I'm talking about the software part. i.e. Building the software that will tell the robot how to walk (over any terrain), process verbal commands (Siri one early example of this), build a house, cook you a meal, etc. That is hard.
Even the software part isn't "hard" theoretically; performing balancing, walking, and such is an engineering challange. With a sufficiently large expert system and data set all problems can likely be solved with a lookup table--again, this is only limited by current technology and processor power. You see this effect when comparing older robots to new ones--people are just finally able to run known algos on hardware they can actually mount on a robot.
I'm still mulling over your other assertions... at any rate, thank you for a good discussion. :)
Even the software part isn't "hard" theoretically; performing balancing, walking, and such is an engineering challange. With a sufficiently large expert system and data set all problems can likely be solved with a lookup table--again, this is only limited by current technology and processor power. You see this effect when comparing older robots to new ones--people are just finally able to run known algos on hardware they can actually mount on a robot.
I'm still mulling over your other assertions... at any rate, thank you for a good discussion. :)