This video is about evolution. Not the biological evolution you’re familiar
with, where living organisms adapt to their environment. No, this is about virtual evolution. About discovering the extent to which video
games and other simulations can shatter the natural order and let us experiment with the
building blocks of life itself. …And the extent to which they can make us
laugh. The ambitious dream of using computers to
simulate life is an old one, and you might be amazed at just how far the technology has
come. So, for this entry into the archive, we’ll
take natural selection into our own hands, and see how far we can push this concept before
things get out of control... Our story begins in 1970, when British mathematician
John H. Conway unveiled a game simply called ‘Life.’ The rules are simple: within a blank void,
you can create tiny pixels called cells, which will reproduce if there are enough cells next
to them, but die out if they get overcrowded or 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 unbelievably complex chain reactions that legitimately
feel… well, alive. Pretty quickly, people discovered that depending
on your initial pattern, specific “species” would emerge with their own unique behavior. Microbe-like colonies of pixels that appeared
to move were nicknamed “Spaceships.” So called “Puffer Trains” traveled along
and left behind a trail of debris. “Rakes” left behind trails of fully functional
Spaceships. This pattern called the Rabbits starts small,
but expands for a ridiculous amount of in-game time before stabilizing. It’s a lot like watching cells under a microscope. Of course, the pixels in the game of life
aren’t actually ‘alive’ by most definitions, but they’re an incredible starting point
that show just how effective simple rules can be at creating complex emergent behaviors. It’s proven to be a game with almost limitless
potential. But we can take things further. Life Engine is a more recent game by the amazing
youtuber Emergent Garden that simulates not just emergent behavior, but natural selection. Like Conway’s game, Life Engine is all about
pixels, but here, different colored pixels have different functions. Orange pixels act as a mouth, allowing lifeforms
to 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, allowing
the smartest virtual lifeforms to avoid each other. When you begin a game in Life Engine, it always
starts with a mass of stationary lifeforms slowly expanding out from the center via reproduction. Yet each time a pixel reproduces, it has a
random chance of changing color, simulating a genetic mutation. As with real world natural selection, some
of the mutations, like movement, prove advantageous, and give that specific lifeform more opportunities
to reproduce. In almost every game, fast moving lifeforms
quickly overwhelm the stationary plant-like creatures, and then settle into an equilibrium
much like on Earth. From here, this simulated petri dish becomes
a battleground, with different species evolving weapons, armor, and evasive maneuvers to out
compete each other. A similar youtuber-created game is The Bibites,
a simulator where endearing virtual creatures compete for food. This game also works on a system of random
mutation and natural selection, with the titular Bibites getting more and more complicated
the more you play. There’s more both games than this, and there’s
a link to them, and all the games and experiments discussed in this video, in the description. There are a lot of incredible games in this
category of cell-like evolution simulators. Like the highly underrated Cute Mold by Erytau,
which lets you simulate the growth of different slime molds. Or the remarkable Clusters by algorithmic
artist Jeffery Ventrella, that renders a microworld of particle systems that begin to display
increasingly complex behaviors not due to some hidden AI, but purely as the result of
a system based on the laws of physics. Many of these games are inspired by Conway’s
Life in some capacity, but all have their own unique spin. There’s one final computational model that
demands attention. Developed by scholar Bert Chan in 2015, Lenia
is an award-winning program derived from Conway's Game of Life that uses continuous generalization
to make everything smooth and ethereal. Using high resolution patterns, Lenia produces
far more complicated types of digital life, with over four hundred individual species
discovered. Lenia patterns have displayed an impressive
number of the traits we typically associate with biological life, including bilaterally
and radially symmetrical body plans, reproduction, and even intercommunication between colonies. Chan has even organized these patterns into
a taxonomic tree of evolution, some of which are so advanced some have called them ‘Mathematical
Lifeforms.’ And sure, by most definitions these patterns
aren’t truly ‘alive’ — they’re just the result of rules interacting in unique
ways. But in a way, our universe and all the incredible
systems of life within are also the result of rules interacting in unique ways. And these simulations are only getting more
advanced… Following the path of life on Earth, it’s
time to leave the virtual tidepool of cell simulators, and take our first, very awkward
steps onto dry land. Evolution is a game by programmer Keiwan that
allows you to build digital animals by connecting joints with bones, and bones with AI controlled
muscle tendons. The game’s neural network will then try
its absolute best to get the creature to walk forwards over multiple generations, with each
attempt appearing as a little ‘ghost’ next to the current model. At first, most simulations are an absolute
mess of failed attempts, but with enough generations, the neural network will eventually get better
at moving, even if it’s not in the way you expect. And that’s because this AI doesn’t know
what ‘walking’ looks like, which is why, at first, it’s quite bad at it. Every movement it learns is through trial
and error — essentially running off its own version of natural selection. And with enough practice, these simulated
