Basic Behaviors evolved rather than unified AI

Suggest a beneficial feature, or idea's for the future.
Post Reply
docsy
Posts: 75
Joined: Sat Jul 29, 2017 8:50 pm

Basic Behaviors evolved rather than unified AI

Post by docsy » Wed Feb 06, 2019 12:13 am

Something that's been iffy about the game lately for me, is the grazing vs trees debacle. creatures tend to just sorta choose randomly based on a unified AI probability, it seems to me. I believe this because even if I turn grass efficiency to 0, creatures still see it as a viable option.

I've been thinking, behavioral biases should probably evolve too. Not complex, new things, but simple "preferences". for example, a creature should be able to evolve a preference for food underwater, or trees rather than grass, effecting the probabilities of them choosing either food source.

Even if one food source is generally better than the other, when competition is high, it becomes worth it to try other foods. So creatures that evolve different biases than the general population can actually survive and thrive. And once they push into their bias and the related niche, they will develop traits along the lines of form follows function.

This will also help with some odd bugs that occasionally happen, where a creature will only eat grass until the grass is depleted, and then EXPLODE with tree-eating. In theory, creatures would evolve the preference that maximizes things from the start, a ratio of "choose grass:choose trees"

User avatar
Quasar
Site Admin
Posts: 1730
Joined: Tue May 29, 2012 2:04 am

Re: Basic Behaviors evolved rather than unified AI

Post by Quasar » Wed Feb 06, 2019 3:09 am

docsy wrote:
Wed Feb 06, 2019 12:13 am
Something that's been iffy about the game lately for me, is the grazing vs trees debacle. creatures tend to just sorta choose randomly based on a unified AI probability, it seems to me. I believe this because even if I turn grass efficiency to 0, creatures still see it as a viable option.
Well that shouldn't happen. Will have to look into that. Maybe the creature's decision making process doesn't factor in grass efficiency...

My goal with behaviour is to make it rational, in the sense that creatures behave in ways that match their stats. Herbivores eat vegetation, carnivores eat meat, omnivores eat both.

I want to avoid 'personality' genes where I can, because a) they're invisible and difficult to perceive as they evolve, and b) they serve as a barrier that slows down the in-game evolution, forcing creatures to evolve along two axes at the same time to take advantage of a new opportunity. There's no real point in allowing carnivores to prefer trees over meat: it would just make them act in ways counter to their own survival.

Generally speaking, though, grass should be on a different axis to tree foliage and especially to fruit. The current meat-veg axis doesn't allow for specialized grazers vs specialised browsers, and that's a problem that needs to be addressed.

User avatar
counting
Posts: 176
Joined: Fri Apr 22, 2016 1:30 am

Re: Basic Behaviors evolved rather than unified AI

Post by counting » Mon Jun 03, 2019 2:38 pm

Quasar wrote:
Wed Feb 06, 2019 3:09 am
docsy wrote:
Wed Feb 06, 2019 12:13 am
Something that's been iffy about the game lately for me, is the grazing vs trees debacle. creatures tend to just sorta choose randomly based on a unified AI probability, it seems to me. I believe this because even if I turn grass efficiency to 0, creatures still see it as a viable option.
Well that shouldn't happen. Will have to look into that. Maybe the creature's decision making process doesn't factor in grass efficiency...

My goal with behaviour is to make it rational, in the sense that creatures behave in ways that match their stats. Herbivores eat vegetation, carnivores eat meat, omnivores eat both.

I want to avoid 'personality' genes where I can, because a) they're invisible and difficult to perceive as they evolve, and b) they serve as a barrier that slows down the in-game evolution, forcing creatures to evolve along two axes at the same time to take advantage of a new opportunity. There's no real point in allowing carnivores to prefer trees over meat: it would just make them act in ways counter to their own survival.

Generally speaking, though, grass should be on a different axis to tree foliage and especially to fruit. The current meat-veg axis doesn't allow for specialized grazers vs specialised browsers, and that's a problem that needs to be addressed.
Find my way back to the forum from the steam community hub.

I have tested tree only island by setting the meat efficiency and grass efficiency to 0, and I can confirm even at 12.0.10 creatures will still target sources that have 0 efficiency. And I've also tested grass only island where I set tree density to 0, and tree efficiency to 0, both carnivores and herbivores will target meat and grass occasionally. One time by accident I restart and forgot to set density to 0 (but tree efficiency is 0) and herbivores will still target trees once they exhausted starting location grass, and then they will keep targeting no-nutrition trees and starve to death.

