One of the few useful resources I find on Facebook is the Bio A.I. Group started by a colleague I met through an AGI (Artificial General Intelligence) initiative.
Today, he shared an article from Stadivm on a multi-use cleaning bot meant for home/hospital use which will, if all goes as intended, disinfect your sheets and kill germs. It immediately made me think of Max. Imprisoned in my basement storage is a bot that was not. Max the robot came into my life a few years back, during the first flush of my excitement over all things bot. He was going to keep my floors pristine and reduce my workload and free us up to do other, better things. Sound familiar, so far? What Max actually did was move forward, hit an obstacle, and trace his path back on the same line backwards. Until he hit the next obstacle. Then, the whole thing started all over again. So, at the end of 2 hours; I had pristine lanes through the larger floor, dust balls swirling just past the reach of an oblivious Max who only cared about his straight line mission. In the worst cases, he mis-identified large debris as a wall or furniture, obstacles he is programmed to move around, and actually, intentionally moved away from them! For his own safety, I had to finally pick him up and put him away. But, not before I spent many hours observing his decision making protocols. Hence, when I read today's article, the first thing I searched for was information on path-determination. Will this bot move any better than Max did? How will it decide where its next cleaning location lies on the sheet?
The article didn't touch upon path selection. I find it short-sighted that most cleaning bot articles and certainly, the marketing, focus solely upon what the bot is supposed to do and not HOW it accomplishes its tasks. I suspect that this is because the media and the rest of us are still fascinated by the fact that the bear dances at all, never mind how well. Perhaps this wouldn't be such an issue if the product design is done right, but in many cases, the product design hasn't made it to the HOW yet either. In my example, what Max and his colleagues should do, is sense obstacles immediately above the wheels and suck them up. Use a scanning function to visually inspect the floor and recognize non-floor objects and then suck them up. As a starting point, good bot design could begin with emulating the best in class human decision making in similar circumstances.
As for the AI Cleaning Robot, if all it will do is suck holes into the one location on the sheets it's managed to navigate to, I'd rather sleep with my germs!