Kill sensory neuron ?
Eric1212 opened this issue · comments
I seen in https://www.youtube.com/watch?v=N3tRFayqVtk that there is a kill action output.
What about if there is a kill sensory neuron ?
Let me explain, if there is kill sensory neuron and a kill action output, evolution could evolve that if an individual A detect a kill from individual B to individual C, that individual A will kill that individual B but because that individual B killed individual C than individual D will not kill indvidual A because it's also detected that individual B did kill individual C.
@davidrmiller I'm not sure it's clear and i understand that this need a lot of mutation and evolution but it's might profoundly change the actual kill you recorded but i suppose that's why you did go up to 200 000 gen...
What you think ?
Hey, that would produce some extremely interesting results. Maybe instead of detecting if another has killed directly, maybe individuals that have killed should have a number of how many others they have killed. (that way, some individuals could try to avoid an aggressive serial killer and others can go out of their way to stop those fiends)
I believe this would be a cheaper computation but I think its slightly off from the result you wanted.
I would be very interested in seeing that simulated. I'm not at a level where I can program something like this but am really curious as to how they would evolve. I hope @davidrmiller notices the sudden uptick in interest via the Youtube vid and makes more content. I highly enjoyed the video and hope he wants to make more! It's not an easy subject to explain but the way it's setup and narrated makes it very accessible.
I'm subscribing to the channel and here just in case @davidrmiller indulges us.
There are kills and then there are potentially sacrifices.
If individual B kills individual C and at the end of the generation there wasn't an improvement in the overall result then hopefully a kill neuron would minimize/eliminate that sort of action.
On the other hand, I've seen a lot of examples running here where a slow individual is going to the right and a bunch of fast individuals stack up behind him, the generation completes and there are fewer survivors than if the first individual behind the slow one sacrificed him and everyone else had a chance to get to the safety area.
However a different alternative, but not one I've noticed, is that when the fast individuals are stuck in this queue, you'd think they might go up or down one step and then continue to the right but I don't see that. If they did then no reason to sacrifice.
Alternatively, if slow individuals noticed they were blocking the road and got out of the way then results would improve overall.
Lots of interesting alternatives...
That's an intriguing idea indeed.
David - your video and code has spurred me toward finally trying to learn c++. No idea if I'll ever get good enough to make a contribution but I'm reading the code and doing online activities to try. Thanks for your generosity.