>    Great robot dance video! How would you respond to a student that asks
>    "Is this AI?".

I agree with Michael Littman that the really impressive thing here is
the dynamic control of the robots' bodies.  When you see the Atlas
humanoid balance on one leg and raise the other high in the air, or
rhythmically shift between standing on one leg and standing on the other
-- things that people can do unconsciously -- it's important to realize
that this is a result of millions of computations per second by
sophisticated control algorithms that took years to develop.

Here's a video that explains this in more detail:
  https://www.youtube.com/watch?v=lXZ6y3lMymM

Many people working in this space use reinforcement learning to train
their robot controllers.  As of 2018, Boston Dynamics did not, but their
CEO, Mark Raibert, said that it was likely they would begin using
machine learning in their future work.  Here's a 2019 paper that talks
about the use of machine learning in robotics and notes that Boston
Dynamics is often assumed to be using this technique, but has not done
so in the development of Atlas:
  https://arxiv.org/pdf/1906.01868.pdf

There is room for a lot more AI in the dance world.  The current video
was probably made using a pre-specified dance sequence for each robot.
The dynamic controller made only tiny adjustments to maintain balance.
The robots could not hear the music and were probably not even aware of
each other's presence, or the presence of a camera operator.  They just
"went through the motions".

That was good enough to make this video, but suppose you wanted the
robot to be able to dance to any similar song by listening to the music?
Suppose you wanted it to dance with a human partner, alternately facing
them, facing away from them, following them around the stage, etc., and
taking care not to crash into them?  That would require computer vision
(which all these robots have) and new reasoning algorithms that could
figure out what the human was doing and was likely to do next.  That's
real AI.

Although DARPA doesn't fund robot dance competitions, the Atlas robots
that participated in the most recent DARPA humanoid robotics challenge
were doing some difficult things, using lots of AI algorithms to drive a
vehicle, plan trajectories to walk around obstacles, and operate
machinery.  But they weren't interacting dynamically with other
unpredictable agents the way humans do when they dance together.  We
still have a long way to go.

-- Dave