In another test I wanted to compare the amount of energy a creature can get to different sources, by setting only one type of food source with non-zero efficiency at a time. The combo of starting diet 1 (carnivore) and grass efficiency 1 (others 0) with 0 tree density, yields a strange starting behavior, where all the starting creatures were just frozen and refused to do anything, just got tired and then sleep and died, not even attacking each others, or tried grass, only when I added tree density not to be 0, then the creatures of diet 1 start to move, even though tree efficiency is set to 0.

In formal test of litter sizes and pregnancy length, I notice carnivores with diet even at 1 will eat grass and trees even though they don't give them lots of energy, but is a crucial part of their survival. They need to sustain a period where most of their surrounding creatures are all their own species, since herbivores usually came or grouped in packs. Hence they need some hunger stopping energy to sustain them. And they eventually evolved to be ambush type hunters, where they will stay in fertile ground, near a tree, eating a lot of grass and force the herbivore packs to targeting the trees, hence carnivores can easily get a target without travelling far and chase. They became a type of lazy hunter. So life always find a way to utilize what they can get, even for behaviors don't make sense in the real world.
King Menander asked Master Nagasena: “I have questions.”
Master replied: “What questions?”
King: “I already asked.”
Master: “I already answered.”
King: “What is THE answer?”
Master: “What is THE question?”
King: “No question.”
Master: “No answer.”

User avatar
White parrot
Posts: 1819
Joined: Wed Sep 12, 2012 4:42 pm

Re: Basic Behaviors evolved rather than unified AI

Post by White parrot » Mon Jul 01, 2019 1:09 pm

Interesting behaviours...
At this point, we shouldn't be surprised by anything nature does. She's like a meth addict whose drug-fueled rampages unfold in slow motion and span millions of years.
Silly Otter wrote:Welcome to the forum.
Please ignore the cultists.

User avatar
counting
Posts: 176
Joined: Fri Apr 22, 2016 1:30 am

Re: Basic Behaviors evolved rather than unified AI

Post by counting » Tue Jul 02, 2019 12:29 am

White parrot wrote:
Mon Jul 01, 2019 1:09 pm
Interesting behaviours...
There are tones of interesting emergent behaviors once I learned how to modify starting carnivore/herbivore gene types, litters, pregnancy and make their species thriving for hundreds of generations, like lair forming (groups of carnivores simply hang out and maintain territories which increase the population density allowing mating easily and waste less with fewer leftover corpses, etc), omnivore hybrids, tree hopping herbivores, etc. The current game has far more potential than simply always evolving to monotone herbivores all over the place. I can now even create island hopping hybrid species that periodically invade as predators and prey species evade/escaping and adapting to extreme conditions. I feel when the terrain editor is included in 0.13 the possibility for creating custom geological conditions, there will be even more surprises.
King Menander asked Master Nagasena: “I have questions.”
Master replied: “What questions?”
King: “I already asked.”
Master: “I already answered.”
King: “What is THE answer?”
Master: “What is THE question?”
King: “No question.”
Master: “No answer.”

User avatar
magmacube_tr
Posts: 880
Joined: Wed Jul 11, 2018 8:01 pm
Location: Now on the depths of the once-volcanic Hasan Mountain

Re: Basic Behaviors evolved rather than unified AI

Post by magmacube_tr » Wed Jul 03, 2019 12:11 am

counting wrote:
Tue Jul 02, 2019 12:29 am
...I feel when the terrain editor is included in 0.13 the possibility for creating custom geological conditions, there will be even more surprises...
Proceeds to make a penis continent with penis creatures

What were you saying :?:
Your everyday cube of molten rock.

User avatar
counting
Posts: 176
Joined: Fri Apr 22, 2016 1:30 am

Re: Basic Behaviors evolved rather than unified AI

Post by counting » Wed Jul 03, 2019 2:57 am

magmacube_tr wrote:
Wed Jul 03, 2019 12:11 am
counting wrote:
Tue Jul 02, 2019 12:29 am
...I feel when the terrain editor is included in 0.13 the possibility for creating custom geological conditions, there will be even more surprises...
Proceeds to make a penis continent with penis creatures

What were you saying :?:
Naturally evolved swimming cold blooded spider sharks that hunt in pack along shorelines of hot desert islands for 2-leg scaly triceratops on a climate extreme continent along with dozens major species.

Image
King Menander asked Master Nagasena: “I have questions.”
Master replied: “What questions?”
King: “I already asked.”
Master: “I already answered.”
King: “What is THE answer?”
Master: “What is THE question?”
King: “No question.”
Master: “No answer.”

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest