to have additional limbs. Each extra node and limb should cost more in terms of energy not only for movement/basic operation but to mimic development of an extra limb.
+Jordan Cooper Constructive criticism time: An idea I just thought of is as follows; give every creature a certain amount of energy they can work with at the start. With every additional limb, they need to divide that energy among those limbs to sustain them. If the energy being sent to those limbs for a generation is below a certain threshold, then the limb should deteriorate as a result. This would lead to the most energy efficient and thusly the most likely to survive design at the end of the simulation. Now all we need is someone who can code this into a usable program and BOOM! Real evolution.
+Jacob Riley For the most part, I concur. However I disagree with the part about a creatures deteriorated body being passed on to the next generation. If somebody gets their arm chopped off, they dont get to have three armed kids. If having fewer limbs is favorable, it should be mutated from reducing the size of the limb or a random mutation to remove a limb.
@@personmcdudeguy Deteriorating a limb doesn't mean chopping it off! Deterioration means the gradual reducing of power/size over sucessive generations by mutation and evolution. Remember that in these simulators most of the slowest half of each generation get killed off, so if something managed to mutate an extra limb that caused it do have too little energy per limb then presumably it would be gone in the same generation purely for this reason. As well as an energy restriction, a more realistic simulator would also need to have variable amounts of 'killing off' each generation. Specifically I would find a way of slowly increasing the distance that 'food' is away from the start point (an perhaps from the last 'food' etc), and all those that don't get at least that far would die of starvation. You will then have variable population sizes, as well as the possibility of extinction if you give them a challenge that's too difficult. You might then also need to allow for variable levels of reproduction to try and counter the occurences of near extinction. I guess allowing for variable 'killing off' might also lead to increasing the time limit, or lifting it completely, and then measuring both distance travelled and time survived for, and then judging them by a combination of both rather than just distance, but I guess the risk there is there could be a creature evolved who goes on infinitely, by always reaching food just before they would otherwise starve.
@@MrDannyDetail Make them have a constant maximum age at the start of the evolution. If a creature passed it's maximum age, it would die. Of course, the maximum age could be increased with mutations.
Also, it's interesting how the super-restricted creatures at the very end almost started to resemble having a proper quadruped running gait (back legs push while front legs are up, front legs push while back legs are up). Of course, in two dimensions, rather than three. On that note, I wonder if the two dimensional limitation is responsible for the typical high-friction-back-and-low-friction-front arrangement you end up with when without as many restrictions.
You know what, I think the tail helped the creatures stabilize when running. That's probably why it kept the tail with a certain amount of friction on the ground.
That guy you ended up with around 23 minutes is awesome! The red and black node are alternating to basically walk while dragging the pink node behind it, keeping it from flipping over, and one of the middle nodes is also moving in time to use inertia to its advantage! Nothing lame about that! Great video.
imo the taildragging creature is the best due to the tail, as despite it being dragged along, it provides some level of stability i think so, anyway. it distributes the centre of mass and widens the contact surface area so that the creature doesn't just flip itself over
I think the reason your creatures are getting "scrunched up" is similar to why being small is an advantage for many real creatures: material strength doesn't necessarily scale with size. A cat can fall three storeys and walk away with bruises, while an elephant risks serious injury merely by falling over. There aren't any giant insects in this era because exoskeletons and lungless respiratory systems don't scale well. Basically if you want to end up with larger creatures you have to offer incentives for size: in the real world it's economies of scale and efficiency of metabolism and food storage, plus extra heft in a fight.
BTW don't leave your simulation running overnight. They will take over! They'll be our overlords. They'll make us run to the right on meter tracks and they'll go "Ooh, look at that. That one ran 26 miles then puked." "That fat one passed out after 5 miles."
If you want to get rid of the standard 'high friction on the back, low friction on the front' creature, what you could do is make the friction of all nodes constant and unchangeable. The only way they can evolve to be faster is by moving more agilely , which seems to be your goal. It's also closer to how animals evolve, they can change their bones and muscles, but rarely do they rely on friction to improve.
You will see creatures shaking for few meters and then stopping i guess ... i think in nodes and muscles environment there exist only one optimal evolutionary design so unless he add some wheels, obstacles or whatever the outcome will be always very similar.
if you want to get graceful movement you need to not limit it by time but energy, give them some energy and wasted movement gets you less distance. while a slow careful movement might get them farther. right now the best ones are the ones most capable of pushing as much energy into their environment to move them as possible, they're sprinters but won't actually make it very far. to do this, they create a single driver and everything else goes friction-less to limit drag. muscles that contract or expand twice as fast would expend 4 times the energy. You then also limit them to muscle movements that don't use an extremely low amount of energy. also limit friction values since a near friction-less "node" is almost impossible in real life, even with vehicles.
Exactly what I was thinking he should see the motion of these particles with limited energy. It is exactly what happens in real life , a 100m sprinter tries to utilise all his energy in that 100 m .
I think is because all the 'muscles' contract simultaneously, maybe if you allow them to have multiple internal clocks it will make them have a more galloping motion
+Antoine Spartaphèze That's exactly what I was thinking. Cary says it's not doing anything, but it's light pink, not white. Looks more like it's evolved the precise amount of drag to keep the whole body steady instead of bouncing around as it's "feet" move. In fact, I'd be willing to bet that spike in the fitness increase was caused by the mutation/evolution of said tail.
You could try an experiment where you run a regular evolution, and also create a separate creature from scratch. Then for each evolution generation, you tweak the single creature a bit; only allowing yourself to change one value at a time. It could be sort of an Evolution vs Intelligent Design thing.
Well your model is great, but it only covers convergent evolution which is why at some point you get only copies of the same creature. If the conditions are always the same then the creatures are always getting more and more specialized. What happens in the real world is that some species won the race not because they did something better, but because they did more things. They were able to survive in more adverse conditions in which the more specialized ones couldnt. If you wish for your model to trully cover evolution as a whole you have to add variables that change for each generation, like for one the ground is stickyer, or the wind is blowing, the gravity is weaker, there's a slope, whatever, you will probably get to see more diverse forms that can deal with more situations that way. At least in theory.
Paulo Mangano I don't think the scenario should change every generation. Instead have a few separate scenarios and run the initial 1k through all of them. Then keeping them separate with a few minor variations to the scenarios see how the creatures change. I think that would be quite cool as earth has many different places and those places are home to animals that have adapted to that area's unique obstacles. Of course all this adds complexity and will greatly increase simulation time.
bensemus x i said every generation because change is often super gradual, and would probably provide more means to introduce variation over just a few vars that would not require that much of processing power, but in a real simulation it would be so gradual it wouldnt matter from a generation to the next.
I think the main way to encourage divergent evolution is to provide different niches for the creatures to move into. Otherwise only one species will prevail, Gause's Competitive Exclusion Principle and all that.
The easiest 'niches' I could see for this model would be furthest moved forward, furthest moved backward, and most ststionary. However, I expect the forward and backward moving designs would end up mirroring one another. Especially since mutation can flip the direction of motion of a creature in a single generation. Still... It would be 3 niches instead of one... And thus we'd see divergence on some level...
I actually like the first part of this video a lot more, as evolution is about being random and unpredictable. Also at the last generation the most effective creatures had grown something like two "legs" that were speeding up the whole structure (black and brownish). Another interesting fact is that with previous evolution series creatures were training their skill in both going backward and forward, but with 8 nodes at the final generation almost every one is progressing to the right no matter what the rules were. Great video, can't wait to see more of those!
