Search Unity

  1. Welcome to the Unity Forums! Please take the time to read our Code of Conduct to familiarize yourself with the forum rules and how to post constructively.

Feedback help for ML agents muscle system

Discussion in 'Physics for ECS' started by Sab_Rango, Mar 26, 2021.

  1. Sab_Rango

    Sab_Rango

    Joined:
    Aug 30, 2019
    Posts:
    121
    Hey! I noticed that all of the physics developers doing their best to make DOTS physics better and better, by
    adding new cool features.

    Yet right now, the most demanding tech is "muscle physics" for ML agents(or the future of AI in human being) is not exists for ML devs to teach their new brain algorithm in 3d physical World.
    Indeed, there is no any proper physics engine to simulate realistic anatomical physics for creature body interaction in the Universe.

    In my view, how Machine Learning for interactable Character agents should be improved by these steps:

    1. ML devs creates any algorithm that assimilates real biological neurons, including cerebral, muscle cache cerebellum neurons.

    2. test those tech in real virtual physical world, and benchmark the ever new brain algorithm with the realistic real time 3d physics.


    What I mean is that in order to feel how our brain simulation code becoming close to real biological brain, we need to test the characters in realistic dedicated 3d character physics engine, and then feel how its close to biological animals.


    There no a better way to begin to write a " biological like Intelligence " code without realistic constrained physics.





    I just want you physics devs to feel how the "anatomical physics engine" is crucial to make a hand made Intelligence code.




    In conclusion, This process is not smth toxic attention. Even I do not spend my all time for this.
    Just guys, consider about little helping to brain developers to initiate their Intelligence code by improving physics engine!
     
  2. milos85miki

    milos85miki

    Joined:
    Nov 29, 2019
    Posts:
    197
    Thanks for the suggestion! We're quite far away from what you described, but it's nice to see this kind of futuristic feedback and hoping this forum discussion will develop.
     
    Sab_Rango likes this.