This video is about evolution.Not the biological evolution you’re familiarwith, where living organisms adapt to theirenvironment.No, this is about virtual evolution.About discovering the extent to which videogames and other simulations can shatter thenatural order and let us experiment with thebuilding blocks of life itself.…And the extent to which they can make uslaugh.The ambitious dream of using computers tosimulate life is an old one, and you mightbe amazed at just how far the technology hascome.So, for this entry into the archive, we’lltake natural selection into our own hands,and see how far we can push this concept beforethings get out of control...Our story begins in 1970, when British mathematicianJohn H. Conway unveiled a game simply called‘Life.’The rules are simple: within a blank void,you can create tiny pixels called cells, whichwill reproduce if there are enough cells nextto them, but die out if they get overcrowdedor end up alone.Simple, right?Well, you activate the simulation, and then…Using these rules and basic starting patterns,Conway’s Game of Life can create unbelievablycomplex chain reactions that legitimatelyfeel…well, alive.Pretty quickly, people discovered that dependingon your initial pattern, specific “species”would emerge with their own unique behavior.Microbe-like colonies of pixels that appearedto move were nicknamed “Spaceships.”So called “Puffer Trains” traveled alongand left behind a trail of debris.“Rakes” left behind trails of fully functionalSpaceships.This pattern called the Rabbits starts small,but expands for a ridiculous amount of in-gametime before stabilizing.It’s a lot like watching cells under a microscope.Of course, the pixels in the game of lifearen’t actually ‘alive’ by most definitions,but they’re an incredible starting pointthat show just how effective simple rulescan be at creating complex emergent behaviors.It’s proven to be a game with almost limitlesspotential.But we can take things further.Life Engine is a more recent game by the amazingyoutuber Emergent Garden that simulates notjust emergent behavior, but natural selection.Like Conway’s game, Life Engine is all aboutpixels, but here, different colored pixelshave different functions.Orange pixels act as a mouth, allowing lifeformsto consume energy.Green pixels produce energy like plants.Blue pixels can move about like animals.Pink pixels do damage to other lifeforms,and purple pixels act like armor against damage.Finally, grey pixels act as eyes, allowingthe smartest virtual lifeforms to avoid eachother.When you begin a game in Life Engine, it alwaysstarts with a mass of stationary lifeformsslowly expanding out from the center via reproduction.Yet each time a pixel reproduces, it has arandom chance of changing color, simulatinga genetic mutation.As with real world natural selection, someof the mutations, like movement, prove advantageous,and give that specific lifeform more opportunitiesto reproduce.In almost every game, fast moving lifeformsquickly overwhelm the stationary plant-likecreatures, and then settle into an equilibriummuch like on Earth.From here, this simulated petri dish becomesa battleground, with different species evolvingweapons, armor, and evasive maneuvers to outcompete each other.A similar youtuber-created game is The Bibites,a simulator where endearing virtual creaturescompete for food.This game also works on a system of randommutation and natural selection, with the titularBibites getting more and more complicatedthe more you play.There’s more both games than this, and there’sa link to them, and all the games and experimentsdiscussed in this video, in the description.There are a lot of incredible games in thiscategory of cell-like evolution simulators.Like the highly underrated Cute Mold by Erytau,which lets you simulate the growth of differentslime molds.Or the remarkable Clusters by algorithmicartist Jeffery Ventrella, that renders a microworldof particle systems that begin to displayincreasingly complex behaviors not due tosome hidden AI, but purely as the result ofa system based on the laws of physics.Many of these games are inspired by Conway’sLife in some capacity, but all have theirown unique spin.There’s one final computational model thatdemands attention.Developed by scholar Bert Chan in 2015, Leniais an award-winning program derived from Conway'sGame of Life that uses continuous generalizationto make everything smooth and ethereal.Using high resolution patterns, Lenia producesfar more complicated types of digital life,with over four hundred individual speciesdiscovered.Lenia patterns have displayed an impressivenumber of the traits we typically associatewith biological life, including bilaterallyand radially symmetrical body plans, reproduction,and even intercommunication between colonies.Chan has even organized these patterns intoa taxonomic tree of evolution, some of whichare so advanced some have called them ‘MathematicalLifeforms.’And sure, by most definitions these patternsaren’t truly ‘alive’ — they’re justthe result of rules interacting in uniqueways.But in a way, our universe and all the incrediblesystems of life within are also the resultof rules interacting in unique ways.And these simulations are only getting moreadvanced…Following the path of life on Earth, it’stime to leave the virtual tidepool of cellsimulators, and take our first, very awkwardsteps onto dry land.Evolution is a game by programmer Keiwan thatallows you to build digital animals by connectingjoints with bones, and bones with AI controlledmuscle tendons.The game’s neural network will then tryits absolute best to get the creature to walkforwards over multiple generations, with eachattempt appearing as a little ‘ghost’next to the current model.At first, most simulations are an absolutemess of failed attempts, but with enough generations,the neural network will eventually get betterat moving, even if it’s not in the way youexpect.And that’s because this AI doesn’t knowwhat ‘walking’ looks like, which is why,at first, it’s quite bad at it.Every movement it learns is through trialand error — essentially running off itsown version