A couple of thoughts on this (not sure if you have drawn these conclusions since this video): It is likely that the piston movement is the simplest , most efficient form of movement for creatures in this world. Thus all shapes and sizes will likely evolve towards being able to perform it. If you add restriction , you are essentially restricting how possible evolving towards this movement type is, thus you will see worse performing creatures the more restrictions you place. (Essentially you are adding a second fitness function: "Do these creatures look like i want them to?") If you look at the way slugs and such move they, do adopt this sort of movement , just in a less effective way than the creatures you have evolved. The way bipeds often walk is essentially two nodes of high friction and then a node structure above them to move one of the "feet" nodes in front of the other. This is less efficient than the "piston like" movement your creatures have adopted. The reason why bipeds evolved to walk this way however is they have more tasks than "walk right on a horizontal plane as fast as possible". My suggestion might be to introduce slopes or steps in the ground and see if the creatures can evolve to over come those tasks. This may lead to non piston-like movement.
I think the reason they always evolve into twitching heaps is because of the way nodes and muscles can freely move through each other. Also because the muscles always have to be moving.
I think one reason why creatures end up like they do, is because they lack bones and there is nothing preventing them from collapsing like they do. Since only factor is "how far right can you go", well, once the creatures get certain setup that works, everything else just... well, they have no use, but since simulation does not allow those parts to disappear, they just dangle around.
I know it's a 5 year old video but the results in the first half actually make sense since the most efficient form of locomotion under this condition is the triangle as shown in previous videos, so no matter how many nodes you add, it will still evolve towards the triangle and any additional nodes will just become vestigial.
I don't know how to code, so may I propose that when a creature passes 15 meters it goes to an higher level and will reproduce and adapt to it's new environment. this will require to simulate more creature, but it will created more complex creatures and more fun. Some levels could select the creatures based on their size, their ability to pass obstacle like bumps and holes. the levels should not change to fast because a to big changes may destroy the creatures not make them stronger. If the simulation runs really long I predict some creature will have something like 20 nodes.
I'm quite impressed with the results of part 1. The double node crawling action is actually pretty complex. I also think that trailing pink node is used as a balance - it's helping keep the creature in position against the floor.
The best creature at 1170 gens was a bit more interesting. It was basically bipedal compared to the tringle. Instead of one black node (unipedal) it had two appendages alternating on the ground (bipedal) so taking advantage of the wasted time in between each single step the tringle achieved. That was very interesting to me. Because there's no balancing organs it's just a bipedal creature pushing its head along the ground in front of it.
I don't think you can make species that are only made of muscles stand up. Ever. They don't have a brain or a computer to balance themselves. They would need a basic AI that re-balance themselves if their average angle gets messed up.
tenedria As said by Ian Malcom from Jurassic Park, "Life finds away." Even if that life is a computer system forcing a group of lines and dots to change...
I think that dragging pink node is really cool because you see completely useless structures in animals all the time due to the fact that an ancestor had it (like an appendix possibly). Also, the fact that they evolved to look like the original creatures from previous videos is an amazing showing of convergent evolution, to have started totally different and end up evolving to be so similar. So cool. Love the videos, thanks!
it makes sense that your last try didn't work as well. caterpillars aren't meant for speed, they are meant for having a point of contact on a walking surface at all times. your other creatures also make sense, they resemble a kangaroo, rapidly pushing with their feet and staying balanced with their tail. I found this all interesting and very logical.
That’s not how the evolution works; it takes the better half of 1000 creatures and then takes each one and duplicates it twice with minor changes. That means you’d have to design 1000 creatures.
The pink tail might not be a bug, but a feature. It pulls on that top node, adding a rotational force to the whole thing, which might act similar to a spoiler on a car.
let the creatures do different tasks, not just walk in a flat surface. maybe add a knee high "water"layer where the space has friction. this could encourage the blobs to stand up.
I have watched a lot of your videos and they are great! Good job and I love the commentary! One thing I see with almost all simulations on your videos is that they are heavily dependant on the initial performance. The twitching creatures are far faster and easier to evolve in a short period of time. Some other interesting ways of travel may be slower to evolve in comparison. I would assume your code heavily benefits the "twitchers" for the first generations and therefore you find a local maxima, quickly, and your gene pool only consists of twitchers. There might be other maxima but as they are quickly "killed" we never get to see them. This would explain your problem with the "twitchers". A solution is to classify species as nature does: some species evolve slower but still maintain to survive and eventually get to be faster than other species. The code implementation would be interesting and a bit tricky with different groups of species. But you are really good at what you do so I would love to see the results! (!)
I love these videos and I love your commentary. You're ridiculously smart, but you're so humble that you speak exactly like a dummy like me, so you're both personable and like an expert. It's like talking to a really cool doctor.
I think the issue with your simulator (why the creatures are twitchy and unrealistic) is that you are simulating motion in two dimensions. Real creatures have evolved to move in three dimensions. Hence why the look the way they do. In your simulation, the twitchy creatures succeed because that motion is the most efficient for two-dimensional motion. If the universe were two-dimensional, I think real creatures would be more twitchy and erratic like those simulated.
I would be really excited to see an evolution video where the creatures are not just competing for higher scores on one fitness function, but on two or more(For example, a setup where creatures are rewarded both for running forward and for jumping as high as possible). There are a few reasons I would find this interesting: 1)It would encourage higher biodiversity even at the late stages. Instead of just one species taking over because it's the best at whatever function you gave, there might be both runners and jumpers, and species that aren't great at either but moderate at both, and everything in between. 2)It might lead to multiple types of strategies being tried even within one of the functions. For example, a population where there are both shakers and wheels that can move forward and are competing with each other. I think this could arise as a result of the increased freedom of change in the late stages of the game, as creatures would be able to trade-off between running and jumping freely. A family of organisms could go from running, to jumping, to running again, but with a different strategy. 3)It would contribute to increased realism, as real life ecosystems rarely only have one goal that organisms can pursue, instead having many different ways to secure reproduction for themselves.
MrSporeowns The main reason I came up with that idea was because I was getting bored with the really low lategame biodiversity, as one species takes over because it's the best. Another idea I had for reducing this effect was some system of dependant relationships between species, so that one species can't just kill off the other without dying out themselves(for example, a predator-prey, symbiotic, or parasitic relationship). This idea may be helped along by requiring organisms to gather energy, because that's the point of predator-prey relationships between organisms. I would tend to think(though couldn't know until further experimentation) that a more productive fitness function would be some sort of weighted addition of the two goals of running fast and jumping high, because if you multiply the two functions together, then generalists would have an advantage as specialists would have near zero score on one of the functions. I think, however, that the multiple fitness functions idea would be the easiest to implement at this stage, as well as having the lowest increased load on the processer, since it adds no new mechanics, just changes the fitness function, and fitness function evaluation accounts for a relatively low amount of computation compared to the actual physics simulation. Food pellets might also be easy to implement, since no new physics are really added, but I don't know how one would distribute food pellets around the arena for the best effect. Interspecies relationships would probably be hard to implement because it would require multiple organisms to occupy the same arena and I'm not quite sure how that would be set up at all.
You actually all ready had homologous structures. When you did your simulations one of the body structures would end up dominating the ecosystem, and the structure that dominated any ecosystem was based on what was the most advantages, just like how homologous structures come to be in the real world.
Your part one creature was boring? You evolved a creature with two front legs, three back legs, and a DAMN TAIL! I don't know what awesomer creature you could've hoped for.
