Introduction to Artificial Life for People who Like AI

Introduction to Artificial Life for People who Like AI

. 14 min read

Artificial Life, often shortened as ALife. What is your first thought when reading those words? A brand of T-shirts? A Greg Egan novel?

For me and hundreds of ALifers, ALife is the bottom-up scientific study of the fundamental principles of life. Just as Artificial Intelligence researchers ponder the nature of intelligence by trying to build intelligent systems from scratch, ALife researchers investigate the nature of “life” by trying to build living systems from scratch.

One of Theo Jansen’s Strandbeests

What is the aim of ALife research?

Our academic field naturally overlaps with biology and chemistry, but also computer science, astrobiology, physics, complex systems, network sciences, geology, evolutionary science, origins of life research, and of course AI and animal behavior studies. Our catchphrase? “Life as it could be.”

A natural place to start explaining this field would be addressing what we mean when we say “life.” You might be thinking that the question “what is life” has been solved a long time ago: something that grows and reproduces is alive. But, that definition is far from the messy scientific reality of “life.” Many video game simulations include animals that grow and reproduce, and although you could certainly find people arguing that those are alive, the consensus is that “grow and reproduce” is not enough to define “being alive.” Likewise, salt crystals grow and cause more crystals to grow around them, but they are not alive.

But wait, what about DNA? In junior high school, you might have learned that DNA is the one common point between all life on Earth. Even if you are looking for life on another planet, DNA is the smoking gun that you should look for, right? Well, things can get more interesting than that.

One crucial thing about DNA, beyond the fact that “it is a double helix made of Adenine Thymine Guanine and Cytosine” (its substrate), is that it encodes information about cells that can be passed from parent to offspring (its function). If you focus on this function, it does not matter what DNA is made of or what shape it takes -- you can encode and transfer information using just anything you like, including 8-letter DNA or a string of 0 and 1 in a computer. Some substrates are better than others under certain conditions, but the function “information transfer” is not dependent on DNA itself. In that sense, ALife is substrate-agnostic.

So, what do we mean when we say “life?” That is what we are trying to figure out! Broadly, just as AI researchers develop programs that in various ways mimic aspects of known human intelligence without necessarily agreeing on what “intelligence” is, we seek to create dynamical systems that mimic aspects of known biological life without necessarily agreeing on what “life” is. In other words, ALifers are after a set of functions that can define life as a process and allows you to “run” it on a suitable platform under certain conditions, just like you can run a piece of software on many different hardware platforms.

These swimming creatures have a very organic feel, yet their body and DNA only have some functions in common with biological organisms.

The more general your rules and the more platforms you can use while maintaining the functions, the better. One of the current list of functions looks like this: a living organism should perform autopoiesis (i.e. the organism should constantly be “rebuilding” itself by exchanging materials with the surrounding environment), respond to stimuli, adapt to its environment, reproduce and transfer imperfect information to its offspring. Yet another definition says that life is just any autopoietic chemical system that can be subject to natural selection. The problem with our current definitions is that if they were productive enough, we would already have built a form of Artificial Life and have a consensus to call it “alive.” So, game on.

With such a general aim, ALife inevitably overlaps with many other fields. What sets ALife apart is its bottom up approach and focus on general principles that can transfer across fields: emergence, information, computation, the relationship between micro- and macroscopic variables...

This is in fact one of the most common criticisms of ALife: its focus on general principles would make it “too metaphorical,” too abstract, too diluted into other fields, making it difficult to pinpoint what kind of research “belongs” to ALife. But research in our field is not metaphorical, as was made particularly clear by the topic of the 2019 ALife conference: “How Can Artificial Life Help Solve Societal Challenges?” The field has yielded concrete advances, both practical and theoretical, as the rest of this article will show.

A short history of ALife

A video introduction to ALife, with excerpts from interviews of ALife researchers.

As a scientific field, ALife was officially born when the American computer scientist Christopher Langton organised the first ALife workshop in 1987. Langton coined the name “Artificial Life” and defined it as “the study of artificial systems that exhibit behavior characteristic of natural living systems.”

Yet the philosophy of ALife is much older than the 1980s: the idea that life is a process that can be recreated in an artificial substrate, like a software that can run on different platforms, is at least as old as the Jewish tale of the Golem, a creature made of mud that comes to life when placed in contact with the right words. There are a few documented examples of people trying to reproduce functions of living organisms in artificial media, such as French engineer Jacques de Vaucanson’s digesting duck (1739), a mechanical “duck” that could ingest food and excrete pre-loaded feces.

