>>9460663first we need stable high bandwidth neural interfaces. Then we need to have a complete revolution in how we attach things to and implant things into the body. Then we need actuators with better torque densities than human muscle, then we need a way to power these actuators. Then we need to reverse engineer the physical and neural basis for sensing touch and proprioception. Then we need artificial skin capable of sensing that last millions of cycles until failure. Pretty much in this order.
High bandwidth neural interfaces we're making progress on, but things like having a massive revolution in implanting stuff in the body is going very slow. Right now, passive implanted components are not as good as their regular counterparts.
>>9460677materials science. Almost all of the issues with neural interfaces are related to materials. Making the interface stable and not kill neurons is a pretty big one.
>>9460709The software and electrical issues are relatively easy compared to the materials issues, you don't need artificial neurons for neural interfaces. In recent neural interface tests, people were able to learn how to use the interface to control a mouse cursor in just a couple minutes
https://news.brown.edu/articles/2018/01/calibrate 