It is said that evolution can be extremely slow, and extremely fast. While these videos due showcase this, there simply isn't enough variables. Use your original concept, of whatever evolves best "wins". However, wouldn't it be neat if 5% of the population, maybe every 5 generations (with a variable for the exact amount of generations (written as 5 plus or minus 2)} that 5% got 3 more nodes, or some other "random" occurrence? Maybe, if the program detects stagnation, it throws in a couple of hurtles, and maybe a slope. Have different obstacles appear, disappear, or even just move from one generation to the next. This would help simulate a real environment. Imagine if there was just a small hole in the ground, even a shallow one, forcing more evolution. Making the situation for these creatures as chaotic as possible, much like real life, would make this far more interesting (right now, I love the concept, and love watching this project evolve [pun not intended]). I realize this would require a LOT of coding, and several changes to the program itself, but I imagine that is what you really want out of this project, and I know that that is what I, as a viewer, want to see.
That final example of the "3 box" structure demonstrates undulatory locomotion, which is exactly how caterpillars and worms move through the environment. I don't know how else to put this, but your program discovered a natural method of movement. You have made great progress! Congratulations!
To be fair, I thought that the creature that evolved in the first part was somewhat interesting. It had two propeller nodes in the back where as the creatures in past parts always evolved to have just one. You dismissed the pink node "tail" as a mistake yet since it lasted for so many generations it seems that it was intentionally kept as some sort of stabilizer. There were still some useless nodes in the middle, though I was pleasantly surprised that more than three ended up being utilized. Also, when dealing with creatures that all have the same structure, there may be a way in which the species chart can be utilized. Since the creatures reproduce by themselves, it might be neat to see what percentage of the remaining creatures originate from each individual from round one. Actually, that would still be interesting to analyze when there's multiple species as well. I think it would be fascinating to see how many remain just after 15 rounds and whether or... (continued in reply)
...not only one remains by the later generations. Finally, I'm interested if you would be able to add another variable in. To start, I seem to recall that the muscle movements are based on a command loop. I'd be interested to know whether that loop could be be sped up and slowed down as a mutation. Also, it would be interesting to see if the muscles could be expanded and contracted multiple times in that loop, but at different rates.
Also, I support the idea of automatically flipping around the individuals that end up in the negative numbers. I've seen before that sometimes the creatures just get entirely flipped around by a single node being moved. That's why in the earlier rounds the ones moving in the wrong direction still had the same amount of movement as those who went right. By killing those who get flipped around, it essentially punishes otherwise perfectly reasonable changes while limiting the number of viable mutations. Because of that, if the creatures are flipped it might result in more varied designs. To be fair, it could have almost no effect though I don't see any harm in trying.
Sorry if someone already said it, but that pink node dressing behind was serving a function. The fact that it stuck around for a hundred generations means it was better than reducing its friction, and then it continued until the end, with over a thousand generations it had to be doing something useful.
Thanks for making these videos, these have taught me a lot about evolution and I usually watch these before i sleep as they calm me down. Anyways keep up with the great work and videos!👌👌 (you should also do this with the vertical jump)
You should Aldo Bump The road , this would avoid the smalls structures and should help more "walking" structures. If tou do that, maybe don't do a very bumpy road since the beginning, maybe go from a quite linear road for the firsts generations to a very spiky one for the avanced generations.
he problem is if he put no restriction he got a dominent 3:3 specie And actually each time he put restrictions , it only reproduce the same pattern but with useless elements evolution try to remove
Well its not about restrictions it is about Coal. the problem of not coming up with new ideas but it seems that Moving fast to right has an efficient layout.Bushing with Rear Legs while front legs are for support. it allso works IRL for example Cats or Cheetahs. or even Antilopes they all have strong back legs that repel them forvard while front legs are for Balance and support.
i have an idea why the creatures always default to the high friction on the left and low on the right. one very important aspect of evolution of a species is the environment they live in. for example, if the species is found in a desert, then of course they wouldn't have developed fins or gills! my guess is that the current default structure is the most efficient on the straight, simple runway that you have now. but if you added an extra bit a randomisation where the ground friction or the gravity of the world can change then you might get different results. it also might be cool to add in some kind of barrier, like water or bushes that slow them down, but don't stop them completely!
at 21:00 you can see that a single node is being dragged along on one muscle. I thought you made restrictions so that the muscles and nodes had the same connections? In the original double-square formation every node had at least 3 muscles connected. How could a creature have one node dragged along on one muscle if you made homogenous restrictions?
I actually really like the creature you ended up with in part 1. It's actually surprisingly animal-like when you think about it: two front legs for balance, two back legs for propulsion, and a tail dragging on the ground behind it.
I think the reason you get the worst reverse going creature going almost as far as the best forward going creature is because the simulation is dominated by its starting conditions. A creature that goes far in the forward direction will go far in the backward direction. The reason it goes forward or backward is due to its initial topple.
Six years late but: What if instead of nodes having differing friction, they had differing weight instead? It'd allow you to also see if they would evolve differently with ice physics
If every 10 generations is about 500 years, then you'd need 1,000,000 generations to get to 50 million years which is more on the scale for the Theory of Evolution. I'd love to see the end of your videos have something on that scale, even if it does plateau.
My recommendation would be to return the ability to have morphological mutations, but just make them rare so there are particular morphology has time to evolve more fitness within that morphology. Changes in morphology that are beneficial anyway will tend to survive, but the more fit a creature gets with a specific morphology, the less likely a change in morphology will be to benefit that creature. I know you wanted to simulate what is seen in natural biology, but the artificial life forms you're evolving are not biological and keep in mind that biology took a long time and an awful lot of generations to evolve the diversity that it currently has. By the way, you mentioned homologous structures in mammals, but the slide you're showing at the time had one non-mammal in with the three mammals it was showing homologous structures on. It's also worth noting that those structure is being homologous doesn't mean that they haven't lost or gained any bones or muscles. Not all mammals have the same number of toes, for example. If you haven't already discovered this for yourself, I think you will find that evolution is much more interesting when you don't try so hard to make it turn out a certain way. Let it surprise you.
I think the problem here is that since the time limit is so short, mad-dashing in an unsustainable manner is the most effective strategy, and the best way to do that is to be compact and pulse erratically.
As much as we want to see some upright locomotion, it's actually much more complex that one would initially think. Sure, most people reading this could stand up right now and not think twice about it, but balancing on two feet requires a lot of coordination to do. That's why many of the creatures in this simulation opt for the path of least resistance: walking is hard, crawling is less hard, jittering randomly with one high friction node and one low friction node is easy.
I think that the reason why the first batch of organisms kept the lagging pink node is because perhaps they actually needed that bit dragging a tiny bit to act as a slight anchor point for contractions amongst the other muscles which worked towards the organism crawling? I think it might have been there to use friction to stabalise the organism and allow some of it's muscles to contract tighter, helping it crawl. Edit: the creatures at 300 in the second batch had trailing pink nodes also, except they just weren't trailing way behind by just one muscle. I think that a low friction trailing node must help with stability or acting as a sort of rear slight frictional anchor for the rest of the muscles to pull on.
I think the elements (dots and muscles) are too simple for really complex interaction. For instance, they have no way to synchronize with others so they just rely on correct timing by luck. And that's why it always degrades basically into the same simple structure, possibly doubled or containing redundant parts if forced to. I'm not sure what you were expecting, but if you try yourself to create a creature with complex movement method, you might find it impossible without adding some extra features for one or both of the elements. Such as some "nerve system" AKA trigger mechanism for the muscles based the state of other muscles. Or I don't know. I haven't tried. Maybe such a creature could exist within the current rules but it's just too unlikely to evolve in the few hundred generations. Or at all, if it would require development into completely other direction than the simple creatures.
