The above could be true if the robot’s controller were hand-coded. But if the robot was trained by RL, then we know there is no internal map, no “if this then that” logic, and no internal route optimizations. But are there any end-to-end RL-trained robots?
I think Tesla got it close to right. Last week, I took a 1,300-mile road trip in a new Tesla with the latest software. I let the car do all the driving. It used the rigid map optimization only for the high-level routing, and then all of the execution was clearly analog network RL trained. I think this is “best of both worlds”. We even were able to dive on one-lane mountain roads that did not have white paint at night, and it handled the case where another car is going the other way. Our car found a way to let the other car by. This was all done with RL.
So my point is… The robot can be very human if it uses millions of hours of human data to train a model. I think the search engine results are not quite up to date.
I am not trying to sell Teslas, but I do think it serves as a good example of what the current state of the art is.
BTW: This book is good if you know little about Reinforcement Learning. It is “The Bible” in its field "Reinforcement Learning:
An Introduction”, second edition, Richard S. Sutton and Andrew G. Barto