You could almost say that the “help” that AI can give us is almost infinite and depends on people's creativity. And this tool helps robots become increasingly “intelligent” and capable of learning on their own various ways to overcome problems.
Researchers at Stanford University have created algorithms that are applied to a vision system, so that robots can overcome different difficulties and many of the obstacles found in their path.
And for their tests they used two standard robots (dog type) to which they added their AI system.
The results are spectacular, and in the video you can see how the robots overcome different obstacles, and when they fail at one of them, they continue trying and learning from their mistakes to end up overcoming it.
A very interesting accomplishment that can have direct applications in the real world.
EN 
ES