For mysterious reasons, De Vaucanson’s mechanical loom for weaving fabric was more successful than his duck; the loom was programmable using punched cards and even inspired Charles Babbage, inventor of the first mechanical computer with Ada Lovelace. Since then, ALife and computing never stopped influencing each other.

Left: De Vaucanson’s “digesting” duck (Wikimedia Commons) Right: Jacquard Loom programmed with punched card, by the inventor of the digesting duck. By User Ghw on en.wikipedia

John von Neumann, most famous for his contributions to mathematics, game theory, and computing, also rigorously researched the conditions for self-replication in cellular automatons that he simulated with paper and pencil. He found rules that allowed a 2-dimensional automaton to build a copy of itself based on internally stored information, and that was before the discovery of DNA! Von Neumann then became interested in the evolution of complexity. Simultaneously but independently, in the early 50s, Norwegian-Italian mathematician Nils Aall Barricelli started investigating the evolution of complex life by running the first ever evolutionary algorithms on military-grade computers.

First implementation of von Neumann's self-reproducing universal constructor (Wikimedia commons)
The "Game of Life", the most famous cellular automaton, created by James Conway in the 1970s.

ALife as a scientific community was born 35 years later with Langton’s workshop, and has since remained small but stable, with yearly ALife conferences and its own journal. The field is commonly divided into 3 sub-fields:

  1. Hard ALife, concerned with hardware, covering robotics and new computing architectures
  2. Soft ALife, concerned with software, covering computer simulations (including AI)
  3. Wet ALife, concerned with wetware, and covering chemistry and biology.

ALife has also always had strong ties with the arts, to the point that Art could be considered as a 4th subfield: ALife simulations are often exhibited in media galleries, ALife-based androids have conducted operas, and the 2018 ALife conference had an art contest.

Lenia – Mathematical Life Forms from Bert Chan on Vimeo.

ALife and AI

ALife is currently experiencing a resurgence of interest from the AI community, but the mutual influence goes way back, with Deep Learning pioneers such as Geoffrey Hinton (Google Brain) having been influenced by the ideas of prominent ALife researchers such as Inman Harvey. Kenneth O. Stanley, head of Uber AI Labs, is also a respected member of the ALife community for his insightful research on Open Endedness. Stanley’s book “Why Greatness Cannot be Planned” touches to the most canonical of ALife’s research topics: Open Ended Evolution (OEE). Stanley supervised the creation of Pic Breeder, an online collaborative OEE art project where pictures “reproduce” and evolve. He also created a new class of Genetic Algorithms used notably for neural network optimization. Stanley’s NEAT algorithm’s most striking feature is its focus on optimizing for diversity of solutions and not solely performance, resulting in solutions that beat the classical performance-only optimization approaches. NEAT was awarded the 2017 International Society for Artificial Life Award for Outstanding Paper of the Decade.

Unsurprisingly, the concept of Genetic Algorithms itself also originates within the ALife community, with John Holland’s pioneering book “Adaptation of Natural and Artificial Systems” in 1975, presenting his work on complex adaptive systems and Genetic Algorithms (a version of which was presented in 1992 at the first European Conference on Artificial Life with another ALife household name, Francisco Varela, who came up with the concept of autopoiesis). Nowadays Genetic Algorithms have their own conference, GECCO.

Karl Sims' evolved virtual creatures

The AI and Machine Learning communities may have forgotten the ALife origins of Genetic Algorithms, but they never stopped being a big area of research in our field. The most famous example is probably Karl Sims’ evolved virtual creatures, but more recent work on GA includes Emily Dolson’s work on the effect of spatial distribution on the speed of evolution or Artem Kaznatcheev’s work on computational complexity in fitness landscapes.

ALife beyond AI

Of course, ALife is not interested solely in the behavior of single creatures[1].

The part of ALife concerned with interactions between creatures can be broadly defined as Swarm Dynamics. From complex swarming patterns in simplistic simulated birds, that inspired the CGI warrior crowds in your favorite movie, to swarming “smart” slime mold made of individual fungi, to swarm chemistry, the overarching principle behind swarm research is the search for emergence. An emergent property is something that is “more than the sum of its parts”: a form of complexity that unexpectedly arises from simpler parts. Life itself is thought of as an emergent process; an isolated pile of molecules is not as “interesting” or complex as a cell made of these same molecules.

