Crazy idea time: real biter evolution

[golfmiketango]

What if we took all the variables that control biter decision-making and individualized them to each biter, and then made spawner code something like this:

Code: Select all

for (i=0; i < MAX_BITER_BEHAVIORAL_ATTRIBUTE_INDEX; i++)
  newly_spawned_biter.behavioral_attribute[i] = spawner.behavioral_attribute[i] * (1 + biter_mutagenic_factor *  (0.5 - drand48()))

in other words, what if biter-babies were just slightly randomized clones of their parent, behaviorally speaking, and the exact tuning of various frobs were left to chance and evolutionary forces? I wonder if this might enable biters to evolve emergent counter-strategies over time, slowly adapting to player innovations not only by becoming bigger and stronger but by literally adapting to whatever the player is doing to eliminate them.

The only problem is, ensuring biters don't just stay away from the player and mind their own business. But surely there is some way to incentiveize conflict. For example, linking spawner fecundity to tribal success at damaging player-owned objects.

The best possible outcome of this approach would be: biters adapt to player strategies with emergent counter-strategy and automatically incorporate circumstantially appropriate genetic specializations, without any explicit code to make this happen.

[Koub]

Did you have too much recreational phytotherapy lately ?
I mean the idea of a true AI is sweet indeed. But AI always comes at a cost, and your UPS/FPS would take a serious hit.

[golfmiketango]

Well the whole idea here is to avoid that problem. I'm not talking about creating some kind of neural network (although, who knows, that might be possible if it were heavily optimized or able to leverage gpgpu). Instead I'm just proposing (this word is too strong; it's not a proposal so much as a wild speculative outburst which I thought might start certain gears crafting in the minds of the developers) that the existing global frobs become individual biter frobs, semi-randomly inherited through biter parthenogenesis.

Biters are already a kind of NPC. There is already an engine to help them form and execute plans, for example. So factorio already has biter AI. But in the current situation, all biters of a given type are created equal, which leads to a very predictable, stable outcome, which in this case is not terribly conducive to interesting gameplay. I'm simply wondering if some of those variables used by the existing AI are too carefully curated and better left to chance.