creatures can evolve movement patterns that are reasonably close to the real thing. And it’s not just walking that this network
is capable of. Depending on the design of your creature,
it might learn to roll, or hop around like a kangaroo, or even take to the skies. And there are more challenging tasks available
than walking in a straight line. You can try to make the poor creature climb
an infinite staircase, or jump over fast moving obstacles. While at times hilarious, there is a strange
tragedy in watching these innocent creatures struggle against gravity. Something about this game really captures
the unfeeling nature of natural selection, where advancement often involves countless
generations of failure. Though cleverly simplified, this is, in many
ways, the story of life on Earth. A similar game to Evolution is Evol Pedal
— a walking simulator created by Evol Games where a lifeform made up of randomly mutating
blocks learns to travel forwards. The game also has an amazing ‘3D version’
where a CGI model of a human will learn to walk by any means necessary, which can have
both amusing and unsettling results. With over 100 moving parts, it’s difficult
for the AI to stay coordinated. You can also adjust the settings to allow
for non-human movements like knees that bend backwards, which often leads to even more
uncanny sights. The game even lets you fully adjust the gravity,
and pushing the setting too far can lead to things breaking down a bit. Such is the price of glorious, glorious evolution… But maybe we’re taking too many steps too
quickly. After all, life on our planet didn’t first
evolve to move on land, but in the much more forgiving environment of water. In 2016, a group of scientists used artificial
evolution to successfully produce swimming, soft-bodied creatures. These simulated lifeforms were made of squares
of virtual tissue, with green representing actively moving tissue, and blue resenting
passive tissues. The AI found that using a combination of both
resulted in the most efficient swimming methods, producing a range of body types convergent
with various forms of life in our own oceans. And to study the relationship between swimming
and walking, some of these creatures were moved onto land, where they found clever ways
to use bodies evolved for aquatic motion to navigate a terrestrial setting. These experiments essentially created virtual
intertidal life, offering intriguing insight into the evolution of early land animals. That same year, another group of scientists
ran a similar soft-bodied creature experiment where a network evolved a body from a single
initial cell. Each evolutionary run produced something different,
but many looked remarkably close to biological life. It's worth remembering that this AI didn’t
know what a leg or a tail is — it just found a way to keep balanced that is convergent
with life on Earth. On the subject of muscle-based locomotion,
one of the most well-known experiments comes from 2013, with a 3D simulation of various
bipedal creatures. These models learned to walk through trial
and error over multiple generations. No motion capture or keyframe animation was
used, with the impressive final results the product of AI learning. The examiners then pitted this AI against
different target speeds, which led to the creatures adjusting their gait accordingly. They even put them through obstacle courses,
throwing virtual boxes at them and adding random slopes to their environments. While sometimes they’d succeed in catching
the AI off guard, the overall results were strikingly consistent. A more recent experiment presented at the
ICLR conference used an analogous method to teach different models to walk. Many of these results resemble a three-dimensional
version of one of the evolution video games, as both are attempting to simulate similar
things. But these scientific simulations are not games
as we understand them, although this technology may one day be used to make virtual entertainment
more lifelike. As amazing as all these simulations and video
games are, for me, there’s always been that faint dream of the one unified game, that
combines cellular models with aquatic and terrestrial locomotion to create the ultimate
evolution-simulating masterpiece. For a lot of people that dream once had a
name — and its name was ‘Spore.’ Released in the late 2000s, the buildup to
Spore was, to put it mildly, completely nuts. As early as 2005 demos of this hyper-detailed
procedurally generated universe simulator were being shown, and it looked incredible. There was a sense among some before its release
that this game, which seemed to merge science with infinite customization, would change
the industry forever. And then the game came out. While Spore did impress with its creativity
and ambition, a lot of people felt the game was significantly shallower and less, uh,
scientific than they were led to believe. To be clear, I’m still quite a fan of this
game despite its eccentricities — just looking at it gives me a huge hit of nostalgia. And it does explore a number of the principles
of evolution and biology in its own unique manner, which I’ve made two separate videos
about. But it certainly takes a more cartoonish approach
to the stages of life, and some features advertised like an underwater multicellular stage weren’t
in the final game. For many, it wasn’t the comprehensive evolution
simulator they were hoping for… Yet the dream of Spore hasn’t vanished just
yet. Recently, there has been a surge of new games
that seem to take some of the concepts of Spore and push them further. One game on this exciting new frontier is
Adapt by Paul Hervé. Currently a work in progress, Adapt is a game
that lets you build your own lifeform much like Spore’s creature stage, although it
has a greater emphasis on biological plausibility and finding your niche through careful evolutionary
choices. And unlike Spore, Adapt lets you build a creature
specially 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 emerging subgenre. One evolution simulator that really impresses
