Absolutely. This is one of the reason why it is really hard to explain to laymans why robotic perception is hard. The conversation usually goes: why can’t you just make it do “obviously good action” when “condition” happens? And since “condition” is something they can just see it is hard to convince them the robot doesn’t have access to it in a clear and unambigous maner.
Rodney Brooks talked about building an ant robot trying to mimic the real thing. He said a real ant has hundreds of thousands to millions of sensors while the robot his team built had 150, and they only barely could handle that. The data living things gather from their environment and the processing used on it is absolutely mind boggling.
