> But I would guess, that is couple orders of magnitude easier than self driving for example. The more I think about this, the more I believe you to be incorr

> But I would guess, that is couple orders of magnitude easier than self driving for example. The more I think about this, the more I believe you to be incorrect! Identifying the correct item in the bin is hard - things are in random arrangements and orientations. They may be buried under other stuff. Packaging continually changes. Things will be in the wrong bin. Things will be packaged by the manufacturer incorrectly! Things expire but you can't rely on expiration dates, which you'd have to be able to read and parse in the first place. Anything in the bin could be extremely fragile. Some things are both fragile and heavy. The act of picking stuff up is hard. Grip too hard, you break it. Grip too soft, you drop it and break it or something else. Grip it the wrong way, you break it. Crush stuff around it rummaging, you break that. Some things must be pinched, others grasped, cradled or supported at multiple points. You break something, then what? You've contaminated who knows what around it roomba style - you can't send stuff out to customers thats covered in shards of glass or unknown liquids. How do you identify when something is broken or defective in the first place? Packaging may be damaged while the item is OK, or the item may be broken while the packaging is intact. Humans can do this because they have a model of the world with interactive feedback. Things with broken glass may sound wrong or just feel wrong when moved. The stakes are lower on average, but putting the wrong item in a cart will eventually kill someone. Consider allergies, medical supplies, the blind or illiterate. All the while, you have to do this for less than $15/hour? I can only see robots currently succeeding for hyper specific items - perhaps those that are low value and heavy for instance.