All is based in “robots can’t see outside text” = robots can’t see reality, but I think they can be programmed to see “outside text” and take data from their “own eyes”.