50 millions “boids”, bird-like creatures which follow simple swarming rules to form complex patterns.

How we historically went from “piles of molecules” to “living organisms” is the topic of Origins Of Life research. It is a somewhat separate field from ALife, but both have significant overlap in the questions they try to answer. Origins Of Life researchers tend to be biologists, geologists, or chemists, most interested in how life “really” (historically) arose rather than in all the ways how it could have arisen. But some are also part of the ALife community, and research what events in the Origins of Life were random and which were truly necessary; what can be reproduced in a different system and what can’t.

Artificial chemistry in particular is a field that branched out from ALife. Artificial chemistry researchers such as Susan Stepney investigate the origins of complexity, the evolution of self-assembly, and other preconditions to the evolution of life.

Despite spanning a wide range of disciplines, all the questions ALife tries to answer are linked: what are the fundamental characteristics of a living organism versus a collection of individual particles? How did these characteristics emerge? And how can we reproduce them from scratch in an artificial system?

Looking forward

In the future of my field, I see 3 big avenues. First, I think we are seeing the first signs of the next AI winter, a period where people lose confidence in AI research and funding dries out. On one hand, overblown claims steal the limelight from genuine advances. On the other hand, some AI practitioners see limits in the Deep Learning boom and in the last few years, have started to look towards ALife for new ideas.

The thing with ALife is that if you manage to build Artificial Life, and apply evolution to it, if you do it right, you must end up with intelligent systems. In my opinion, there cannot be AI without ALife first, and therefore my vision of the future is the merging of both fields.

This big merger would be part of an even bigger advance for ALife: the synthesis of Open Ended Evolution (OEE) in an artificial system. OEE is this idea that some systems get exponentially more complex with time, and that complexity never ceases to increase. Life on Earth is said to be such an open ended system. Creating OEE in a computer or in a chemical system would mean that you start from something simple, maybe a soup of molecules, or an empty simulation, and get immense complexity out of it, like living animals or maybe conscious beings.

The last, and probably biggest, event that could happen for ALife would be the discovery of life on another planet. Unfortunately, for now, we only know one type of life. It is hard to do science when your sample size is N=1. Finding any other kind of life would give us tremendous knowledge about what is important and what doesn't matter to build life.
It would completely change the way we approach the question of “what is life” and it would lead to unprecedented advances in ALife, both in theory and in practice.


To close this primer, let me share a few papers presented at the 2019 Conference on ALife (find all published papers in open access here). This purely subjective selection includes one paper on Hard ALife, one on Wet ALife, one on Soft ALife, and one Art ALife paper. For more resources to deepen your ALife knowledge, scroll to the end of this article.

Hard ALife: “The ARE Robot Fabricator: How to (Re)produce Robots that Can Evolve in the Real World”, by Matthew F. Hale, Edgar Buchanan, Alan F. Winfield and Jon Timmis

Evolution famously exploits random mutations to find acceptable optimisations to a problem. Evolutionary algorithms can be used to design both the body and the controllers of robots, but that usually happens in simulation before a solution is implemented in the real world. This last step involves a lot of human labor, and often many failures, as the real world is always harsher than simulations. This paper proposes a system that builds and evolves small robots entirely in the real world, autonomously or with minimum human intervention. With this system, you would have no “reality gap” between the world the robot evolved in and the real world: everything happens in the real world! [2]

Wet ALife: “Synthetic Biology in the Brain: A Vision of Organic Robots”, Ithai Rabinowitch

Synthetic biology is what happens when you use natural biology to make artificial systems, or alternatively when you use engineering to make living systems. Genetic engineering has been particularly successful, and in this paper Rabinowitch explains how he managed to create new synapses inside a C. Elegans worm’s brain by only modifying its DNA.