I love how the first set of simulations lead to a creature with (mostly) four nodes consistently on the ground doing the actual moving, and one extra one way out the back that does have a node on the floor, but with essentially eventually near-zero friction (I would predict), which means it's as if it wasn't quite touching the floor. Basically it has evolved into a (mostly) quadrapedal (four-footed) creature with, effectively, a tail on the back for stability, and the frictiony nodes/feet being the back-most ones is eqiuvelant to having stronger back limbs providing most of the locomotive power. It all seems pretty close to real-life faster running creatures to me.
this feels like a giant version of one of those reality tv shows where there are challenges and you get voted off but if you win you get immunity that round
Gives the things the mechanisms of evolution; Hates the results and calls them cheating. Arbitrarily decides to prevent generational muscle atrophy and vestigial limbs because modern mammals look exactly like the proto amphibeans they descended from a half billion years ago. Yep, makes perfect sense. Of course their performance was sub par; evolution is about change and you kept strangling every change they made.
the reason nothing but twitchy jittering is produced is that that's all your physics engine supports. The bodies don't even have a property of mass, which is probably the central quantity in mechanics. I don't think you're gonna get a majestic galloping unicorn out of this unless you change it.
Actually the first run ended up doing one of the things you wanted. It eventually made use of some of those extra nodes by having a pair of high friction and low friction "legs" instead of just one. Also I'm pretty sure that tail is doing something, though I could be wrong.
Might I suggest using the absolute value of distance, so that the reverse direction also matters? Also that 'drag along pink node', is actually a 'wheelie bar' that is preventing the creature from lifting to it has more traction.
I rather liked seeing how you progressively put more limitations on the variables. The less freedom they had the less they could adapt. Kind of fun to see how trying to put your own ideals on a system does not workout as hopes. I liked seeing the experimentation. Thanks for sharing.
This is incredible! Dude how can you not like your new creature? It evolved a stabilizing tail. I'd say you wound up with a more complex creature than last time - pretty cool. Now I shall return to watching part 2
Hi, I like what you are doing and here are some things I have to say: 1) If you run the first version of the simulation often enough yourself, you may have noticed that 90% of the time the triangles are the final stable species. This is because triangles are the SIMPLEST, and this means it's easy to have positive mutation and easy to maintain positive mutation. This is reflected in nature as most species on Earth are bacteria. 2) A reason why the first simulator is not too riveting may be because all the organisms are test for the SAME NICHE--namely, all they are tested for is their ability to move to the right, whereas in nature, organisms perform various tasks and a species can excel in one task but avoid another task in which it is not superior. A "sand-box-y" simulator in which all organisms are present in the same environment and may interact with one another may allow some biodiversity, though it will take some creativity. 3) If you want the first simulator to have interesting (or more specifically complex) creatures, perhaps you should allow the user to manually kill a creature and replace it with an offspring of a liked creature. Of course, this means the simulation is no longer doing "natural selection", and the programming involved may be difficult, but it does allow the user to develop a kind of organism they like: for instance, I like complex creatures and would happily wipe out the triangles in order to remove competition (sorry triangles)
I love how the first creature is essentially the triangle, with the high friction on the left, and the rest of the creature is just there to pull it off the ground.
20:59 I think the pink node is actually serving to provide tension to hold up the node at the very top. If it turned white, that top node would fall forward and then all the nodes would be on the ground, I think, since the front nodes are white so they don't provide any force to hold it upright.
This is an awesome video, the creatures in the first simulation had a hanging tail but they also were using multiple feet. Best part was that the creatures in the last simulation were making wave like motions to move forward. You certainly undersold this video a bit.
Hi Carykh. I gotta say. Actually, the first part of the video is the most interesting from an evolution point of view. Because in the end, your creatures had 2 propelling limbs (I dare say legs) and a tail! How couldn't you see?! hehe. It was amazing. While the legs propel the creature forward, they depend on the front structure (let's say head) to elevate the high friction nodes up. And the tail, while having some friction, also has stiffer muscles that, attached to the head, help it to stay up high and to lift the legs. So there you go! A complex creature composed of a head, 2 legs and a tail. Also. The fact that every creature evolves into a mush instead of developing a structure is because they are basically composed of muscles and joints (nodes), with no skeleton. So in the end there is no real support for joints to stay put, since muscles will soften and flex over generations. The same would happen to humans if we didn't have a skeleton. We would suddenly become a mush of muscle twitching around ;-) I hope you can take your code back to that part 1 point in time and from there add components that mimic the function of bones. Regardless, congrats on the job. I found it to be really interesting.
The dragging tail seemed very important to those creatures. It was helping to push the "spine" nodes upwards, which pulls one of the "hind legs" up so that the other hind leg can make contact with the ground. I thought it was really awesome.
Have you tried putting an energy restriction on the creatures? One major reason real-life organisms can't solve problems just by spazzing out is that doing so is very energy inefficient. Perhaps that is what is missing.
I think this is the right answer for a couple reasons: all that movement across all those limbs is expensive like you say. The other consideration is that it costs a lot for organism to have additional limbs. Each extra node and limb should cost more in terms of energy not only for movement/basic operation but to mimic development of an extra limb.
+Jordan Cooper
Constructive criticism time:
An idea I just thought of is as follows; give every creature a certain amount of energy they can work with at the start. With every additional limb, they need to divide that energy among those limbs to sustain them. If the energy being sent to those limbs for a generation is below a certain threshold, then the limb should deteriorate as a result. This would lead to the most energy efficient and thusly the most likely to survive design at the end of the simulation. Now all we need is someone who can code this into a usable program and BOOM! Real evolution.
+Jacob Riley For the most part, I concur. However I disagree with the part about a creatures deteriorated body being passed on to the next generation. If somebody gets their arm chopped off, they dont get to have three armed kids. If having fewer limbs is favorable, it should be mutated from reducing the size of the limb or a random mutation to remove a limb.
@@personmcdudeguy Deteriorating a limb doesn't mean chopping it off! Deterioration means the gradual reducing of power/size over sucessive generations by mutation and evolution. Remember that in these simulators most of the slowest half of each generation get killed off, so if something managed to mutate an extra limb that caused it do have too little energy per limb then presumably it would be gone in the same generation purely for this reason.
As well as an energy restriction, a more realistic simulator would also need to have variable amounts of 'killing off' each generation. Specifically I would find a way of slowly increasing the distance that 'food' is away from the start point (an perhaps from the last 'food' etc), and all those that don't get at least that far would die of starvation. You will then have variable population sizes, as well as the possibility of extinction if you give them a challenge that's too difficult. You might then also need to allow for variable levels of reproduction to try and counter the occurences of near extinction. I guess allowing for variable 'killing off' might also lead to increasing the time limit, or lifting it completely, and then measuring both distance travelled and time survived for, and then judging them by a combination of both rather than just distance, but I guess the risk there is there could be a creature evolved who goes on infinitely, by always reaching food just before they would otherwise starve.
@@MrDannyDetail Make them have a constant maximum age at the start of the evolution. If a creature passed it's maximum age, it would die. Of course, the maximum age could be increased with mutations.
"GONE BORING" hell nah; this stuff is the best
Also, it's interesting how the super-restricted creatures at the very end almost started to resemble having a proper quadruped running gait (back legs push while front legs are up, front legs push while back legs are up). Of course, in two dimensions, rather than three. On that note, I wonder if the two dimensional limitation is responsible for the typical high-friction-back-and-low-friction-front arrangement you end up with when without as many restrictions.
I think the gone boring is a reference to gone wrong or gone sexual
Frez Person
+usavictor1 No shit?
20:38 That moment you realise it has a tail and two hind legs for propulsion
Shit just got real
i love them
@@lukesmith8896 yep that thing needs to survive if it can evolve a way to jump swim and run
You know what, I think the tail helped the creatures stabilize when running. That's probably why it kept the tail with a certain amount of friction on the ground.
