I think this perspective is now demonstrably false. It was basically the defacto position of most roboticists until a few months back. But we've already seen the cost of humanoid robots fall precipitously (even without the kind of mass production that genuinely useful home and real world intelligence would bring).

Which seems more likely, that cheap humanoids will continue to improve in dexterity, battery life, strength etc. Leveraging the vast investment in research into real world task training. Or - that we'll redesign millions of everyday objects to fit the limitations of current task specific robots?

If you answer the latter, I've got a whole bunch of mid century quirky labour saving inventions in the back room I'd like to sell you.

I can't believe someone shares my point of view. We have already many non humanoid robots to do boring stuff (underwater maintenance, stuff like that). For some reasons people tend to think that for robots to do stuff close to humans, we need human like robots.

We do not. I want a swarm of insects-like robots to fight mosquitos. A snake one that fixes my pipe. A spider-like robot that cleans surfaces. Etc.

I will suggest mix of both: some everyday tools are already built for being used with hands and legs (e.g opening a door). But indeed adding some different grip types, additional arms, etc, could be helpful.

Exactly what I came here to say. A robot doesn't have to look like a human. Just a few arms that does the job neatly is more than enough.