Concepts
Background on the ideas Emergent Life is built out of: artificial life, emergence, evolutionary search, neural controllers and the ecology of a world organisms can change. These pages explain each concept on its own terms first, then say plainly how it is used here and where this project falls short of it.
Explore documents this implementation, with the real variable names, dimensions and numbers. These pages are about the wider ideas, and are useful whether or not you care about this particular world. Where a concept page makes a claim about Emergent Life specifically, it links to the Explore page that carries the evidence.
Foundations
What is artificial life?
Life as it could be, rather than life as it is. The questions the field asks, the three strands of soft, hard and wet, and what it is not.
What emergence actually means
A usable test: is the behaviour written down anywhere in the rules? Weak versus strong emergence, and why it is hard to demonstrate honestly.
Cellular automata and the Game of Life
Four rules, and gliders, guns and universal computation come out. Rule 110, von Neumann's self-reproducing machine, and where the field started.
Digital organisms: Tierra, Avida and Polyworld
Parasites nobody wrote, a Nature paper on how complex features arise, and the trade between an abstract world and an embodied one.
Evolution and search
Genetic algorithms
The five-step loop, the exploration and exploitation dial, and the place where the method stops resembling evolution: the fitness function.
Fitness landscapes
Peaks, valleys and neutral networks, plus the three assumptions the metaphor hides and why this project does not use the term.
Open-ended evolution
Why biology keeps producing new kinds of thing and simulations stop. The major transitions, and an honest account of where this project stands.
Niche construction
Organisms change their environment and descendants inherit the changed version. What that does and does not license you to claim.
Brains and learning
Neuroevolution
Evolving a network instead of training one. Why it works without gradients or labelled answers, and what that costs in sample efficiency.
Recurrent neural networks
The loop that gives a network a past. Leaky state, the vanishing gradient problem, and why evolutionary methods sidestep it.
Hebbian learning
Fire together, wire together, why the plain rule is unstable, and how eligibility traces connect an outcome now to a decision several steps ago.
What is a neural network?
Numbers in, numbers out, arithmetic in between. The only interesting question is where the weights came from, and here they were inherited.
Do neural networks think like brains?
No, and the places the comparison fails are the interesting ones: spikes, chemistry, and how the weights were actually arrived at.
Is evolution random?
Mutation is random, selection is the opposite, and evolution is what you get running one through the other. With a case you can watch.
What survival of the fittest actually means
Fittest does not mean strongest, and the phrase is nearly circular. What a world with no fitness function at all reveals about it.
Can complexity come from simple rules?
Settled since 1970. The open question is whether it keeps going, and what nine months of one world does and does not show.
Is artificial life actually alive?
Every behavioural property on biology's own checklist is here and measurable. Whether that makes it alive is a question about the word.
What is life?
There is no accepted definition and biology has largely stopped expecting one. Why that is a discovery rather than a failure.
Could a simulation be conscious?
No observation would settle it. What can be said instead, and why the absence of any reward signal here is the part that matters.
Can a simulation suffer?
Organisms starve and die here thousands of times a day. The honest answer, and what it does and does not rule out.
Free will in a deterministic world
Every creature here is fully determined, and each one still decides. Both are facts about the same system, which is most of the argument.
Explore the implementation
The technical layer: how all of this is actually built here, with the real numbers generated from the code rather than typed in.