That guy you ended up with around 23 minutes is awesome! The red and black node are alternating to basically walk while dragging the pink node behind it, keeping it from flipping over, and one of the middle nodes is also moving in time to use inertia to its advantage! Nothing lame about that! Great video.
23:00
(GONE BORING) (KIND OF DISAPPOINTING) (A BIT OF A LET DOWN)(NONSENSICAL GIBBERISH) (CONTAINS PROLAPSE) (THESE GUYS ARE LIKE TURTLES KIND OF)
kind of
I lol'd way too hard when the contains prolapse turned out to be actually in there and not you making it up.
You win.
contains prolapse? Lmfao wtf
(ALMOST DIED OF BOREDOM)
26:35
imo the taildragging creature is the best due to the tail, as despite it being dragged along, it provides some level of stability
i think so, anyway. it distributes the centre of mass and widens the contact surface area so that the creature doesn't just flip itself over
I think the reason your creatures are getting "scrunched up" is similar to why being small is an advantage for many real creatures: material strength doesn't necessarily scale with size. A cat can fall three storeys and walk away with bruises, while an elephant risks serious injury merely by falling over. There aren't any giant insects in this era because exoskeletons and lungless respiratory systems don't scale well. Basically if you want to end up with larger creatures you have to offer incentives for size: in the real world it's economies of scale and efficiency of metabolism and food storage, plus extra heft in a fight.
"Boring"... I honestly couldn't be doing something more interesting with my time.
Do some programming yourself.
"hello darknes my old friend"
I feel like Mykola miss read what you said like I did at first...
@@sadduck1737 misread*
@@sadduck1737 Afternoon miss led*
BTW don't leave your simulation running overnight. They will take over! They'll be our overlords. They'll make us run to the right on meter tracks and they'll go "Ooh, look at that. That one ran 26 miles then puked." "That fat one passed out after 5 miles."
Paul TheSkeptic. That's completley unrealistic... No fat guy could run 5 miles
Hey, yes I could. Well, maybe with some rests.
I think he meant meters, since the creatures ran meters.
xDD
Imagine a 2d creature Seeing 3d. they would ?
If you want to get rid of the standard 'high friction on the back, low friction on the front' creature, what you could do is make the friction of all nodes constant and unchangeable. The only way they can evolve to be faster is by moving more agilely , which seems to be your goal. It's also closer to how animals evolve, they can change their bones and muscles, but rarely do they rely on friction to improve.
Chrizz yeah you’re right!
As a human who is taller than wide, I found that I really related to this video. Good job!
.Could adding energy restrictions on amounts of movement limit twitching??? such as given 500 units of energy, go as far as possible!
That would be cool. Maybe moving thick muscles and high friction nodes takes more energy, too.
yea great idea.
You will see creatures shaking for few meters and then stopping i guess ... i think in nodes and muscles environment there exist only one optimal evolutionary design so unless he add some wheels, obstacles or whatever the outcome will be always very similar.
Thought about that one as well, but maybe even less restricting and more like a general balance factor by comparing energy used to distance travelled.
500 yums*
if you want to get graceful movement you need to not limit it by time but energy, give them some energy and wasted movement gets you less distance. while a slow careful movement might get them farther. right now the best ones are the ones most capable of pushing as much energy into their environment to move them as possible, they're sprinters but won't actually make it very far. to do this, they create a single driver and everything else goes friction-less to limit drag.
muscles that contract or expand twice as fast would expend 4 times the energy. You then also limit them to muscle movements that don't use an extremely low amount of energy.
also limit friction values since a near friction-less "node" is almost impossible in real life, even with vehicles.
Exactly what I was thinking he should see the motion of these particles with limited energy. It is exactly what happens in real life , a 100m sprinter tries to utilise all his energy in that 100 m .
I think is because all the 'muscles' contract simultaneously, maybe if you allow them to have multiple internal clocks it will make them have a more galloping motion
It had galloping motion in earlier simulations already.
It already has the muscles be set to different parts of the internal clock. They dont move at the same time
20:38 it just developed a tail!
Yeah, and that's a actually a really cool way of lowering its center of mass to ensure that its motor nodes cxcan stay on the ground.
Antoine Spartaphèze ikr
Antoine Spartaphèze yeah! It's like a spoiler on a car I was actually looking through the comments to see if anyone else figured that out!
It develops a second moving leg at around 20:55
+Antoine Spartaphèze That's exactly what I was thinking. Cary says it's not doing anything, but it's light pink, not white. Looks more like it's evolved the precise amount of drag to keep the whole body steady instead of bouncing around as it's "feet" move.
In fact, I'd be willing to bet that spike in the fitness increase was caused by the mutation/evolution of said tail.
You could try an experiment where you run a regular evolution, and also create a separate creature from scratch. Then for each evolution generation, you tweak the single creature a bit; only allowing yourself to change one value at a time. It could be sort of an Evolution vs Intelligent Design thing.
nice
Creation vs Evolution?
Well your model is great, but it only covers convergent evolution which is why at some point you get only copies of the same creature. If the conditions are always the same then the creatures are always getting more and more specialized.
What happens in the real world is that some species won the race not because they did something better, but because they did more things. They were able to survive in more adverse conditions in which the more specialized ones couldnt.
If you wish for your model to trully cover evolution as a whole you have to add variables that change for each generation, like for one the ground is stickyer, or the wind is blowing, the gravity is weaker, there's a slope, whatever, you will probably get to see more diverse forms that can deal with more situations that way. At least in theory.
Paulo Mangano I don't think the scenario should change every generation. Instead have a few separate scenarios and run the initial 1k through all of them. Then keeping them separate with a few minor variations to the scenarios see how the creatures change. I think that would be quite cool as earth has many different places and those places are home to animals that have adapted to that area's unique obstacles. Of course all this adds complexity and will greatly increase simulation time.
bensemus x i said every generation because change is often super gradual, and would probably provide more means to introduce variation over just a few vars that would not require that much of processing power, but in a real simulation it would be so gradual it wouldnt matter from a generation to the next.
Paulo Mangano I think it would be better for certain creatures have different conditions. not a change every generation
I think the main way to encourage divergent evolution is to provide different niches for the creatures to move into. Otherwise only one species will prevail, Gause's Competitive Exclusion Principle and all that.
The easiest 'niches' I could see for this model would be furthest moved forward, furthest moved backward, and most ststionary.
However, I expect the forward and backward moving designs would end up mirroring one another.
Especially since mutation can flip the direction of motion of a creature in a single generation.
Still... It would be 3 niches instead of one...
And thus we'd see divergence on some level...
I think the fact that every time you do these simulations they always end up look fairly similar means that they really are evolving properly.
In part 1 they actually cheated your system by making one node redundant... they only wanted 7 nodes!
I actually like the first part of this video a lot more, as evolution is about being random and unpredictable. Also at the last generation the most effective creatures had grown something like two "legs" that were speeding up the whole structure (black and brownish). Another interesting fact is that with previous evolution series creatures were training their skill in both going backward and forward, but with 8 nodes at the final generation almost every one is progressing to the right no matter what the rules were. Great video, can't wait to see more of those!
You should have changed it to *GONE ASEXUAL!*
I'm a mod of the TWOW subreddit but this is secretly the real reason I'm subscribed to this channel :$
I'd love to see this system used to produce creatures that are good at other tasks. like perhaps, climbing between two narrow walls
A couple of thoughts on this (not sure if you have drawn these conclusions since this video):
It is likely that the piston movement is the simplest , most efficient form of movement for creatures in this world. Thus all shapes and sizes will likely evolve towards being able to perform it.
