This field is incredibly cross-disciplinary... all I come in with is a little software, a little AI, and a lot of interest
materials science, mechanical engineering, software, AI, electrical eng, philosophy..
get up and running with pybullet
it's kind of big, and my computer runs it really slow, which isn't a good sign. Hopefully colab will be helpful for this. May need to figure out how I can put all my computers resources just on that program (or use work computer... but I don't really want to do this because if I leave work for this kind of work then I want to be fully prepared for what I may be getting into)
The example he shows is like a globe of shocks, that can protect what is in the middle (scientific equipment, valuables, a person?, etc.)
rigid beams, connected with wires/springs
designing robots is non-intuitive, which is why they use evolutionary algorithms. That means they aren't coming up with all these crazy designs with their own puzzle solving abilities
oddly, this makes the field feel much more approachable (as someone with a software background)
Create task environment
Create the robot
Create robot's brain, (neural net?)
Use evolutionary algorithm to optimize brain so the robot performs the desired task
evolutionary algorithm vs Reinforcement Learning vs classification
it is different, but it is still a form of mathematical optimization
didn't get further into this