of natural selection.And with enough practice, these simulatedcreatures can evolve movement patterns thatare reasonably close to the real thing.And it’s not just walking that this networkis capable of.Depending on the design of your creature,it might learn to roll, or hop around likea kangaroo, or even take to the skies.And there are more challenging tasks availablethan walking in a straight line.You can try to make the poor creature climban infinite staircase, or jump over fast movingobstacles.While at times hilarious, there is a strangetragedy in watching these innocent creaturesstruggle against gravity.Something about this game really capturesthe unfeeling nature of natural selection,where advancement often involves countlessgenerations of failure.Though cleverly simplified, this is, in manyways, the story of life on Earth.A similar game to Evolution is Evol Pedal— a walking simulator created by Evol Gameswhere a lifeform made up of randomly mutatingblocks learns to travel forwards.The game also has an amazing ‘3D version’where a CGI model of a human will learn towalk by any means necessary, which can haveboth amusing and unsettling results.With over 100 moving parts, it’s difficultfor the AI to stay coordinated.You can also adjust the settings to allowfor non-human movements like knees that bendbackwards, which often leads to even moreuncanny sights.The game even lets you fully adjust the gravity,and pushing the setting too far can lead tothings breaking down a bit.Such is the price of glorious, glorious evolution…But maybe we’re taking too many steps tooquickly.After all, life on our planet didn’t firstevolve to move on land, but in the much moreforgiving environment of water.In 2016, a group of scientists used artificialevolution to successfully produce swimming,soft-bodied creatures.These simulated lifeforms were made of squaresof virtual tissue, with green representingactively moving tissue, and blue resentingpassive tissues.The AI found that using a combination of bothresulted in the most efficient swimming methods,producing a range of body types convergentwith various forms of life in our own oceans.And to study the relationship between swimmingand walking, some of these creatures weremoved onto land, where they found clever waysto use bodies evolved for aquatic motion tonavigate a terrestrial setting.These experiments essentially created virtualintertidal life, offering intriguing insightinto the evolution of early land animals.That same year, another group of scientistsran a similar soft-bodied creature experimentwhere a network evolved a body from a singleinitial cell.Each evolutionary run produced something different,but many looked remarkably close to biologicallife.It's worth remembering that this AI didn’tknow what a leg or a tail is — it just founda way to keep balanced that is convergentwith life on Earth.On the subject of muscle-based locomotion,one of the most well-known experiments comesfrom 2013, with a 3D simulation of variousbipedal creatures.These models learned to walk through trialand error over multiple generations.No motion capture or keyframe animation wasused, with the impressive final results theproduct of AI learning.The examiners then pitted this AI againstdifferent target speeds, which led to thecreatures adjusting their gait accordingly.They even put them through obstacle courses,throwing virtual boxes at them and addingrandom slopes to their environments.While sometimes they’d succeed in catchingthe AI off guard, the overall results werestrikingly consistent.A more recent experiment presented at theICLR conference used an analogous method toteach different models to walk.Many of these results resemble a three-dimensionalversion of one of the evolution video games,as both are attempting to simulate similarthings.But these scientific simulations are not gamesas we understand them, although this technologymay one day be used to make virtual entertainmentmore lifelike.As amazing as all these simulations and videogames are, for me, there’s always been thatfaint dream of the one unified game, thatcombines cellular models with aquatic andterrestrial locomotion to create the ultimateevolution-simulating masterpiece.For a lot of people that dream once had aname — and its name was ‘Spore.’Released in the late 2000s, the buildup toSpore was, to put it mildly, completely nuts.As early as 2005 demos of this hyper-detailedprocedurally generated universe simulatorwere being shown, and it looked incredible.There was a sense among some before its releasethat this game, which seemed to merge sciencewith infinite customization, would changethe industry forever.And then the game came out.While Spore did impress with its creativityand ambition, a lot of people felt the gamewas significantly shallower and less, uh,scientific than they were led to believe.To be clear, I’m still quite a fan of thisgame despite its eccentricities — just lookingat it gives me a huge hit of nostalgia.And it does explore a number of the principlesof evolution and biology in its own uniquemanner, which I’ve made two separate videosabout.But it certainly takes a more cartoonish approachto the stages of life, and some features advertisedlike an underwater multicellular stage weren’tin the final game.For many, it wasn’t the comprehensive evolutionsimulator they were hoping for…Yet the dream of Spore hasn’t vanished justyet.Recently, there has been a surge of new gamesthat seem to take some of the concepts ofSpore and push them further.One game on this exciting new frontier isAdapt by Paul Hervé.Currently a work in progress, Adapt is a gamethat lets you build your own lifeform muchlike Spore’s creature stage, although ithas a greater emphasis on biological plausibilityand finding your niche through careful evolutionarychoices.And unlike Spore, Adapt lets you build a creaturespecially adapted for a semi-aquatic lifestyle,which is a neat addition.The game is currently free to play on Steam,and is a promising addition to this emergingsubgenre.One evolution simulator that really impresseswith its sheer amount of detail is The Sapling.Created by Wessel