If you add restriction , you are essentially restricting how possible evolving towards this movement type is, thus you will see worse performing creatures the more restrictions you place. (Essentially you are adding a second fitness function: "Do these creatures look like i want them to?")
If you look at the way slugs and such move they, do adopt this sort of movement , just in a less effective way than the creatures you have evolved.
The way bipeds often walk is essentially two nodes of high friction and then a node structure above them to move one of the "feet" nodes in front of the other. This is less efficient than the "piston like" movement your creatures have adopted. The reason why bipeds evolved to walk this way however is they have more tasks than "walk right on a horizontal plane as fast as possible".
My suggestion might be to introduce slopes or steps in the ground and see if the creatures can evolve to over come those tasks. This may lead to non piston-like movement.
I think the reason they always evolve into twitching heaps is because of the way nodes and muscles can freely move through each other. Also because the muscles always have to be moving.
This new Evolution Simulator video was definitely worth 35 minutes of my time. Good job as always Cary :)
I think one reason why creatures end up like they do, is because they lack bones and there is nothing preventing them from collapsing like they do.
Since only factor is "how far right can you go", well, once the creatures get certain setup that works, everything else just... well, they have no use, but since simulation does not allow those parts to disappear, they just dangle around.
I know it's a 5 year old video but the results in the first half actually make sense since the most efficient form of locomotion under this condition is the triangle as shown in previous videos, so no matter how many nodes you add, it will still evolve towards the triangle and any additional nodes will just become vestigial.
I don't know how to code, so may I propose that when a creature passes 15 meters it goes to an higher level and will reproduce and adapt to it's new environment. this will require to simulate more creature, but it will created more complex creatures and more fun.
Some levels could select the creatures based on their size, their ability to pass obstacle like bumps and holes.
the levels should not change to fast because a to big changes may destroy the creatures not make them stronger.
If the simulation runs really long I predict some creature will have something like 20 nodes.
I'm quite impressed with the results of part 1. The double node crawling action is actually pretty complex. I also think that trailing pink node is used as a balance - it's helping keep the creature in position against the floor.
The best creature at 1170 gens was a bit more interesting. It was basically bipedal compared to the tringle. Instead of one black node (unipedal) it had two appendages alternating on the ground (bipedal) so taking advantage of the wasted time in between each single step the tringle achieved. That was very interesting to me. Because there's no balancing organs it's just a bipedal creature pushing its head along the ground in front of it.
I really want to download this, buy a server box, and let it go for a week straight.
Have fun with your powerbill :)
I'm pretty sure that problem would be great to put it on a GPU.
you don't download it, it's online :)
frogonlilypad on what website
I haven't seen the whole video yet, but... try hurdles with the homologous structure species! Maybe that would force them to stand up!
It'd be interesting to see if any "tumblers" evolved, and did flips over hurdles, like a slinky.
Okay, I watched the whole thing... I kinda still want to see this. It might encourage actual strategic movement!
You could download the source code and try it out.
I don't think you can make species that are only made of muscles stand up. Ever. They don't have a brain or a computer to balance themselves. They would need a basic AI that re-balance themselves if their average angle gets messed up.
tenedria As said by Ian Malcom from Jurassic Park, "Life finds away."
Even if that life is a computer system forcing a group of lines and dots to change...
I think that dragging pink node is really cool because you see completely useless structures in animals all the time due to the fact that an ancestor had it (like an appendix possibly). Also, the fact that they evolved to look like the original creatures from previous videos is an amazing showing of convergent evolution, to have started totally different and end up evolving to be so similar. So cool. Love the videos, thanks!
Can you release Evo. Sim. to the public?
Look at Part 1/4 of the evoloution sim series. It has it in the description to a website that plays it
Thanks!
FSG he did
Vladimir Tiffany he knows
FSG #FSGOUT
it makes sense that your last try didn't work as well. caterpillars aren't meant for speed, they are meant for having a point of contact on a walking surface at all times. your other creatures also make sense, they resemble a kangaroo, rapidly pushing with their feet and staying balanced with their tail. I found this all interesting and very logical.
Make a sandbox mode where you can build your creature and watch it evolve!
That’s not how the evolution works; it takes the better half of 1000 creatures and then takes each one and duplicates it twice with minor changes. That means you’d have to design 1000 creatures.
U can have 1000 copies of that creature an dlet them perfect the muscle contraction
The pink tail might not be a bug, but a feature. It pulls on that top node, adding a rotational force to the whole thing, which might act similar to a spoiler on a car.
let the creatures do different tasks, not just walk in a flat surface. maybe add a knee high "water"layer where the space has friction. this could encourage the blobs to stand up.
I like how they perform worse when you put more and more restrictions on them. Go evolution!
I have watched a lot of your videos and they are great! Good job and I love the commentary!
One thing I see with almost all simulations on your videos is that they are heavily dependant on the initial performance. The twitching creatures are far faster and easier to evolve in a short period of time. Some other interesting ways of travel may be slower to evolve in comparison. I would assume your code heavily benefits the "twitchers" for the first generations and therefore you find a local maxima, quickly, and your gene pool only consists of twitchers. There might be other maxima but as they are quickly "killed" we never get to see them.
This would explain your problem with the "twitchers". A solution is to classify species as nature does: some species evolve slower but still maintain to survive and eventually get to be faster than other species. The code implementation would be interesting and a bit tricky with different groups of species. But you are really good at what you do so I would love to see the results! (!)
I love these videos and I love your commentary. You're ridiculously smart, but you're so humble that you speak exactly like a dummy like me, so you're both personable and like an expert. It's like talking to a really cool doctor.
I think the issue with your simulator (why the creatures are twitchy and unrealistic) is that you are simulating motion in two dimensions. Real creatures have evolved to move in three dimensions. Hence why the look the way they do. In your simulation, the twitchy creatures succeed because that motion is the most efficient for two-dimensional motion. If the universe were two-dimensional, I think real creatures would be more twitchy and erratic like those simulated.
Niklas Leet lets get a 3D evolution sim YAY
I would be really excited to see an evolution video where the creatures are not just competing for higher scores on one fitness function, but on two or more(For example, a setup where creatures are rewarded both for running forward and for jumping as high as possible). There are a few reasons I would find this interesting:
1)It would encourage higher biodiversity even at the late stages. Instead of just one species taking over because it's the best at whatever function you gave, there might be both runners and jumpers, and species that aren't great at either but moderate at both, and everything in between.
2)It might lead to multiple types of strategies being tried even within one of the functions. For example, a population where there are both shakers and wheels that can move forward and are competing with each other. I think this could arise as a result of the increased freedom of change in the late stages of the game, as creatures would be able to trade-off between running and jumping freely. A family of organisms could go from running, to jumping, to running again, but with a different strategy.
3)It would contribute to increased realism, as real life ecosystems rarely only have one goal that organisms can pursue, instead having many different ways to secure reproduction for themselves.
MrSporeowns The main reason I came up with that idea was because I was getting bored with the really low lategame biodiversity, as one species takes over because it's the best. Another idea I had for reducing this effect was some system of dependant relationships between species, so that one species can't just kill off the other without dying out themselves(for example, a predator-prey, symbiotic, or parasitic relationship). This idea may be helped along by requiring organisms to gather energy, because that's the point of predator-prey relationships between organisms.
I would tend to think(though couldn't know until further experimentation) that a more productive fitness function would be some sort of weighted addition of the two goals of running fast and jumping high, because if you multiply the two functions together, then generalists would have an advantage as specialists would have near zero score on one of the functions.