with its sheer amount of detail is The Sapling. Created by Wessel Stoop, this polygon-art
sandbox game lets you design every organism in an ecosystem, and then simulate their interactions
over multiple generations. And I do mean every organism — unlike most
games in this genre, you’re not just designing animals, but plants too, and even things like
algae. And the depth of these systems are absolutely
mind-blowing. When placing a plant, you need to consider
how many leaves it has relative to the sunlight, and the length of its roots, and the amount
of ground water in the area, and the soil softness, and the temperature, and the wind
strength. Different plant colors realistically grow
better or worse depending on the type of star your planet orbits. There’s a full system of flowers you can
add to plants that produce different amounts of pollen that affects reproductivity. You can also use bioluminescence in plants
to attract pollinators — or use bioluminescence in animals to mislead prey. In terms of creatures, there are flying lifeforms,
detailed customization options for aquatic life, different eyes that can see different
ranges of infrared and ultraviolet light. You can even adjust the instincts of each
animal, using a modular system of commands to 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 goliath
of a team, these single-developer games of course don’t have the same resources behind
them, but the incredible results show that there are still passionate people who believe
in this idea — this dream of a game that could 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, Thrive
is a game quite unambiguously born out of frustration with the final product of Spore. A communal project open to all who wish to
contribute, the game takes a far more ambitious and scientific approach to Spore’s gameplay
loop. In the game, you start as a single, realistic
cell in a tidepool. As you advance, your cell splits into a complex,
multicellular organism, becoming a 3D model that can be fully shaped by the player, with
different organs serving unique functions. Over the next few stages, this creature grows
in size and intelligence, eventually able to communicate using a multifaceted language
system and create technology. Until eventually the technology compounds
to the point where you ascend to a god-like level of power and can control the entire
galaxy. It’s all so cool. It also doesn’t exist. Since Thrive a crowd-sourced game, anyone
on the internet can come up with ideas for Thrive, but actually implementing those ideas
is another story. Most of the images people share of the game’s
later stages is actually art by the talented Sciocont. Right now, over a decade after the game’s
inception, the only playable stage is the first one: The Microbe Stage. And to be clear, this initial stage is very
interesting. It’s an exceedingly complex and realistic
cell simulator, where instead of eating cartoon plants, you’re balancing your Glucose, Ammonia,
and Phosphate levels, in many ways making it a game about virtual chemistry. While this attention to scientific detail
makes the game fairly difficult, it’s a truly admirable product. Another cool thing about Thrive is that while
you build and design your cell, other cells in the environment randomly mutate and adapt,
with any given game creating hundreds of interconnected microbe lineages, all with their own randomly
generated scientific name. Most people seem to be aware that the other
stages of Thrive aren’t coming for a while. But it is pretty inspiring to see so many
people drawn in by this idea, and shows that the dream of the ultimate evolution simulator
hasn’t gone extinct just yet. Before this journey through virtual evolution
concludes, there’s one last thing I have to show you. A simulation that does something incredible
not by thinking big, but by thinking small. Really small. OpenWorm is an international open science
project with the sole purpose of simulating every cell of a type of worm that grows just
under 1mm in length. That includes every neuron in its body, the
wiring that connects these neurons, muscle cells, organs, and even its genetic code. The current model is so complete that when
placed into an environmental simulation, it can move like a worm. As incredible as this recreation is however,
it is still not a perfect match, with replicating the electrical signals within the brain a
future hurdle. But it may only be a matter of time… As our ability to simulate life gets more
and more accurate, I can’t help but wonder about our responsibilities. Over the course of making this video I caused
thousands of virtual lifeforms to expire for the purpose of education and entertainment. And sure, these ‘creatures’ were pixels,
just ones and zeros. But looking at how our ability to simulate
evolution has ‘evolved,’ it’s clear that things are advancing rapidly. If one day we are able to truly render a universe
in a box, filled with virtual creatures indistinguishable from biological life, I wonder if we’ll
have to start having more serious conversations about what we do with that power? We know how people play The Sims, after all… Perhaps one day, we’ll think of those first
pixels placed in Conway’s Game of Life as comparable to the first cells dividing in
Earth’s primordial oceans — the opening of a Pandora’s Box that marks a new chapter
in the story of life. And what that chapter looks like will be up
to us… Existential musings aside, the future of this
subgenre looks quite promising. There are plenty of other games that feature
evolution in some way that I didn’t get a chance to talk about. If you found this subject interesting, once
again, there are links to all the games and scientific papers discussed in this video
down below, with more information about the incredible people who are paving the way in
simulating evolution. As always, thanks for watching. If you enjoyed this entry, please lend your
support by liking, subscribing, and hitting the notification icon to stay up to date on
all things Curious. See you in the next video.
