Project image
)}
$17,653
pledged of $20,000pledged of $20,000 goal
307
backers
0seconds to go
Funding Unsuccessful
The project's funding goal was not reached on Thu, April 12 2018 5:05 PM UTC +00:00
$17,653
pledged of $20,000pledged of $20,000 goal
307
backers
0seconds to go
Funding Unsuccessful
The project's funding goal was not reached on Thu, April 12 2018 5:05 PM UTC +00:00

Q&A: AI and Traits

Posted by Tom Johnson (Creator)
2 likes

For this update, I thought I would cover a few more of the questions I've received from backers and interested people in other parts of the internet.

1. How does the AI work? Can creatures make emotional bonds with each other?

There are two layers to creature AI. At the lowest level, there's a nervous system where inputs travel down a pipeline, are processed by neurons in different ways, and then are eventually sent down to control the contraction of muscles.

A step above that, there's a behavior tree which controls some of the inputs to the lower level nervous system AI. This system implements more large-scale behaviors and responds to higher level details of the creature's situation. For example, the behavior tree would choose a target swimming direction based on where it thinks it is likely to find food, and the nervous system has this swim direction as an input to guide its motion.

Both systems are customizable by the player and by natural selection, and they are basically programming languages unto themselves, so they should be capable of creating really diverse array of behaviors. For the suggestion of emotional bonds, I like the idea a lot. It may be something I will have to experiment with to see how well it works as part of the overall AI system, but I think that would add substantially to the ability of the game to generate interesting stories, the way that games like Dwarf Fortress or Rimworld do.

2. How will predatory and defensive mutations work? Can things like claws, constricting appendages, poisonous barbs, projectiles, and other such appendages evolve in the game?

Evolutionary arms races between predator and prey populations are definitely something I want to support. Some mutations like this should just naturally fall out of the way that existing systems are implemented: for example, prey AI evolving to freeze in the presence of a predator, since a moving creature is easier to detect than a still one. Other examples might include behaviors like swimming in schools, or having skin that closely blends into the surrounding area. I also hope to add additional traits, such as varying fields-of-vision and vision depth, skin toughness, carapaces, bio-luminescence, lures, bite strength, and more. Traits like these that don't come with obvious downsides will impose additional costs in terms of the energy and nutrients that a creature requires, so that whether or not a specific trait pays off for a creature depends on the details of its environment and what other creatures live there.

Thanks to everyone who asked questions, and thank you for reading.

Tom

Tom Harriss and Frederick Wrigley like this update.

Comments

Only backers can post comments. Log In
    1. Tom Johnson 2-time creator on

      I think you are right. It's a really good question. I could be getting this wrong, but the impression I got was that the genes tell a cell to divide or change type or move based on their chemical environment, and the bilateral symmetry basically comes from a bilateral symmetry in the initial chemical distribution rather than the 'instructions' directly (for example, specific hormones are different at the top and bottom of a creature, distinguishing them; this doesn't happen on the left and right sides). Biologists seemed to think there were selection pressures that favored bilateral symmetry when it came to moving (it helping to steer and to not have different amounts of drag on each side of the body), so I think it may be a combination of both symmetry being easy to make and being important for directed movement (and the fairly small population size, compared to Earth, of creatures in the game). Thanks for writing that, I think I learned a lot for the game from trying to answer.

    2. Greg on

      I wonder if it's easier for real life biology to encode, build, and control symmetrical components than anything asymmetrical, and that's why there's a strong tendency towards symmetry. Any biologists around that could clarify/correct me?

      Because if that's the case, I'd be surprised if the simulation favored symmetry as heavily - there just isn't the same kind of fundamental push for symmetry as real life has.

    3. Tom Johnson 2-time creator on

      Hi Tom! I think that it should, over time. I think that for real fish, most of the thrust is generated by their tail fins, and the symmetric fins (e.g. pectoral fins) are more about balance and steering (with exceptions in some species, of course). Since many creatures in the trailer were tested for just swim speed in any direction, there wasn't a strong force pushing for them to evolve such symmetric fins. When I test for target-seeking behaviors, I tend to see more symmetry (though still not as much as in real life). I'm still adjusting this, and I may see if I can adjust the way that the DNA is written so that symmetric body parts are more likely to appear when mutations occur. This may also potentially help target-seeking fish to evolve more quickly as well. Thanks for your question.

    4. Tom Harriss
      Superbacker
      on

      Will the creature development tend toward symmetry?