I think, however, that the multiple fitness functions idea would be the easiest to implement at this stage, as well as having the lowest increased load on the processer, since it adds no new mechanics, just changes the fitness function, and fitness function evaluation accounts for a relatively low amount of computation compared to the actual physics simulation. Food pellets might also be easy to implement, since no new physics are really added, but I don't know how one would distribute food pellets around the arena for the best effect. Interspecies relationships would probably be hard to implement because it would require multiple organisms to occupy the same arena and I'm not quite sure how that would be set up at all.
20:50
IT DEVELOPED A SECOND BACK LEG FOR MOVEMENT!
Anybody else noticed that?
You actually all ready had homologous structures. When you did your simulations one of the body structures would end up dominating the ecosystem, and the structure that dominated any ecosystem was based on what was the most advantages, just like how homologous structures come to be in the real world.
The creature at 12:00 looks cute. It's like a little snail with triangle wings. Don't think of the nodes as intersecting, but rather in 3D.
Your part one creature was boring? You evolved a creature with two front legs, three back legs, and a DAMN TAIL! I don't know what awesomer creature you could've hoped for.
It is said that evolution can be extremely slow, and extremely fast. While these videos due showcase this, there simply isn't enough variables. Use your original concept, of whatever evolves best "wins". However, wouldn't it be neat if 5% of the population, maybe every 5 generations (with a variable for the exact amount of generations (written as 5 plus or minus 2)} that 5% got 3 more nodes, or some other "random" occurrence? Maybe, if the program detects stagnation, it throws in a couple of hurtles, and maybe a slope. Have different obstacles appear, disappear, or even just move from one generation to the next. This would help simulate a real environment. Imagine if there was just a small hole in the ground, even a shallow one, forcing more evolution. Making the situation for these creatures as chaotic as possible, much like real life, would make this far more interesting (right now, I love the concept, and love watching this project evolve [pun not intended]). I realize this would require a LOT of coding, and several changes to the program itself, but I imagine that is what you really want out of this project, and I know that that is what I, as a viewer, want to see.
That final example of the "3 box" structure demonstrates undulatory locomotion, which is exactly how caterpillars and worms move through the environment. I don't know how else to put this, but your program discovered a natural method of movement. You have made great progress! Congratulations!
WHERES YOUR SPECIAL RELATIVITY VIDEO PART 3 AND 4?
Ikr
The first part is interesting to me. The creatures developed a 2-feet-1-tail system, and I'm curious about what the tail does.
*Why not make nodes able to rotate their muscle and whatever is attached?* Please like so he sees this
John Doe I know I'm late but muscles able to rotate nodes are better because they are natural wheels
thanks for coming out with a new one, ive been waiting 11 months for this. hope you do some more!
To be fair, I thought that the creature that evolved in the first part was somewhat interesting. It had two propeller nodes in the back where as the creatures in past parts always evolved to have just one. You dismissed the pink node "tail" as a mistake yet since it lasted for so many generations it seems that it was intentionally kept as some sort of stabilizer. There were still some useless nodes in the middle, though I was pleasantly surprised that more than three ended up being utilized.
Also, when dealing with creatures that all have the same structure, there may be a way in which the species chart can be utilized. Since the creatures reproduce by themselves, it might be neat to see what percentage of the remaining creatures originate from each individual from round one. Actually, that would still be interesting to analyze when there's multiple species as well. I think it would be fascinating to see how many remain just after 15 rounds and whether or... (continued in reply)
...not only one remains by the later generations.
Finally, I'm interested if you would be able to add another variable in. To start, I seem to recall that the muscle movements are based on a command loop. I'd be interested to know whether that loop could be be sped up and slowed down as a mutation. Also, it would be interesting to see if the muscles could be expanded and contracted multiple times in that loop, but at different rates.
Also, I support the idea of automatically flipping around the individuals that end up in the negative numbers. I've seen before that sometimes the creatures just get entirely flipped around by a single node being moved. That's why in the earlier rounds the ones moving in the wrong direction still had the same amount of movement as those who went right. By killing those who get flipped around, it essentially punishes otherwise perfectly reasonable changes while limiting the number of viable mutations. Because of that, if the creatures are flipped it might result in more varied designs. To be fair, it could have almost no effect though I don't see any harm in trying.
Should make it so the more there are of a certain type, the higher the chance of mutation. Might prevent some plateaus?
Sorry if someone already said it, but that pink node dressing behind was serving a function. The fact that it stuck around for a hundred generations means it was better than reducing its friction, and then it continued until the end, with over a thousand generations it had to be doing something useful.
evolution is about getting the most efficient creature not the coolest.you have unrealistic expectations.
Thanks for making these videos, these have taught me a lot about evolution and I usually watch these before i sleep as they calm me down. Anyways keep up with the great work and videos!👌👌 (you should also do this with the vertical jump)
You should Aldo Bump The road , this would avoid the smalls structures and should help more "walking" structures.
If tou do that, maybe don't do a very bumpy road since the beginning, maybe go from a quite linear road for the firsts generations to a very spiky one for the avanced generations.
5 years ago. I began big dive ai with you and ended with you.
Each time you add more restrictions, the performance gets worse. Clearly you need fewer restrictions, not more.
he problem is if he put no restriction he got a dominent 3:3 specie
And actually each time he put restrictions , it only reproduce the same pattern but with useless elements evolution try to remove
Well its not about restrictions it is about Coal. the problem of not coming up with new ideas but it seems that Moving fast to right has an efficient layout.Bushing with Rear Legs while front legs are for support. it allso works IRL for example Cats or Cheetahs. or even Antilopes they all have strong back legs that repel them forvard while front legs are for Balance and support.
+mayaknife Seems lik eintelligent design won't be so intelligent after all ...
i have an idea why the creatures always default to the high friction on the left and low on the right. one very important aspect of evolution of a species is the environment they live in. for example, if the species is found in a desert, then of course they wouldn't have developed fins or gills! my guess is that the current default structure is the most efficient on the straight, simple runway that you have now. but if you added an extra bit a randomisation where the ground friction or the gravity of the world can change then you might get different results. it also might be cool to add in some kind of barrier, like water or bushes that slow them down, but don't stop them completely!
at 21:00 you can see that a single node is being dragged along on one muscle.
I thought you made restrictions so that the muscles and nodes had the same connections?
In the original double-square formation every node had at least 3 muscles connected. How could a creature have one node dragged along on one muscle if you made homogenous restrictions?
Oh, forget it. I watched the whole video XD
it's muscles got so weak you can't see them
I actually really like the creature you ended up with in part 1. It's actually surprisingly animal-like when you think about it: two front legs for balance, two back legs for propulsion, and a tail dragging on the ground behind it.
You should take the absolute value of the distance and reverse the muscles to even it out, you might consider this cheating but i dont
I think the reason you get the worst reverse going creature going almost as far as the best forward going creature is because the simulation is dominated by its starting conditions. A creature that goes far in the forward direction will go far in the backward direction. The reason it goes forward or backward is due to its initial topple.
Mathis o
Six years late but: What if instead of nodes having differing friction, they had differing weight instead? It'd allow you to also see if they would evolve differently with ice physics
If this was an app that would be cool
If every 10 generations is about 500 years, then you'd need 1,000,000 generations to get to 50 million years which is more on the scale for the Theory of Evolution. I'd love to see the end of your videos have something on that scale, even if it does plateau.
Finally!