This is particularly impressive because genes are many steps removed from something as concrete as one synapse on one specific neuron in the brain of a worm! Nematodes are great models for neural network research: unlike rats or humans, they only have a few hundreds neurons and we know the exact mapping of each one of them. Because their brain is so streamlined, small modifications can cause big changes in behavior. By creating just one synapse, Rabinowitch managed to reverse an avoidance behavior into a seeking behavior, effectively programming a very simple “organic robot”. [3]

Soft ALife: “On Sexual Selection in the Presence of Multiple Costly Displays” by Clifford Bohm, Acacia L. Ackles, Charles Ofria and Arend Hintze

Some animals display extravagant attributes (like colorful feathers or elaborate dances) to woo possible mates, despite the possibility that these displays make them more likely to be eaten by predators. The paradox is that ideally, you would want your mate to be the one the less likely to die a stupid death, not more likely. So how did costly displays evolve? One of the possible explanations is the “sexy sons” hypothesis, where a female choses a “sexy” male in hopes that her sons will inherit the sexiness and be chosen by as many female mates as possible, leading to a self-reinsforcing sexy but costly loop. This paper uses a bottom-up approach to show that these dynamics do evolve in a system where some traits are needlessly costly, even if these traits have no other advantages than guaranteeing you some sexy sons [4].

Art ALife: “Edge of Chaos: Artificial Life based interactive art installation” by Vasilija Abramovic and Ruairi Glynn

Have you heard about the edge of chaos? Discovered by the same Christopher Langton who founded ALife, and originally limited to cellular automatons, the edge of chaos refers to the space between order and chaos where changing the value of a variable makes a system transition through different high-complexity phases, without being completely ordered or completely chaotic. Some studies explore the idea that brains operate optimally at the edge of chaos, and life itself is said to exist at the edge of chaos. Inspired by this concept, Vasilija Abramovic and Ruairi Glynn built a giant installation where kinetic objects are controlled by cellular automatons. The installation reacts to the presence of visitors, in a way that is neither predictable nor chaotic [5].

ALife resources

First and foremost, join us at the next ALife conference in Montreal: the theme is “New frontiers in AI: What can ALife offer AI?”

To learn more about basic ALife concepts:

To keep up to date with recent advances in ALife:

To keep up with conferences and events:

To connect:

More Art ALife to enjoy:

ALife Sci-Fi:

  1. (The definition of “single creature” itself is contentious: is it more fruitful to consider “you” as a single creature, or as a collection of human and non-human microbial cells?) ↩︎

  2. Hale, M.F., Buchanan, E., Winfield, A.F., Timmis, J., Hart, E., Eiben, A.E., Angus, M., Veenstra, F., Li, W., Woolley, R. and De Carlo, M., 2019, July. The ARE Robot Fabricator: How to (Re) produce Robots that Can Evolve in the Real World. In The 2019 Conference on Artificial Life (pp. 95-102). ↩︎

  3. Rabinowitch, I., 2019, July. Synthetic Biology in the Brain: A Vision of Organic Robots. In The 2019 Conference on Artificial Life (pp. 654-655) ↩︎

  4. Bohm, C., Ackles, A.L., Ofria, C. and Hintze, A., 2019, July. On Sexual Selection in the Presence of Multiple Costly Displays. In The 2019 Conference on Artificial Life (pp. 247-254). ↩︎

  5. Abramovic, V. and Glynn, R., 2019, July. Edge of Chaos: Artificial Life based interactive art installation. In The 2019 Conference on Artificial Life (pp. 493-494) ↩︎

Author Bio
Lana is an ALife researcher at Sony Computer Science Laboratories and at the Earth Life Science Institute in Tokyo. These views are her own and do not represent her employers. She received her Masters working on Good Old Fashioned AI, got bored of building smart models for dumb robots and switched to ALife for her PhD, looking to create robots that would build their own smart models. These days she works mostly on predictive coding and the origins of perception, as well as tools for fairer science communities. She was recently elected to the board of the International Society for Artificial Life. Follow her at @sina_lana.

Header from "Lenia — Biology of Artificial Life", used with permission of Bert Wang-Chak Chan. Special thanks to Amy Tabb, David Ha, Andrey Kurenkov and Atsushi Masumori for their comments pre-publication. Many thanks to the twitter folks who recommended good ALife

For attribution in academic contexts or books, please cite this work as

Lana Sinapayen, "Introduction to Artificial Life for People who Like AI", The Gradient, 2019.

BibTeX citation:

author = {Sinapayen, Lana},
title = {Introduction to Artificial Life for People who Like AI},
journal = {The Gradient},
year = {2019},
howpublished = {\url{ } },

If you enjoyed this piece and want to hear more, subscribe to the Gradient and follow us on Twitter.