>>520150899
>It took 5 minutes to put 2 glasses and a fork into a dishwasher.
All of the demos for this kind of task they do various tricks, where they might use plastic instead of porcelain and they never do it live, it's always a carefully rehearsed video that they can just not include any mistakes it makes.
Now to be fair, the benefit of these robots is that even if they were slower doing things like dishes, laundry, general cleaning like wiping down surfaces, putting toys away and so on, is that this thing could be doing all this while you're at work or shopping or just some other task.
But at the moment none of them are there. Things like the warehouse jobs, boxes are a lot simpler, they have barcodes and so on. So even some humanoid robot doing something like stacking boxes it isn't the same "fine motor" controlled task like holding a cup and giving it to a person or folding clothes or something like that. The demos are all using tricks like the robot is remote controlled or replaying human motions back.
I think it could be like the self driving, that 90% of being on the road (like on a highway going straight) isn't hard to automate. It's that 10% of things where a human knows what to do because they have life experience or "common sense" to know what to do. A good example is Tesla's self driving got freaked out by a bicycle on the back of a car. It detected bicycle alright but it thought it was flying and also going across the highway. Now the "solution" was to invent a new category of object: car+bike, but that's not a proper solution. And the many other edge cases like that are why it's never quite there yet. That said I can believe Elon saying "next year" because he saw progress and assumed it would be done, but he's been saying "next year" for many years now.
I think these household robots will have similar "edge cases" in the real world (that are conveniently not in the demos or nobody has even thought of them), it's "fake it until you make it"