I was sort of hoping the 'toothbrush test' would involve the robot being able to clean your teeth... Years ago my father worked in robotics. The product was a

I was sort of hoping the 'toothbrush test' would involve the robot being able to clean your teeth... Years ago my father worked in robotics. The product was a very simple robot that took a piece out of a die-casting machine and placed it on a conveyor belt. There was nothing to program of any complexity, this arm had a fixed trapezium with different length sides that performed all the required motion, some sines/cosines/tangents involved to get the fixed path right. Most of the bits for the robot came from the R+S components catalogue, so usual industrial control gear, all 'tried and tested' with warranty, not 'home made'. However, did it go well? The robot only had to do one job right but it rarely did. It probably only ever was one robot, customers were not exactly two-a-penny for this machine, even though it was in theory a very good machine. It would require on-site attention every week and some weekends. There was no fundamental design flaw, the heat had been considered, but it was a constant war of attrition on the reliability front. I would have to ask how that ended, all I know is that it was doomed, just not an economically viable adventure. Yet this was the simplest of robots for the simplest of tasks, solving a known problem that seemed to make sense to solve. The main business survived the venture into robotics (sold to the same customers) but a loss had to be accounted for, people were laid of and the industrial unit vacated. A lot of hard work was put into it and the whole company was behind it. But it failed, reliability and support just could not be done economically. So what are robot arms to be for outside of manufacturing, where Google wants to be with their customers? If Google did get into the sales of these robot arms and 'learn along the way' there really is a risk that support for customers, 'beta customers', might be as all-consuming as it was for my father with his very simple robot. Then there are no resources to actually develop or iterate the product. You need to have the original creators debugging and fixing the robots in the field and there is a high likelihood that this work would take 10x the amount of available resources making actual development nigh on impossible. Much better to take the long route and get the product right before bringing it to market and avoid the trap. Building robots is okay if that is your core business or if you have some massive factory that needs them. Here the cost benefit analysis can be done, things tested and deployed complete with a team of engineers to keep everything running. If selling adverts is your main business then maybe not.