Hi THT
+Rony McJony heya
+Rony McJony also, slap
oh no, its you
+JC7146 yes. It is
My recommendation would be to return the ability to have morphological mutations, but just make them rare so there are particular morphology has time to evolve more fitness within that morphology. Changes in morphology that are beneficial anyway will tend to survive, but the more fit a creature gets with a specific morphology, the less likely a change in morphology will be to benefit that creature. I know you wanted to simulate what is seen in natural biology, but the artificial life forms you're evolving are not biological and keep in mind that biology took a long time and an awful lot of generations to evolve the diversity that it currently has. By the way, you mentioned homologous structures in mammals, but the slide you're showing at the time had one non-mammal in with the three mammals it was showing homologous structures on. It's also worth noting that those structure is being homologous doesn't mean that they haven't lost or gained any bones or muscles. Not all mammals have the same number of toes, for example. If you haven't already discovered this for yourself, I think you will find that evolution is much more interesting when you don't try so hard to make it turn out a certain way. Let it surprise you.
Add attractiveness based on symmetry
I think the problem here is that since the time limit is so short, mad-dashing in an unsustainable manner is the most effective strategy, and the best way to do that is to be compact and pulse erratically.
I'm not a human
Oh hi FryUaj
Oh hi Lyndon Choong
(omg Senpai noticed me!)
lol
You're profile picture tells me you're a lemon.
As much as we want to see some upright locomotion, it's actually much more complex that one would initially think. Sure, most people reading this could stand up right now and not think twice about it, but balancing on two feet requires a lot of coordination to do. That's why many of the creatures in this simulation opt for the path of least resistance: walking is hard, crawling is less hard, jittering randomly with one high friction node and one low friction node is easy.
"it's super slow! but it this is the price you pay!" *proceeds to skip forward a few seconds*
I think it'd be interesting to allow the user to input a creature so that you can test theories on how some of these creatures move.
I've been waiting this for so long yay! Nice job as always, kind of sad evolution didn't come up with some fuzzy nice big tall hairy legs
I think that the reason why the first batch of organisms kept the lagging pink node is because perhaps they actually needed that bit dragging a tiny bit to act as a slight anchor point for contractions amongst the other muscles which worked towards the organism crawling? I think it might have been there to use friction to stabalise the organism and allow some of it's muscles to contract tighter, helping it crawl.
Edit: the creatures at 300 in the second batch had trailing pink nodes also, except they just weren't trailing way behind by just one muscle. I think that a low friction trailing node must help with stability or acting as a sort of rear slight frictional anchor for the rest of the muscles to pull on.
I think the elements (dots and muscles) are too simple for really complex interaction. For instance, they have no way to synchronize with others so they just rely on correct timing by luck. And that's why it always degrades basically into the same simple structure, possibly doubled or containing redundant parts if forced to.
I'm not sure what you were expecting, but if you try yourself to create a creature with complex movement method, you might find it impossible without adding some extra features for one or both of the elements. Such as some "nerve system" AKA trigger mechanism for the muscles based the state of other muscles.
Or I don't know. I haven't tried. Maybe such a creature could exist within the current rules but it's just too unlikely to evolve in the few hundred generations. Or at all, if it would require development into completely other direction than the simple creatures.
I love how the first set of simulations lead to a creature with (mostly) four nodes consistently on the ground doing the actual moving, and one extra one way out the back that does have a node on the floor, but with essentially eventually near-zero friction (I would predict), which means it's as if it wasn't quite touching the floor. Basically it has evolved into a (mostly) quadrapedal (four-footed) creature with, effectively, a tail on the back for stability, and the frictiony nodes/feet being the back-most ones is eqiuvelant to having stronger back limbs providing most of the locomotive power. It all seems pretty close to real-life faster running creatures to me.
this feels like a giant version of one of those reality tv shows where there are challenges and you get voted off but if you win you get immunity that round
Have a predator chase them? So two evolving roles? Predators and prey. When a predator overtakes the prey, the prey is eliminated?
Gives the things the mechanisms of evolution; Hates the results and calls them cheating. Arbitrarily decides to prevent generational muscle atrophy and vestigial limbs because modern mammals look exactly like the proto amphibeans they descended from a half billion years ago.
Yep, makes perfect sense.
Of course their performance was sub par; evolution is about change and you kept strangling every change they made.
the reason nothing but twitchy jittering is produced is that that's all your physics engine supports. The bodies don't even have a property of mass, which is probably the central quantity in mechanics. I don't think you're gonna get a majestic galloping unicorn out of this unless you change it.
Actually the first run ended up doing one of the things you wanted. It eventually made use of some of those extra nodes by having a pair of high friction and low friction "legs" instead of just one. Also I'm pretty sure that tail is doing something, though I could be wrong.
Might I suggest using the absolute value of distance, so that the reverse direction also matters?
Also that 'drag along pink node', is actually a 'wheelie bar' that is preventing the creature from lifting to it has more traction.
I rather liked seeing how you progressively put more limitations on the variables. The less freedom they had the less they could adapt. Kind of fun to see how trying to put your own ideals on a system does not workout as hopes.
I liked seeing the experimentation. Thanks for sharing.
This is incredible! Dude how can you not like your new creature? It evolved a stabilizing tail. I'd say you wound up with a more complex creature than last time - pretty cool. Now I shall return to watching part 2
Love how even the last version was still doing the same kind of movement, just with the middle box floating in the air.
Hi, I like what you are doing and here are some things I have to say:
1) If you run the first version of the simulation often enough yourself, you may have noticed that 90% of the time the triangles are the final stable species.
This is because triangles are the SIMPLEST, and this means it's easy to have positive mutation and easy to maintain positive mutation. This is reflected in nature as most species on Earth are bacteria.
2) A reason why the first simulator is not too riveting may be because all the organisms are test for the SAME NICHE--namely, all they are tested for is their ability to move to the right, whereas in nature, organisms perform various tasks and a species can excel in one task but avoid another task in which it is not superior.
A "sand-box-y" simulator in which all organisms are present in the same environment and may interact with one another may allow some biodiversity, though it will take some creativity.
3) If you want the first simulator to have interesting (or more specifically complex) creatures, perhaps you should allow the user to manually kill a creature and replace it with an offspring of a liked creature. Of course, this means the simulation is no longer doing "natural selection", and the programming involved may be difficult, but it does allow the user to develop a kind of organism they like: for instance, I like complex creatures and would happily wipe out the triangles in order to remove competition (sorry triangles)
I love how the first creature is essentially the triangle, with the high friction on the left, and the rest of the creature is just there to pull it off the ground.
20:59 I think the pink node is actually serving to provide tension to hold up the node at the very top. If it turned white, that top node would fall forward and then all the nodes would be on the ground, I think, since the front nodes are white so they don't provide any force to hold it upright.
I feel like he should make random mutations more frequent, then they’d stagnate less.
This is like the 5th evolution video I have watched. Subscribed
I'm sure that if you kept it going for a really long time, they would eventually start using all of their nodes
This is an awesome video, the creatures in the first simulation had a hanging tail but they also were using multiple feet. Best part was that the creatures in the last simulation were making wave like motions to move forward. You certainly undersold this video a bit.
Hi Carykh. I gotta say. Actually, the first part of the video is the most interesting from an evolution point of view. Because in the end, your creatures had 2 propelling limbs (I dare say legs) and a tail! How couldn't you see?! hehe. It was amazing. While the legs propel the creature forward, they depend on the front structure (let's say head) to elevate the high friction nodes up. And the tail, while having some friction, also has stiffer muscles that, attached to the head, help it to stay up high and to lift the legs.
So there you go! A complex creature composed of a head, 2 legs and a tail.
Also. The fact that every creature evolves into a mush instead of developing a structure is because they are basically composed of muscles and joints (nodes), with no skeleton. So in the end there is no real support for joints to stay put, since muscles will soften and flex over generations. The same would happen to humans if we didn't have a skeleton. We would suddenly become a mush of muscle twitching around ;-)
I hope you can take your code back to that part 1 point in time and from there add components that mimic the function of bones.
Regardless, congrats on the job. I found it to be really interesting.