Expanding auto-evo into the multicellular stage scares people. I think the main reason is the combinatoric magnitude of multiplying the number of ways you could add a cell times the number of organelles you could add to a cell (gonna feel strange not calling it “a microbe” for a while). This, fortunately, is a matter that mutation strategies were conceived to address.
Here’s an example set of mutation strategies that can make a wide variety of creatures. Lets start with this simple starting clump:
-
Differentiate forward: make a new cell type, mutate it once, place it on the front of the central column
-
Differentiate Backwards: As before, but you add it to the bottom of the central column:
-
Grow Forwards(cell_type): For all instances of that cell type, if the front-left and front-right positions are empty, but another instance of that cell in that spot:
-
Grow Backwards(cell_type): As before, but with the rear-left and rear-right spots:
With these four rules, here’s a few species I can make with five mutations:
-
Differentiate forward, Differentiate backward, grow forward (blue cell)
-
Differentiate forward, grow forward(red), grow forward(red)
-
Differentiate backwards, grow forwards(red), grow forwards(red), grow backwards (yellow), grow backwards(red)
These shapes can be made by not iterating over possible positions in the grid, only iterating over mutation strategies. That gives you 2 + 2 * the number of cell types possible moves to try, with the differentiate strategy probably inherently having some subtypes that involve special differentiating mutations in the new cell (but not every miche has every mutation as a mutation strategy). This set of rules also happens to force all species to be bilaterally symmetrical, and won’t permit, for example, this species to emerge
It’s possible to limit (or eliminate) these restrictions with enough mutation strategies, but I would actually discourage it. First off, this might result in some really cursed looking creatures, but more importantly we have a much better source of originality: the player. Even a small change a player makes early in multicellular changes the set of possible designs for that run, most of which won’t look at all like a copy-cat of what the player actually made. As species start having more shapes and taking on “personality”, making sure the same species don’t show up game after game will become more important. Remember: the goal of mutation strategies is not to capture all possible actions in the editor, or even to ensure that any body plan can be made eventually, but to ensure a wide enough variety of potentially viable body plans to create diversity.
So what scares me in multicellular auto-evo? The selection pressures. In microbe stage, placement of organelles doesn’t matter most of the time, and microbes are are roughly circular exempting some very long line cells that don’t evolve until very late in microbe stage, if at all. In multicellular, cells of different types are on opposite parts of the body, and geometry starts to matter. Grasping tentacles are very advantageous (and awesome), but if they are not near the mouth, don’t pull things towards a mouth, or are surrounded on all sides by a hard shell, they do nothing.
Clever use of mutation strategies might help us a little here: the differentiation strategy that mutates a cell to have pulling cillia might only branch off from a cell that can engulf, for example. Simple geometry rules can also help: If you can engulf, but all six adjacent slots have a cell in them, then that doesn’t matter very much. In many (particularly heterotrophic) cases, however, I expect things to get tricky, and we will likely need to be smarter.
I think multicellular is the right time to start bringing up the idea of miches not directly reflected in the gameplay itself. For example, drought resistant plants are important in an alien biosphere, but we might never actually implement droughts in-game, because we decide generations aren’t lasting that long in later stages for example. Particularly while we still work out how gameplay really works in multicellular, I would suggest starting out with a few “silly” miches just to get some basic designs on the field, and start playing with and against them:
- a “snowflake” miche that rewards maximizing surface area
- a “blob” miche for minimizing surface area
- a miche for maxing out single-target damage (without starving to death)
- a miche for the widest “mouth” (vaguely defined)
As other parts of multi-cell solidify, a couple of these might solidify into something backed by science and or gameplay, and others will get removed. With our dreams of getting multicellular done much faster than we did microbe, however, I don’t think we have time to wait until the gameplay of multicell is fully fleshed out before scaffolding for auto-evo is built, nor do I think we will be able to try out gameplay very well until other species are forming into anything interesting.
