There's an interesting discussion on Hacker News right now about a cellular automata scheme called Wireworld.

https://news.ycombinator.com/item?id=12047338

Here are a few of the interesting links from the thread:

• https://en.wikipedia.org/wiki/Wireworld
• http://www.quinapalus.com/wi-index.html - wireworld computer
• http://dested.com/projects/wire/ - javascript implementation of wireworld computer
• http://nicmalan.com/wireworld/ - interactive online wireworld system
• http://golly.sourceforge.net/ - cellular automata system with wireworld implementation
• https://github.com/martinkirsche/wired-logic - image-based construction of logic circuits

Due to its easily interpreted semantics, it strikes me that Wireworld is a good system for exploring artificial life. I think Wireworld models should be relatively easier to "disassemble" for analysis than some other simulation alternatives.

The ability to analyze artificial life models may prove to be incredibly important. By way of analogy, neural nets are extremely difficult to analyze once they have been trained; it is not easy to know "how" a neural net knows what it learned. Likewise, fMRI recordings can not be analyzed to figure out what somebody is thinking - yet - despite all we've learned about neuroscience so far.

Wireworld might be sufficiently powerful while being analyzable. Wireworld has at least two interesting properties: 1) human-understandable semantics; and 2) sufficiently complex, Turing-complete operation.

As a final thought, I think genetic algorithms might be particularly interesting in the context of Wireworld. Intuitively, the likelihood of randomly constructing interesting simulations with Wireworld seems higher than with other simulation environments (e.g. other cellular automata). This is potentially a step forward: a complex problem space can be explored using an evolutionary search process without sacrificing the ability to analyze and interpret the results.