Stoop, this polygon-artsandbox game lets you design every organismin an ecosystem, and then simulate their interactionsover multiple generations.And I do mean every organism — unlike mostgames in this genre, you’re not just designinganimals, but plants too, and even things likealgae.And the depth of these systems are absolutelymind-blowing.When placing a plant, you need to considerhow many leaves it has relative to the sunlight,and the length of its roots, and the amountof ground water in the area, and the soilsoftness, and the temperature, and the windstrength.Different plant colors realistically growbetter or worse depending on the type of staryour planet orbits.There’s a full system of flowers you canadd to plants that produce different amountsof pollen that affects reproductivity.You can also use bioluminescence in plantsto attract pollinators — or use bioluminescencein animals to mislead prey.In terms of creatures, there are flying lifeforms,detailed customization options for aquaticlife, different eyes that can see differentranges of infrared and ultraviolet light.You can even adjust the instincts of eachanimal, using a modular system of commandsto change how they react to different creatures,sounds frequencies, and specific colors.It’s absolutely off the chain.Especially when you consider that like Adapt,The Sapling was developed by just one person.Compared to Spore which was made by a goliathof a team, these single-developer games ofcourse don’t have the same resources behindthem, but the incredible results show thatthere are still passionate people who believein this idea — this dream of a game thatcould simulate life like never before.And if we’re going to talk about dreamers,we have to talk about Thrive.Beginning development over a decade ago, Thriveis a game quite unambiguously born out offrustration with the final product of Spore.A communal project open to all who wish tocontribute, the game takes a far more ambitiousand scientific approach to Spore’s gameplayloop.In the game, you start as a single, realisticcell in a tidepool.As you advance, your cell splits into a complex,multicellular organism, becoming a 3D modelthat can be fully shaped by the player, withdifferent organs serving unique functions.Over the next few stages, this creature growsin size and intelligence, eventually ableto communicate using a multifaceted languagesystem and create technology.Until eventually the technology compoundsto the point where you ascend to a god-likelevel of power and can control the entiregalaxy.It’s all so cool.It also doesn’t exist.Since Thrive a crowd-sourced game, anyoneon the internet can come up with ideas forThrive, but actually implementing those ideasis another story.Most of the images people share of the game’slater stages is actually art by the talentedSciocont.Right now, over a decade after the game’sinception, the only playable stage is thefirst one: The Microbe Stage.And to be clear, this initial stage is veryinteresting.It’s an exceedingly complex and realisticcell simulator, where instead of eating cartoonplants, you’re balancing your Glucose, Ammonia,and Phosphate levels, in many ways makingit a game about virtual chemistry.While this attention to scientific detailmakes the game fairly difficult, it’s atruly admirable product.Another cool thing about Thrive is that whileyou build and design your cell, other cellsin the environment randomly mutate and adapt,with any given game creating hundreds of interconnectedmicrobe lineages, all with their own randomlygenerated scientific name.Most people seem to be aware that the otherstages of Thrive aren’t coming for a while.But it is pretty inspiring to see so manypeople drawn in by this idea, and shows thatthe dream of the ultimate evolution simulatorhasn’t gone extinct just yet.Before this journey through virtual evolutionconcludes, there’s one last thing I haveto show you.A simulation that does something incrediblenot by thinking big, but by thinking small.Really small.OpenWorm is an international open scienceproject with the sole purpose of simulatingevery cell of a type of worm that grows justunder 1mm in length.That includes every neuron in its body, thewiring that connects these neurons, musclecells, organs, and even its genetic code.The current model is so complete that whenplaced into an environmental simulation, itcan move like a worm.As incredible as this recreation is however,it is still not a perfect match, with replicatingthe electrical signals within the brain afuture hurdle.But it may only be a matter of time…As our ability to simulate life gets moreand more accurate, I can’t help but wonderabout our responsibilities.Over the course of making this video I causedthousands of virtual lifeforms to expire forthe purpose of education and entertainment.And sure, these ‘creatures’ were pixels,just ones and zeros.But looking at how our ability to simulateevolution has ‘evolved,’ it’s clearthat things are advancing rapidly.If one day we are able to truly render a universein a box, filled with virtual creatures indistinguishablefrom biological life, I wonder if we’llhave to start having more serious conversationsabout what we do with that power?We know how people play The Sims, after all…Perhaps one day, we’ll think of those firstpixels placed in Conway’s Game of Life ascomparable to the first cells dividing inEarth’s primordial oceans — the openingof a Pandora’s Box that marks a new chapterin the story of life.And what that chapter looks like will be upto us…Existential musings aside, the future of thissubgenre looks quite promising.There are plenty of other games that featureevolution in some way that I didn’t geta chance to talk about.If you found this subject interesting, onceagain, there are links to all the games andscientific papers discussed in this videodown below, with more information about theincredible people who are paving the way insimulating evolution.As always, thanks for watching.If you enjoyed this entry, please lend yoursupport by liking, subscribing, and hittingthe notification icon to stay up to date onall things Curious.See you in the next video.
