Boston Dynamic's BigDog never fails to amaze or amuse, but it has moved on into the slightly threatening category. It can now throw a cinder block further than seems reasonable.
The first thing to say is that, yes, that is an arm where you might expect a head. It looks odd, but where are you going to put a fifth limb on a quadruped? My guess is that the arm is intended for loading and unloading operations but from the video you can see that this is no simple static lifting device. The arm can do things you might not think possible courtesy of the rest of the robot. Software co-ordinates movements between legs and arm so that the whole body gets involved in the throwing action. Much like a human throws, only we just do it - BigDog needs an explicit algorithm.
As the description of the video says:
" The goal is to use the strength of the legs and torso to help power motions of the arm. This sort of dynamic, whole-body approach to manipulation is used routinely by human athletes and will enhance the performance of advanced robots."
If you pay attention to the way the legs and body move just before and during the throw you will see that it is a very complex maneuver.
There is no information on the software but clearly it has to adjust the position of the robot to react to the way the block has been gripped and is moving. Couple it with good sensor input and BigDog could navigate and use its arm to clear an area or load/unload its back pack.
It is interesting to note that, while many of the comments to the You Tube video are predictions of the robot uprising, this particular activity isn't really AI and the robot isn't autonomous. However, it is part of a general development of a range of robot platforms that borrow from biological systems. At the moment these systems are at the stage of demonstrating what they can do with the help of complex software control - think of this as a software autonomic nervous system. The next step is to add an AI sensory system that can plan and direct the robot. When this happens the silly comments might not be so silly.
The idea that computers play a key role in mathematical proof is well accepted today by all but the purist of mathematicians, but the latest example of proof by computer enters a new realm - a proof 2 [ ... ]