ConwayLife.com

I am newbie to Cellular Automatas (CA), find them very intriguing and want to learn more about them.

I find there are already lot of simulation type solutions for practical problems eg. Montecarlo simulations, agent based simulations etc.

I find all cellular automata simulations can be done in agent based simulations too.

My question is, what is the unique awesomeness that CAs bring to table ? Do CAs reduce/simplify the solution development time drastically ? or do they provide unique insights that are hard to grasp in other simulation models? if no one bothered about CAs will we be poorer in knowledge (understanding of physics/math problems), won't we just get along fine with other type of simulations? or do CAs are simpler and special case of agent based simulations (its power is its simplicity and visualization) and hardly CAs bring anything new ?

Can someone point out cellular automata solutions which are much superior to other simulation solutions?

I can think of few things where CAs shine over other traditional simulations:
a) Visualization
b) Simplicity
c) Demonstrating emergence

I would like to hear more from you guys.

TL;DR version:
If some one says 'CAs are simply rehash of agent based simulations', how will you educate him CAs are not mere simplified rehash of agent based simulations but lot more!

Thanks!

From my understanding, cellular automata are a very simple precursor to the concept of agent-based modeling. It's rather pointless to ask what would happen "if no one bothered about CAs" because Von Neumann and others progressed through computational models from CA to more involved microsystems.

So yes, agent-based models are probably superior in their scope to CA, provided someone tailors them to solving a practical problem. Pedagogically though, I feel the concept of emergent phenomena is most obvious and aesthetically pleasing in CA. Game of Life in particular persists just from how simple and visual it is, that for me is what drew me in in the first place.

I have read about conversations between Stephen Wolfram and Richard Feynman regarding the possibility that the underlying computation driving our observable universe is a CA. It is an enticing model simply because it allows for every particle to carry minimal information with it and act locally. I believe a recent book from Wolfram tries to expand on this even further? (This is only information I heard secondhand from a friend)

To sum up:
CAs: Fundamental, aesthetic model. Fun to play with.
Agent-based models: Useful when tailored to a problem. Like a CA with extra features.

(Of course, anyone else feel free to weigh in!)

yeah... CAs could be seen as "simulation primitives", the very bare bone things you need to perform a simulation. Add little bells and whistles we get Agent based simulation.

I see CAs as nice simple data structure for computing (just like queues, stacks and trees), using proper data structure enormously helps in algorithm designs improving their elegance and efficiency.

Personally, what I find impressive with CAs are, just few innocent looking simple rules can result in unbelievable complexity and novelty. This is against our natural intuition of 'complex creator' creating lesser complex objects (i.e man creating a pot, god creating world etc.). Every time a CA runs I find my intuition crumbling screaming "no way..." and the results pleasantly amusing me with "yes way ...". Hardly any agent based simulation has managed to bring this in your face inversion of my intuition! Agent based simulation has always felt like huge factory running with cogs but CAs seem to capture the essence of 'complex things from simpler things' much elegantly!

lifeisagame wrote: what is the unique awesomeness that CAs bring to table

CA is not simulation tool, at least not that good. It's somewhere between computer science and engineering, and it teaches the basics of computational theory in pretty original way.

I like CA because of the designs that come around from basic "searchable" components.

In general in CA we have two major areas:
1. Basic component mining.
2. Construction of high level designs using those components.

What brought my interest back to CA after long break is Gemini. Now I'm also focusing on search tools, as a programmer this is very fun part of GOL.

simsim314 wrote:

lifeisagame wrote: what is the unique awesomeness that CAs bring to table

CA is not simulation tool, at least not that good. It's somewhere between computer science and engineering, and it teaches the basics of computational theory in pretty original way.

Every now and then there's a way to set up a simulation problem as a multi-state CA, that seems just about as elegant as any other possible solution. I guess I'm thinking of things like solving mazes.

A stranger example is generating prime numbers using a two-cell starting pattern.

Come to think of it, electrons are cheap: here's a copy of Adam Goucher's pattern to generate the sequence of prime numbers. Prime-numbered dots head northwest at light speed; adjacent integers are two cells apart.
The above pattern can be run in Golly after you copy the following rule table text and do a quick File > Open Clipboard:

dvgrn wrote:Every now and then there's a way to set up a simulation problem as a multi-state CA, that seems just about as elegant as any other possible solution.

If we open this issue. I was thinking about such question: assuming we have "molecular CA" in which each generation takes exactly O(1) time. That means assuming we can make "molecular computer" simulating some CA rule.

Can we build parallel computer that will work faster than the current computers? For now all CA based computers are sequential, while CA itself is very extremely parallel system. Can we come up with a design of computer, that uses say 2-3% of computational capacity of CA grid?

Article about molecular CA.