I’ve written a lot about robotics and I can tell you it’s incredibly, incredibly difficult — and expensive — to develop robotic arms that could grab soft, easy-to-damage items like that and toss them into packaging without bruising them
and even if you can get the correct robotic arm and gripper, everything else has to be designed so that the box is in the exact right place, etc.
that final move, where the worker grabs the box flips it shot and moves it off? That’s literally a whole other nightmare for a completely different robotic system
this is why it’s comparatively easier to impress people with “AI” that just manipulates images and words – they don’t need any of this proprioception, flexibility, and common-sense reasoning about the physical world
This physical-world stuff requires massive amounts of everyday smarts
@clive
To be fair, you wouldn't toss them at all :) unboxing is hard, but the complications here are (1) not enough space for the robot (2) food safe grippers are tricky (3) she's probably implicitly rejecting gross ones that you wouldn't see in this speedrun :) @yurnidiot
@clive@yurnidiot Yeah. I shouldn't do driveby workcell design :-) but there are in-line box-closers/sealers/labellers (they'd just drop in further down the conveyor belt.) And I've worked on half-pallet workcells (intended for easy reconfiguration, ie. "we're doing a batch that we know the robot can't handle, pallet-jack the robot out and put a person in for a shift or two") that would fit there - but at least 4ish years ago couldn't get anywhere near human rates.
@clive@yurnidiot would definitely need a lot of work to get something even vaguely price-and-speed-competitive with the human. (This is how roboticists end up being in favor of raising the human wage floor - not on ethics grounds, but because robots are basically never competitive with exploited humans :-)
Add comment