The Powder Toy

TPT evolution? by perogiepro

perogiepro
perogiepro
99 / 13
24th Aug 2015
22nd Dec 2015
We are the SFPI, we are searching for mechanical life that can detect danger and think. essentialy basic AI. Our end goal is to make one that can replicate and change, so it will """evolve"""
machine intelligence experiment neuron life survival challange evolution spfi

Comments

  • msasterisk
    msasterisk     25th Aug 2015
    Well, it wouldn't need depth perception, it just needs a general idea of which direction the bad stuff is and and which direction the good stuff is.
  • Schmolendevice
    Schmolendevice     25th Aug 2015
    Except that making any neural network support 'perception of an environment' (yes I have pondered the structure of a simple thinking brain/hamster bot aware of its surroundings) and trig based depth perception is rather complex.
  • msasterisk
    msasterisk     25th Aug 2015
    You are very smart, Schmolendevice. Thats a GREAT IDEA!! :D
  • Schmolendevice
    Schmolendevice     25th Aug 2015
    Ah. But maybe for TPT's sake we can imagine that the only thing we would need to simulate a green cone cell is DTEC(PLNT)? DTEC(CONV) could be like a no no for the creature to walk towards. Having strategically placed DTEC of different radius increments can give you the option to use trig to determine the depth or relative position of an object.
  • perogiepro
    perogiepro     25th Aug 2015
    ok msasterisk ill do that. plus isnt there a way to filter green ligt (not make passing light green, only let green in ) but i think im getting it wrong
  • msasterisk
    msasterisk     25th Aug 2015
    We would use dtec, but we want it to only see green light...
  • Schmolendevice
    Schmolendevice     25th Aug 2015
    An idea of mine is to just use plain old DTEC for 'eyes.' Just use to prop edit tool to change the tmp2 to the desired 'radius' and have its output go to WIFI which is fed into the brain's neural inputs. Technically, any biological sensory processing circuit, no matter how simple or complex, could be considered as a neural network. I've watched vids of simulations with moving and eating agents that only need up to 5 neurons.
  • msasterisk
    msasterisk     25th Aug 2015
    Ok, I've solved the blind spot problem, basically all the surface area you can spare, make a light sensor. Then the only blind spots will be the red lines and the light emitters. Can't figure out how to filter the light, so it gets attracted to all reflectve things. And the light is still burning hot.
  • perogiepro
    perogiepro     25th Aug 2015
    ok ill post to the forum then.
  • Schmolendevice
    Schmolendevice     25th Aug 2015
    Basically if it takes a month to reach a breakthrough in 'biological TPT,' (hopefully not) future viewers can still refer to the original thread which does get bumped up on TPT wiki upon new posts.